To our knowledge this is the first systematic review of the prevalence, incidence, associated factors and risk factors for TSP among the general population. We elected to study and report all these epidemiologic characteristics to present a comprehensive picture of what is known about TSP in the general population. Findings are consistent with clinical anecdotes suggesting that TSP is common in the general population, particularly during adolescence, yet interpretation of the data are difficult due to heterogeneity. Despite the large ranges identified in TSP prevalence, predominantly as a result of variability in definitions of pain, these data support the contention that TSP is prevalent among youth. Whether it should be considered as a discrete and important clinical presentation in youth and perhaps treated as such [
11], would depend on how often TSP coexists with other spinal pain conditions. Conversely, in adulthood there are insufficient data to draw a similar conclusion. Unfortunately, the data reviewed do not provide comprehensive information about the impact of TSP on function. Nonetheless, up to 10% of adolescents experienced TSP that interfered with school or leisure [
45] and TSP prevalence seemed to be highest during backpack use. Although some data have been collected regarding associated and risk factors for the condition, these are relatively scarce, highlighting the need for further epidemiologic research directed towards this condition. Most importantly, a consistent approach with respect to defining pain characteristics and reporting prevalence and incidence data is urgently needed among researchers to allow meaningful comparisons between studies.
Study design and quality assessment
The NHMRC supports a 4-point rating scale (excellent, good, satisfactory, poor) for each of the 5 essential components of a body of evidence: evidence base, consistency of results, clinical impact, generalisability, and applicability [
35]. The majority of studies included in this review were cross-sectional in design, limiting inferences about causality and prognosis for TSP. Thus, in terms of the evidence base reviewed (relating primarily to study design), it may be rated as poor according to NHMRC criteria. Prospective cohort studies are therefore required to provide a more robust evidence base for prognostic factors and the clinical course of the condition across the lifespan. Moreover, these studies would also provide important information to clinicians regarding the natural history of TSP and ultimately trajectories in certain clinical groups. Nonetheless, we suggest that the evidence presented is satisfactory with respect to consistency, clinical impact, generalisability and applicability. Generally, the method quality of the included studies was good, with only one older study being rated particularly low (4/15) [
42]. More than 90% of the studies reviewed used an appropriate study design, used appropriate analysis methods, reported results in terms of statistical significance, provided a commentary on the clinical relevance, and reached appropriate conclusions given the results presented. The most significant method quality issue identified was a lack of sample size justification, with only 2 (6.1%) studies addressing this criterion with a power calculation. However, in observational studies concerning prevalence and incidence a sample size estimate is not as critical as in intervention studies, since demonstrating a difference between groups is rarely needed. We acknowledge, however, that power is needed in order to detect relationships in epidemiologic studies. The independent reviewers also identified biases in many studies, primarily related to sampling bias. We considered a sample bias to be present if the response rate to a questionnaire was less than 80% [
70]. Finally, only 33.3% and 42.4% of the included studies presented evidence about the validity and reliability, respectively, of the outcome measures used. The comprehensive quality assessments performed in this study highlight areas for improvement in research design and reporting in the context of spinal pain.
TSP prevalence
The range of prevalence estimates of TSP in the general population was broad. Similarly, a broad range of TSP prevalence was reported in a review of TSP among adult working populations [
7]. The wide prevalence range is partially a reflection of the influence of age and gender. However, even within an age range and gender category, prevalence estimates were highly variable. For example, point prevalence in children ranged from 14–38% among males (based upon 2 studies) and 9–72% among females (2 studies). This may result from the variable operational definitions or study inclusion criteria for pain cases in cross-sectional studies. Notably, the studies on young people were often focussed on pain related to school bags and workstations which may also have influenced the reported rates. The operational definition issue has been identified as a major limitation in the comparability between prevalence studies in low back pain research [
29].
Variability in operational definitions of TSP may also influence the interpretation of TSP across prevalence periods. As highlighted in Figure
2, children had higher TSP prevalence than adults for one month prevalence, while the reverse was observed for one year prevalence. This inconsistency may partly be explained by recall period, where children are less likely to recall events over a longer duration. However, a more likely explanation may be the differences in operational definitions of TSP between studies. Data for one month prevalence of TSP in youth was sourced from four studies [
11,
50,
57,
67] while only one study was available for adult data [
55]. The operational definition for TSP in the adult study was "frequent pain in the upper back" compared to "any pain" or "pain duration ≥ 1 day" in the youth studies. Fewer adults were likely to report frequent pain as compared to definitions that were unrelated to pain frequency. Similarly, for the one year prevalence, six adult studies contributed to the data and adopted a definition of "any pain" [
13,
46,
47,
68], "pain duration for ≥ 1 week" [
51], or "pain after work" [
39], while one youth study reported a definition of "pain interfering with school or leisure" [
45]. The lower prevalence in the youth study was likely related to a definition of pain which needed to be associated with a functional impairment.
To assess the effect of study quality on the prevalence ranges we excluded the 8 studies which scored less than the mean method-quality score (< 10/15). Excluding these studies had a minimal effect on prevalence other than raising the lower limit of 7-day prevalence to 2.8% (from 0.5%), raising the lower limit of 1-year prevalence to 15.0% (from 3.5%), and leaving one study reporting a lifetime prevalence of 15.6% (from 15.6–19.5%). This finding is consistent with an earlier study which reported prevalence estimates of neck pain to be unrelated to study quality [
71]. We did not perform a similar sensitivity analysis based on NHMRC Hierarchy of Evidence rank since the majority of studies were ranked as level IV, thus over representing this type of study design relative to others.
We identified 6 prevalence periods and 4 incidence periods in the literature for TSP. Prevalence data were distributed relatively equally across the 6 periods, other than the 3 month period where only one report of TSP prevalence was made in one study [
66]. Therefore, it seems that there is not only a lack of consensus with respect to definitions of pain and inclusion criteria between studies, but also the most appropriate prevalence period to investigate. The 7-day and 1-year prevalence periods were the most commonly reported (25.8% and 22.6% respectively), consistent with an earlier systematic review of neck pain [
71]. Moreover, they are also consistent with recall periods in the Nordic Musculoskeletal Pain Questionnaire [
72], which is one of the most commonly used assessment tools in musculoskeletal research. Although recall bias may be more problematic with longer recall periods (e.g. 1-year) [
73], shorter time frames (e.g. 7-day) may miss episodes of pain. In light of evidence which supports the validity of recalling pain intensity for at least a 3-month recall period [
29], we suggest that for chronic and disabling spinal pain, recall bias is less likely to be threatened. Nevertheless, such variability in definitions renders interpretation of the data somewhat difficult, and this issue has been highlighted previously as a limitation in the comparability of spinal pain research [
27‐
29,
71]. A recent international Delphi study concluded that definitions for prevalence studies on low back pain should include, at a minimum, the site of low back pain, symptoms observed, time frame of the measure, and severity [
29]. Arguably, these same criteria should be applied to TSP studies. Consistent with the consensus of an international working party for low back pain research [
29], we recommend a 1 month prevalence period for studies investigating TSP.
Similar to findings in this review, prevalence estimates for low back and neck pain vary widely in the literature. Among the general adult population, the point, 12-month and lifetime prevalence for low back pain ranges from 5.6–28.4%, 22–65%, and 11–84% across various studies [
46,
74‐
76], while for neck pain these estimates are 5.9–38.7%, 16.7–75.1%, and 0.2–71% across various studies [
71,
77] and for TSP the 12-month prevalence ranges from 15–34.8% [
13,
46,
47,
51,
68]. Notwithstanding the wide prevalence ranges and variability of spinal pain definition, it appears that TSP may be a significant a problem in adulthood. In adolescents the point, 12-month and lifetime prevalence for low back pain ranges from 1.0–35.8%, 7.0–50.8%, and 7.0–72.0% respectively [
25] and for neck pain the 12-month and lifetime prevalence range from 7.6–13.0% and 3.0–28.0% respectively [
25]. In this review the point, 12-month and lifetime for TSP in adolescents (13–20 years) was reported to range from 4.0–41.0%, 4.2–9.7%, and 15.6–19.5% suggesting comparable significance to low back and neck pain in adolescents. Notably, the point prevalence for TSP in children was even higher (4.0–72.0%), suggesting the magnitude of the problem of TSP, in terms of prevalence, to be greatest in youth. This may be one reason why there are a greater number of studies using childhood and adolescent cohorts compared to adult cohorts. However, further interpretation of the impact of TSP should be made with due consideration to the severity and disability associated with the experience of TSP.
Higher TSP prevalence in females is consistent with general reports of musculoskeletal pain in adults [
78], adolescents [
79,
80] and children [
81]. A higher prevalence of self-reported pain among females may be due to differences in physical activity, musculoskeletal maturity, posture, endocrine and psychosocial characteristics as well as different physiological mechanisms for pain perception between genders [
82], which should be investigated.
Associated factors and risk factors for TSP
In children and adolescents, TSP was associated with female gender, postural changes associated with backpack use, backpack weight, other musculoskeletal symptoms, participation in specific sports, chair height at school, and difficulty with homework, while poorer mental health and age transition from early to late adolescence were significant risk factors for TSP. In adults TSP was associated with concurrent musculoskeletal symptoms and difficulty in performing activities of daily living and there were no studies reporting risk factors. Although the limited data describing the associated and risk factors for TSP established predominantly with bivariate analyses render interpretation of its aetiology difficult, the factors identified in this review suggest that musculoskeletal growth, biomechanical loading, concurrent musculoskeletal pain and psychosocial characteristics are important mediators. Therefore, a biopsychosocial framework would seem appropriate for conceptualising TSP aetiology among the general population who are free of other pathology.
Strengths and limitations
Strengths of this review include a systematic review method, in particular the use of an appropriate critical appraisal tool for observational epidemiological literature. Additionally, tailoring of the search strategy, via the use of broad search terms, was employed to capture studies that reported on the prevalence, incidence or risk of TSP even where these parameters were not the study's primary objective. This search approach formed the rationale of an earlier review [
71]. However, the findings presented here should be interpreted within the limitations of the review. Firstly, the studies included were generally from high income countries and therefore the data reported may not represent TSP from a global perspective [
83]. Future studies should examine whether any differences exist in TSP experiences between ethnic groups. Secondly, studies published in languages other than English were not reviewed. Thirdly, studies reporting epidemiologic data for TSP in discrete occupational groups were not included, as risk factors for spinal pain would likely be influenced by, and differ between occupations and not be representative of the general population. Finally, studies in this review involving children and adolescents were derived from samples of schoolchildren and therefore do not represent children who do not attend school, for example those involved in child labour [
84]