Risk factors for episodes of back pain in emerging adults. A systematic review

The transition from adolescence to adulthood is a sensitive period in life for health outcomes, including back pain. The objective was to synthesize evidence on risk factors for new episodes of back pain in emerging adults (18–29 years).


| INTRODUCTION
Back pain, mostly in the lower back, is an enormous global health problem in low-income as well as in high-income countries (Hartvigsen et al., 2018). It is strongly associated with activity limitations and work restrictions, and is the leading cause of years lost to disability (YLD) worldwide (Global Burden of Disease Study, 2015;Vos et al., 2012). Consequently, back pain leads to substantial economic costs for individuals, families and societies. This is seen particularly in highincome countries with highly developed healthcare systems (Lopez, Mathers, Ezzati, Jamison, & Murray, 2006;Murray & Lopez, 1996). The Global Burden of Disease Study 2010 showed an overall prevalence of back pain of 9.2% (Vos et al., 2012), with the highest prevalence found in females and those aged 40-80 years. Back pain appears mostly without a clear pathologic cause. According to the Lancet series from 2018, individuals with physically demanding occupations or comorbidities, and those with lifestyle risk factors such as smoking, obesity and low physical activity level seem to be most at risk (Hartvigsen et al., 2018). Often, modifiable and non-modifiable factors occur at the same time in complex interactions influenced by psychological and social contexts and socioeconomic status (Hartvigsen et al., 2018). The period of transition from adolescence to adulthood -"emerging adulthood (Arnett, 2000)" -has attracted relatively little attention in the back pain literature, despite the high prevalence of low back pain (up to 30%) in the age range associated with this life stage (Hoy et al., 2012). Emerging adulthood, roughly capturing the age range between 18-29 years of age (Arnett, Zukauskiene, & Sugimura, 2014) is characterized by transitions (Swanson, 2016) involving exposure to new educational and workplace environments, the strengthening and loosening of social and family ties, romantic relationships and changing lifestyles. We hypothesized that exposures and transitions in this life stage may be relevant to the risk of back pain, and in turn be impacted by back pain at this time. Reviews on risk factors for low back pain in younger populations are performed primarily on adolescents, focusing on growth and maturation (Swain et al., 2018) and exposure to sports (Trompeter, Fett, & Platen, 2017). Calvo-Munoz and colleagues (Calvo-Munoz, Kovacs, Roque, Gago Fernandez, & Seco Calvo, 2018) systematically reviewed 61 longitudinal studies on young individuals from childhood to adolescence, and found inconsistent results on associations between low back pain and demographic, clinical, biological, family, psychological, ergonomic and lifestyle factors, but consistent associations with older age and participation in competitive sports. From a life course perspective, investigating risk factors for back pain in children, adolescents and emerging adults would help reinforce the concept of sensitive periods for intervention, and may identify preventive interventions with impact throughout the life course (Dunn, 2010). To our knowledge, there are no systematic reviews on risk factors for episodes of back pain in emerging adults. Thus, the purpose of this study was to summarize risk factors for new episodes of back pain in emerging adults in a systematic review of the literature.

| 21
ØIESTAD ET Al. outcome was a "first episode" of back pain. Back pain had to be measured either during the period of emerging adulthood (age 18-29 years) or in adulthood (age 30 years and above). The risk factors had to be measured either during adolescence (up to age 18 years) or during emerging adulthood (age 18-29 years). Risk factors reported as being measured in a pain free condition prior to the outcome were considered. Only full-length publications in English or Scandinavian languages were included. Studies were excluded if they involved individuals with back pain caused by specific pathologies (fracture, ankylosing spondylitis, spondyloarthritis, spinal stenosis, infection, neoplasm or metastasis), or specific conditions (pregnancy) and disc herniation with radiating pain. Three authors (GH, ATT and BEØ) independently selected studies retrieved from the searches. Disagreements were resolved through discussions. In cases where consensus could not be reached, another author (MG) was approached for selection of studies.

| Risk of bias assessment
The risk of bias assessment for each study was performed using questions adapted from Hayden, Windt, Cartwright, Cote, and Bombardier (2013) and recently recommended (Riley et al., 2019). The assessment tool included questions covering six important criteria when evaluating validity and bias of studies on prognostic factors: study participation (did the study sample represent the population of interest?); study attrition (did data from participants not lost to follow-up accurately represent the sample?); risk factor measurement (were the risk factors similarly measured for all study participants?); outcome measurement (was the outcome similarly measured for all the study participants?), confounding measurement (were important potential confounding factors accounted for in the regression analyses?), and analyses (were the statistical analyses appropriate and was the primary outcome reported?). Each criterion was judged on the basis of the reporting of the studies. The included studies were divided between two pairs of authors (BEØ and MG; GH and ATT). For each included study each pair of authors independently carried out the risk of bias assessments in which the six criteria were given either low, medium or high risk of bias. Each study was then given a "total risk of bias assessment score" as follows: low if five or six criteria had low risk of bias; moderate if four criteria had low risk of bias and high if one to three criteria had low risk of bias (last column Table 2). In cases where the majority of criteria were moderate risk of bias, a total moderate risk of bias was given. Disagreements were resolved with a discussion to reach consensus. Disagreements that could not be resolved were referred to a third review author (one author from the other pair). The operationalization of each criterion is reported in Appendix S3.

| Data extraction and synthesis
Two pairs of authors (GH and ATT, MG and BEØ) carried out the data extraction on the following variables: author and study year, population source, sample size eligible at study inception, sample size at follow-up assessment, participant age and sex, length of follow-up, back pain measurement and reported risk factors for back pain. To be able to get an overview of the risk factors for back pain, we mapped the potential risk factors from each study according the International Classification of Functioning, Disability and Health (ICF) domains a) body functions, b) body structures, c) activities and participation and contextual factors (environmental and personal; Appendix S4).
We conducted a descriptive narrative synthesis of risk factors. In the interest of identifying consistent, reproducible findings on risk factors, we summarized reported results from the studies from both unadjusted and adjusted statistical analyses for factors that had been investigated by three or more studies. We emphasized results from the adjusted analyses when interpreting and presenting the summary. For studies using the same study participants, we extracted results from the study with best methodological quality. Meta-analysis was not performed due to study heterogeneity (clinical and statistical diversity).

Years of follow-up Low back pain assessment and prevalence (%) at follow-up
Adams et al. No pain at baseline and follow-up: 24% Not pain at baseline, but pain at follow-up: 33% Pain at baseline but no pain at follow-up: 10% Pain at baseline and follow-up: 32%

Years of follow-up Low back pain assessment and prevalence (%) at follow-up
Coenen et al.

Years of follow-up Low back pain assessment and prevalence (%) at follow-up
Nemoto et al.  No association: n = 7 (Adams 1999, Mitchell 2010, Nemoto 2013, Nemoto 2012, Poussa 2005, Power 2001, Salminen 1995 No association: n = 3 (Hestbaek 2008, Hestbaek 2006 Are No association: n = 3 (Baranto 2009, Lundin 2001, Nemoto 2012 No association: n = 1 (Nemoto 2013) High Note: n, number of studies. Positive association means that if the risk factor is present, the risk of back pain is increased (e.g. a smoker has higher risk of back pain). Negative association means that if the risk factor is present, the risk of back pain is decreased (e.g. a smoker has lower risk of back pain). Number of study participants is from column #4 in Table 1. In Lunde et al. number was unavailable and the included participants were counted. In cases with several follow-ups, the number of study participants from the longest follow-up was chosen. In cases where a study presented both unadjusted and adjusted analyses, we chose to present results from the adjusted analyses only. In crude analyses for BMI and physical activity and adjusted analysis for history of back pain, Adams and Mannion found the same results from the same study participants. We excluded Mannion in both these cases. Body height was measured in several articles, however, either was height only adjusted for in the analysis or no results were shown, resulting in three papers that reported proper results for height.

| Study quality
The risk of bias assessment revealed that nine studies had low risk of bias, 26 studies had moderate risk of bias and 14 studies had high risk of bias ( Table 2). The criteria where most studies had high risk of bias were Study Attrition (n = 21) and Confounding measurement (n = 17).

| Risk factors for back pain
More than 150 potential risk factors were investigated in the 49 studies included (Appendix S4). Among all potential risk factors investigated, there were nine risk factors investigated in three or more studies. These were age, sex, height, body mass index (BMI), smoking, physical activity level, a history of back pain, job satisfaction and structural imaging. The reported results for each of these risk factors and the overall quality of evidence are presented in Table 3. Nine studies reported a significant association between a history of back pain and a new episode of back pain in adjusted analyses (Adams et al., 1999;Feyer et al., 2000;Greene et al., 2001;Hestbaek et al., 2006b;Kroner-Herwig et al., 2017;Mikkonen et al., 2013;Monnier et al., 2016;Roy & Lopez, 2013;Videman et al., 2005; Table 3). Inconsistent associations were reported from adjusted analyses between back pain and age, sex, height, BMI, smoking and physical activity level. One study reported adjusted analysis for job satisfaction and one other for structural imaging findings, both reported no association to back pain.
Of the nine studies that showed an association between a history of back pain and a new episode of back pain, five studies measured both the risk factor and back pain within 18-29 years of age (Feyer et al., 2000;Greene et al., 2001;Monnier et al., 2016;Roy & Lopez, 2013;Videman et al., 2005). Two studies measured the risk factor in emerging adulthood, and back pain in adulthood (Adams et al., 1999;Kroner-Herwig et al., 2017) and two assessed the risk factor before emerging adulthood and outcome in emerging adulthood (Hestbaek et al., 2006b;Mikkonen et al., 2013). The nine studies included more than 9,000 study participants, and the risk of bias was moderate in six of the studies, high in two, and low in one of the nine studies.

| DISCUSSION
This systematic review summarizes the evidence for risk factors for back pain in emerging adults. Nine risk factors were investigated in three or more studies, revealing that having a history of back pain consistently seems to be a risk factor for a new episode of back pain in emerging adults. Inconsistent results were found for age, sex, height, BMI, smoking and physical activity level with studies reporting both associations and no associations in adjusted analyses. The results for job satisfaction and structural imaging findings showed no association both in unadjusted and adjusted analyses, but only one study had conducted adjusted analysis for job satisfaction and one other for structural imaging findings. This systematic review showed that a history of back pain was a consistent risk factor for future episodes of back pain in young people across many studies. The studies with adjusted analysis included in total over 9,000 study participants and had overall moderate study quality. Previous reviews on other populations show similar findings. For instance, da Silva et al. (da Silva et al., 2017) reported that a history of previous episodes of back pain was the only significant predictor for low back pain in adults, but the result was based on two studies only (Hancock et al., 2015;Stanton et al., 2008). Taylor et al. systematically reviewed risk factors for first-time low back pain, and for a new episode from a pain free baseline in subjects 18 years or older (Taylor, Goode, George, & Cook, 2014). They reported that no consistent risk factor emerged as predictive of first-time low back pain, but that a history of low back pain was a consistent predictor of future incident low back pain for those who were pain free at baseline. Our intention was also to distinguish risk factors for first and new episode of back pain. It was however, challenging to ascertain if included studies had investigated first ever episode of low back pain, new episodes of low back pain or a mixture of both. Although some studies defined their outcome as a first episode of back pain, it was apparent even in these studies that a history of previous episodes was a risk factor, underlining the difficulty in identifying a true, single point of onset for back pain (Adams et al., 1999;Mannion et al., 1996). According to the present systematic review, conflicting results were found for smoking and BMI as possible risk factors for back pain in adjusted analyses. Huguet et al. reported moderate quality evidence for a trend that regularly smoking in childhood or adolescence may be associated with later low back pain (Huguet et al., 2016). Their results were based on four studies in which two of them are included in our review (Hestbaek et al., 2006a(Hestbaek et al., , 2006Mikkonen et al., 2008) and they show opposite results in adjusted analyses. Their other two studies that showed an association between smoking and low back pain investigated in children and adolescents, thus a direct comparison with data from this report could not be done. In line with our results, Huguet found no association between pain and BMI. However, with different age groups and pain sites included in the two reviews comparisons of the results could not be directly done. O'Sullivan and colleagues summarized their work from the Raine Study cohort in Australia where they found that low back pain is prevalent at the age of 14 years and it increases to adult prevalence by 22 years of age. They stated that predictors of disabling low back pain in adolescence, as in adults, are multidimensional, including female sex, negative back pain beliefs, poor mental health status, somatic complaints, involvement in sports and altered stress responses. (O'Sullivan, . This has not been demonstrated in other studies. Lack of replicable high quality studies and inconsistent results for risk factors for back pain on adolescents and emerging adults highlight the need to address these age spans more rigorously. Importantly, trajectories for long-term musculoskeletal health may start to manifest and emerge during these complex transitionary periods. Furthermore, emerging adulthood is a period of life when implementation strategies for lifestyle changes may be effective (Wing et al., 2016).

| Strength and limitations of the study
This systematic review was conducted and reported in accordance with the PROSPERO protocol and the PRISMA guidelines (Appendix S1). The systematic searches were performed by one librarian and replicated by another librarian with comprehensive search strategies in all the relevant databases and in close collaboration with the authors. The search results included studies from four continents and covered both registry and population cohort studies. Our initial intention was to investigate modifiable risk factors for back pain in emerging adults, aged 18-29 years, however we decided to expand the scope of the review and included studies with either the outcome, risk factor(s) or both measured in the age span 18-29 years. Thus, some of the studies included study populations aged 16 years at baseline and 18 years at follow-up while other studies included study participants in the upper limit of the age span. The wide age span led to a number of distinctive sample frames being used, for example school for recruiting adolescents; more selective universities/ workplaces/military for recruiting people in emerging adulthood; and population cohorts and registries generally capable of moving across both. Consequently, these data provided strong evidence to explore comparative risk factors across different life stages. We were unable to perform a meta-analysis as only a few studies had measured the same risk factor and the outcome measurement differed in terms of back pain duration and often included a mixture of study participants with both first and new episodes of back pain. Finally, we intended to apply the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework for assessing quality of evidence if that was possible (Guyatt et al., 2011). However, it was challenging to use the GRADE framework appropriately for the included cohort studies, due to large heterogeneity in the study samples, statistical analyses and handling of possible confounders.

| Implications
We found little evidence to support or refute our theory that exposures and transitions in emerging adulthood may be relevant to the risk of back pain. Future studies should examine more carefully the effect of key transitions or developmental processes occurring in emerging adulthood to inform whether or how interventions may need to be adapted to different stages of life. High quality prospective cohort studies evaluating risk factors for first ever episode of back pain with frequent follow-ups over several years, and with a comprehensive set of measured risk factors, including biological, psychological and social factors, are needed.

| CONCLUSION
There is moderate quality evidence that a history of back pain is a risk factor for new episodes of back pain in emerging adults. There are no or unclear associations for age, sex, height, BMI, smoking and activity level. No associations were found between job satisfaction and structural imaging findings and episode of back pain.