This review evaluated the results of 13 high-quality studies on the association between physical activity and neck and low back pain. We found heterogeneity among studies as to aspects such as study design, study population, type of exposures measured, methods of exposure assessment, statistical analysis, and data presentation. Thus, the analysis of the results was limited to qualitative summary. Based on the limited number of studies and the heterogeneity among studies, the results indicated limited evidence for no association between physical activity during leisure time and neck pain in the working population. Strong evidence was found for no association between physical activity and neck pain among school children. Conflicting evidence was found for the association between physical activity and low back pain in both general population and school children.
Of 13 high-quality studies, the items in the criteria list rated as negative in most studies were physical activity at work time assessment, physical activity measurement tool, and frequency of data collection during follow-up period.
Most studies solely measured physical activity level at leisure time, which may not reflect actual daily physical activity. Physical activity at work time should be assessed and included as part of daily physical activity. When physical activity at work is taken into account, workers who have sedentary activity during work, such as office workers, may have considerably different physical activity level compared with workers whose job characteristics are more physically demanding, such as nurses or refuse collectors. Therefore, future research should consider measuring physical activity at both work and leisure time in order to be more representative of an individual’s daily physical activity level.
Common physical activity level measurement methods include self-reported questionnaire, interviewing, and objective instrumentation (i.e., an accelerometer). Most studies employed self-reported questionnaire or interviewing. Only 2 out of 13 included studies used objective instruments to assess physical activity level. Many of the subjective methods had problems with reliability and/or validity. Moreover, objective methods were found to report different results than those obtained from subjective methods . Verbunt et al.  indicated that self-report measurements may lead to under- or overestimation of physical activity level, which may result in bias in the association between physical activity and musculoskeletal pain. An objective measure is preferable for assessing physical activity level. Its advantages include having greater validity and providing both quantitative and qualitative assessment of physical activity with minimal burden on participants. During physical activity monitoring, not only mean physical activity levels, but also a classification of physical activities (such as standing, sitting, and locomotion) can be collected. Nowadays, physical activity monitors are becoming more and more convenient. However, high cost and restricted registration time are still barriers. Future research should attempt to use an objective measure to evaluate physical activity level.
The follow-up period of exposure and disease for the studies varied considerably, ranging from 3 to 25 years for physical activity level and from 1 month to lifetime for neck or low back pain. Of three cohort studies, only one study recorded data every year for 3 years , whereas the rest of the studies recorded data at the beginning and the end of study only. No data collection regarding exposure and disease during follow-up period may pose a threat of recall bias. This bias may result in an under- or overestimation of the risk of association with an exposure. Kremer et al.  reported that patients with pain significantly underestimated their activity level. Schmidt and Brands  found that patients were less capable of estimating their physiological level of exertion during a performance test situation than healthy controls. Future studies should pay more attention to the frequency of data collection during their follow-up period, and it is recommended that data are collected at least every 3 months or are obtained from a continuous registration system.
Evidence for association between physical activity and neck pain
Studies were conducted in substantially varying groups of subjects, including school children, workers, and the general population. One may argue that the effect of physical activity level in different population groups might be different, particularly between adolescents and adults. This seems to be the case for neck pain. When the effect of physical activity level was separately analyzed for workers and school children, there was limited evidence for no association in workers and strong evidence indicating no association in school children.
Performing physical work, adopting awkward working postures and having sedentary lifestyle are common for workers, while such activities are rare in an adolescent population. Epidemiological studies have shown that adopting awkward working postures for prolonged time combined with having sedentary lifestyle have been found to be associated with neck pain [10, 25, 35]. Therefore, increased physical activity level in workers may be beneficial for preventing neck pain. However, the preventive effect of increased physical activity level on neck pain may not be so obvious in adolescents, who usually do not stay in awkward positions  and are more physically active than adults . Thus, future research should be more specific regarding the study population and taking the impact of work status on physical activity into account. In addition, due to the low number of high-quality studies, more research is needed to confirm our findings in this respect.
Evidence for association between physical activity and low back pain
The body of evidence regarding the role of physical activity level and low back pain is somewhat more inconsistent than that for neck pain. Even with the separate analysis of the effect of physical activity on low back pain in adolescents and adults, the conflicting evidence still existed. One of the possible explanations for inconsistent findings among studies may relate to heterogeneity in methods of exposure assessment among studies. To assess the physical activity level in patients with musculoskeletal pain, an objective measure is a preferable measurement device to self-report measurement . Wedderkopp et al. , who used accelerometers to measure physical activity level, reported that low level of physical activity increased the risk of low back pain in school children. Being physically active may lead to improved physical fitness, which consequently reduces the risk of low back pain and helps the back to function better . However, the rest of the studies employed self-report measurements to examine physical activity level, which are prone to the risk of recall bias. For example, those without low back pain may be more likely to consider themselves to be physically active than those with low back pain or those who are physically active may be more likely to consider their back to be in better condition than those who are less physically active, even if this is not the case . Due to conflicting results, more high-quality studies are needed before a final conclusion can be reached regarding the effect of physical activity on low back pain.
Methodological quality of included studies ranged between 43 and 88%, with eight of 17 studies scoring between 43 and 57%. In this review, a priori cut-off point of >50% was used, which might have influenced the level of evidence and potentially the results of the review. Thus, we assessed the effect of the cut-off point used in the methodological quality assessment on the level of evidence. Shifting the cut-off point from >50 to >60% or shifting the cut-off point from >50 to >40% would not have influenced our levels of evidence at all.
The strength of evidence was divided into five levels. However, in an earlier study by Hamberg-van Reenen et al. , three levels of evidence were used, i.e., (1) strong evidence: consistent findings in multiple high-quality studies; (2) moderate evidence: consistent findings in one high-quality study and in at least one low-quality study, or consistent findings in multiple low-quality studies; (3) inconclusive evidence: inconsistent findings in multiple studies, or the results based on one or no study provided findings for or against an association. Changing the method to assess the strength of evidence into the one used by Hamberg-van Reenen et al.  would not have altered our conclusions.
Limitations of this review
There are a number of methodological limitations of this systematic review that are noteworthy. First, the search strategy was limited only to full reported publication in English. The possibility of publication and selection bias cannot be ruled out, which may affect the results of the review. Second, we summarized the results from studies with substantial heterogeneity. This may explain the observed variation in the results among studies. Future research is required to indicate whether differences in these aspects affect the association between physical activity on one hand and neck and low back pain on the other. Last, quality assessment tools to appraise observational studies are less well established than those for randomized controlled trials. As no universally accepted quality assessment tool for observational studies exists, the methodological quality assessment used in the present review was based on the assembly of criteria lists in the previous reviews [23, 43]. It is believed that the items included in the criteria list assessed the important components to validate these types of studies.