Background
With over 100 million people reporting discomfort in muscles or joints within the European population, the prevalence of musculoskeletal disorders (MSD) is a severe problem [
1]. As the fourth greatest cause of overall ill health and the leading cause of disability, this it is also a global concern [
2]. In the working population, the presence of MSD may result in reduced work ability [
3,
4] and an increased sickness absence [
1,
5].
The Norwegian Directorate of Health recommends young adults (age 13–17) and adults (age 18+) a minimum of 30–60 min of daily leisure time physical activity (PA) to gain positive health benefits. This is a rather common advice and is also promoted as the 10 000 steps per day slogan [
6].
Previous studies have not been able to produce consistent results on the association between PA and musculoskeletal pain [
7,
8]. A cross-sectional study published in 2011 indicated that PA (increased frequency, duration and intensity) seemed to reduce level of general chronic pain in adults [
9]. This was also later shown through a longitudinal study design, with five data collection points over a 12 month period. There seemed to be a relation between the two variables when close in time, indicating that subjects reported less pain at times when they reported more exercise [
10]. The discussion will then be if the decrease in pain leads to more exercise or if increased exercise reduces the pain.
Studies have shown that pain occurring in children/adolescents often are persisting [
11,
12] and it is suggested that pain formed in younger years may be related to pain in adulthood [
13]. As part of The Young Finns study the researchers found low level of PA to be an independent risk factor for low back pain (LBP) in 24 to 39 year olds [
14]. This relationship was also reported by others studying a similar age group [
15] and concerning general pain in children [
16]. Some of the few prospective studies carried out on school children have not been able to show that PA at baseline can predict LBP at follow-up [
17,
18]. However, most studies on this topic are cross-sectional or have short follow-up periods and/or a limited frequency of time points during follow up. Thus, larger prospective studies focusing on leisure time PA and its association to LBP in young are lacking [
7]. The prevalence of back pain in young increases with age [
19] and entering working life may additionally introduce a set of risk factors for MSD [
20‐
22], and therefore possibly increase the risk of developing pain [
23]. A study following both level of PA and development of LBP in young adults during their transition from school into their first years of working life may give important information on causal relationships. Several authors suggest that young subjects should be focused on to prevent present and future LBP and that PA should be part of the solution [
19,
24,
25]. However, the evidence on the role of PA in this setting is scarce and information based on longitudinal studies should be of importance when determining if PA may prevent LBP in young stepping into adult life [
26,
27].
The overall aims of this study was to prospectively examine the prevalence and course of LBP in young adults transitioning from school to working life over a 6.5 year period. The specific objective was to investigate associations between PA level and LBP using multiple time point measurements.
Discussion
In this study we found a significant effect of time on LBP. Further, there was a weak protecting, but non-significant association between PA and LBP in this 6.5 year follow-up among young adults entering working life.
The overall decreasing trend in LBP from baseline to follow-up shown in this study has also been shown in other studies of LBP [
35], and may be due to an initial attention being directed towards the pain area. Additionally, large population studies have shown that the prevalence of LBP is increasing from age 12 to around age 20, where it is seen a flattening of LBP reporting toward the age of 40 [
19]. Even though the overall tendency was a decreasing level of LBP, males seemed to be more irregular in their reporting of LBP than females, thus the tendency was clearer amongst females. In our study we had a mean prevalence of LBP of 43 % amongst females and 31 % amongst males, which is somewhat higher than what is found in general for 17–24 year olds in the Norwegian working population [
36]. The higher levels of reported LBP among young females compared to young males are shown in several studies [
24,
37].
A recent review on the association between PA and LBP showed inconsistent results [
7], being unable to conclude if PA could have negative or positive effects on development of LBP in adults. A study on adult cleaners which recorded PA and LBP weekly for a year found similar to our study a indicated weak protecting, but non-significant (
p = 0.08) effect of PA [
38]. In one of few longitudinal studies (3 year follow-up) on the relationship between PA and LBP in adolescents it was found at baseline and follow-up a lower frequency of PA and decreased strength amongst subjects with initial LBP [
18]. On the contrary, a 4-year follow-up study on 10 to 19 year olds showed a significant relationship between increased PA and a more common history of LBP [
39]. Others have also suggested that participation in sports may cause rather than prevent back pain in young, possibly in combination with growth. Still, the evidence for such relationships are lacking [
27]. In adults it has been indicated that physical fitness rather than self-reported leisure time PA is more strongly related to LBP and that physical fitness measures could show clearer relationships [
40]. In the study presented in this paper we did measure hand strength and shoulder endurance at baseline and found this to be significantly correlated to the self reported PA measure used. A study investigating the muscle strength on 5489 adolescent men at the age of 17–19 years could not relate low upper body muscular strength to self-reported musculoskeletal pain in adulthood 17 years later [
41]. Thus, the authors were not able to confirm their hypothesis that low physical fitness was a risk factor for future musculoskeletal pain.
Generally the prevalence of any LBP had a decreasing trend during follow up, while the opposite was seen for PA levels. Gender stratification of data showed that females during the first years of follow-up (T0 – T5, 20 months) did have a significantly lower level of PA together with significantly higher level of LBP compared to males. An increase in PA in females starting at T7 (28 months) was seen together with a decrease in later reporting of any LBP, equalizing the differences between genders in reported PA seen from T0 to T5 and differences in LBP seen from T2-T8 (4 to 32 months). In Fig.
2c and d we see this increase in PA together with the decrease in LBP with time for females. Together with the borderline significance (
p = 0.06) for PA on LBP in stratified analysis for females it is tempting to indicate a relationship. However, we do also see this reduction trend in LBP (despite higher level of variation in reporting) for males even though PA level are more stable in this group. Additionally it is important to not over interpret unadjusted descriptive data like the once seen in Fig.
2. Considering a potential latency in the effect of PA on LBP one could in future longitudinal studies evaluate the effect of PA at T0 on LBP at T1, effect of PA at T1 on LBP at T2 etc. (or similar strategies). In our study the time intervals between PA level collections differed, so a strategy using a discrete time variable with distinct time point effects was more appropriate.
A recent summary of previous reviews on interventions to reduce sedentary time for the young found that such interventions are generally successful, although effects seem to be small [
42]. An intervention aiming to reduce LBP by home exercise and back health education in young did produce promising results with reduction in several LBP related outcomes after the intervention period of 12 weeks [
43]. On the other hand, it is not from this study design possible to state if the reduction in pain was merely a result of healing with time. In our study we did see that the effect of time on LBP should be taken into consideration. With the design in our study we focused on information on PA and LPB taken pairwise from same time points over the whole follow-up period to obtain solid data concerning the association between these two variables. Thereby we have in this study a great foundation to address short-term effects over a long follow-up period. We approached the whole period by using the time variable, random effects of each participant and following the course of PA and LBP descriptively, but cannot determine if a certain level of physical activity may protect against LBP in the future. Others have shown positive short-term effects of different exercise programs to reduce LBP in adolescents, but have failed to determine any long-term effects [
44,
45]. This may indicate that this relationship is in the need of continuous maintenance to bear fruit.
An obvious strength of this study is the longitudinal design with repeated measures of both exposure and outcome variable over a 6.5 year period. This gave us the opportunity to describe the course of low back pain over several years in young subjects in their transition from school into working life and further enabled us to look at the relationship between PA and LBP in this group over time. The investigated group of individuals in this study is of major importance, considering the lack of information on the potential effect of PA on LBP in this group and the importance of providing strategies to reduce musculoskeletal disorders in the population. Using multiple imputation for the BMI and tobacco variables made us able to replace missing values with plausible values providing less chance of biased estimates. The high number of imputations (
m = 40) is relatively conservative with the goal of reducing power falloff [
34].
The majority of both exposure and outcome data in this study is obtained through self-reports. This is a practical and efficient way to gather repeated measures over long periods of time. Still, it is well known that all self-reports may have limitations due to recalling and averaging values for previous weeks or month. Such issues may lead to both under and overestimating of the variables of interest [
46]. In our study this may have affected both the main exposure variable PA and outcome variable LBP since both variables require the respondents to recall and average variable levels from the previous 4 weeks. Further, perceived level of strain considered as PA and perceived level of pain intensity may also be influenced by individual differences. However, initial misreporting should be of less concern when categorizing the variables in larger response groups prior to analysis, like done in this study. Even though such merging of response categories may lead to loss of information it is important to obtain a reasonable amount of participants in the response categories used. For the PA variable we argue that it is primarily of importance to differentiate between if low or high level of physical activity may have associations to low back pain, rather than between numbers of days being physically active. Variables directed towards concrete psychological aspects was not included in the adjusted linear mixed models in this study, something that could have affected the model output in an unknown manner due to its possible association to pain occurrence. We did however include variables concerning psychosocial dimensions that have previously shown to be associated with LBP [
47]. We did have pronounced fall in response rate during follow-up from the initial high response rate at baseline, where 420 out of 496 invited responded. Some of the explanation to the relatively high loss to follow-up could be due to the initial questionnaire being provided in a school setting with time provided for this activity to be carried out. This may have lead to a higher proportion of participants initially, with the drawback of lower threshold for dropping out at a later stage due to lack of motivation. The high participation rate at baseline does however reduce possibility for a selection of subjects with e.g. high levels of LBP into the study. Possibility to generalize the findings in this study to other educational groups of young adults or to other nationalities is unknown, as there are differences between educational choices and origin of nationality when it comes to both health and lifestyle [
29].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
KBV, TNH and MW contributed to the design and data collection of this study. LKL provided the idea for the manuscript, were responsible for analysis and interpretation of the data and drafting the paper. TNH, KBV, MK and MW assisted in analyzing and interpreting data, and revised the manuscript. All authors collaborated on the conceptualization and the design of the paper. All authors have given approval for the final version of the manuscript to be published and are comfortable with taking responsibility for its content.