Low back pain (LBP) affects most adults at some point in their lives, some 85–95% of cases being classed as ‘non-specific’ (i.e. without identifiable underlying pathology) [1, 2]. In recent decades, it has consistently been the leading cause globally of years lived with disability [3], and in Japan, it is one of the most common causes of disability, with a reported lifetime prevalence of more than 80% [4]. In the workplace, it is a costly problem, not only impairing the health of employees, but also reducing productivity [5]. The largest societal costs arise from cases in which the pain is disabling [6].

Various risk factors for the development of LBP have been identified previously, including mechanical stress from occupational activities such as lifting, bending, twisting and manual handling [7], and also psychosocial factors such as low mood, somatizing tendency (a tendency to worry about common somatic symptoms), job dissatisfaction, and adverse health beliefs about the causes and prognosis of back disorders [7–12]. Moreover, epidemiological studies indicate that most people with a history of LBP experience a recurrence within a year [13–16]. Thus, the occurrence of LBP is an important predictor of future episodes [7, 8, 17–20].

Most of the research on these risk factors has been conducted in Western populations, but it is possible that because of cultural and environmental differences, their impacts are not the same in other countries [21]. In an earlier prospective cohort study of Japanese workers who had been symptom-free for at least 1 year, we found that, in accordance with observations in Western populations [7, 22–24], past history of LBP, interpersonal stress at work, and frequent occupational lifting were all important predictors of disabling LBP [25]. Before that study, risk factors for new onset LBP, and in particular the role of psychosocial aspects of work, had not been properly assessed through prospective epidemiological research in Japan, and there remains a need for further investigation to confirm its findings.

We therefore conducted a new longitudinal study, as part of an international investigation of risk factors for musculoskeletal pain and associated disability, the Cultural and Psychosocial Influences on Disability (CUPID) study, which focused on workers aged 20–59 years from 47 occupational groups in 18 countries [26–30]. Using data from the CUPID study, we again assessed risk factors for new onset of disabling LBP among Japanese workers.

These results indicate that past history of LBP and working long hours were risk factors for the new onset of disabling LBP among Japanese workers who had been symptom-free during the 12 months before baseline. In addition, risk was increased in participants who reported occupational lifting, although not significantly at a 5% level.

In the present investigation, the incidence of disabling LBP was relatively low (6.1%) which may reflect our strict definition of disability (interference with work), as well as the requirement for a long symptom-free period before baseline. It has previously been proposed that an episode of LBP can be classed as new if it occurs after a period of at least 1–3 months without symptoms [38]. However, LBP is commonly recurrent within a year [13–16]. Moreover, a recent systematic review indicated that only 33% of patients in a primary care setting have recovered from non-specific LBP at a year after onset, whereas approximately 65% still experience pain [39]. Give these findings, we felt justified in requiring a 12-month symptom-free period at baseline, when exploring risk factors for new episodes, although we recognize that the criteria are to some extent arbitrary. In our earlier study, the incidence of new disabling LBP during 2 years of follow-up in workers who had been without LBP for more than 12 months before baseline was 3.9%, which is a little lower than in the current investigation [25].

We found that past history of LBP was the strongest and most significant risk factor for new disabling LBP, with an OR of almost three. This accords with our earlier study in Japan [25], and also with observations in Western populations [7, 8, 17–20]. It may be that the occurrence of a back problem renders an individual more vulnerable to future episodes of LBP (e.g. through changes in spinal structure and function or in the central processing of pain). Alternatively, the association might reflect continuing exposure to risk factors that were responsible for the initial development of the back problem. In our analysis, the association with past history of LBP was present after adjustment for other risk factors, but there may have been other important determinants of LBP that we did not assess.

In addition to past history of LBP, working ≥60 h per week and lifting weights of ≥25 kg by hand carried significantly elevated risk in analyses that adjusted for age and gender, the association with occupational lifting falling just short of significance when risk estimates were mutually adjusted. Biomechanical loading of the spine from manual handling tasks such as lifting, has been found experimentally to be greater in the presence of demands for mental processing that induce stress [24, 40]. Moreover, working overtime has been reported to increase risk of musculoskeletal disorders such as LBP [41]. While excessive working hours, perhaps entailing physical exhaustion as well as mental strain, could of itself lead to LBP, it might also act by potentiating the risks from spinal strain as a consequence of heavy lifting.

Long working hours may also reflect an element of “workaholism” in which an employee, whether for personal reasons or in response to an over-demanding job, spends excessive time at work to the detriment of his or her personal life [42]. This too is a previously reported risk factor for disabling LBP [43].

An association with long working hours was not apparent in our earlier study [25]. On the other hand, that investigation found new incidence of disabling LBP aside was significantly related to interpersonal stress at work, a finding that was not replicated in the current analysis. These differences may reflect differing characteristics of the populations studied. For example, in the earlier investigation, the participants were mostly male (88.3%) and office workers (76.1%). Alternatively, they could have occurred by chance. They underline the need for replication of results, especially when multiple risk factors are examined without strong prior expectations, and there is therefore greater potential for false positive results.

That said, the findings of the present study are not clearly different from those in Western populations. Divergence from other countries in the factors affecting new onset of disabling LBP might perhaps have been expected as a consequence of cultural differences. However, a trend to westernization in Japan may have reduced those differences. Alternatively, our questionnaire may not have covered risk factors that would differ from those in other countries or cultures.

Some limitations of our investigation should be noted. First, the generalizability of the results may be limited because the study sample was not fully representative. For example, the proportion of female participants was small in comparison with that in the national workforce of Japan. Second, because information about exposures and symptoms was obtained by self-report, some degree of misclassification is likely. Physical exposures, such as heavy lifting, might be assessed better using objective measures. Because of constraints on the total length of the questionnaire, the ascertainment of interpersonal stress was based on a single question rather than the longer Brief Job Stress Questionnaire [44] that we had used to assess psychosocial factors including interpersonal stress in our earlier study. In addition, there is a possibility of recall bias, given that the presence and severity of LBP, both at baseline and follow-up, were ascertained retrospectively. For example, participants with physically demanding jobs may have been more likely to recall symptoms and difficulty with work. Third, because the outcome was relatively infrequent, statistical power was limited. Lastly, although the present analysis included most of the well-established risk factors for new onset LBP, as well as other potential risk factors that have been suggested by earlier studies, it is possible that some important determinants, perhaps distributed differentially by occupational group, were overlooked, leading to unrecognized residual confounding. Given these limitations, our results should be interpreted with caution.

In conclusion, our findings suggest that among Japanese workers, as elsewhere, past history of LBP is a major risk factor for the development of new episodes of disabling back pain. They give limited support to the association with occupational lifting that has been observed in earlier research, both in Japan and in Western countries. In addition, they suggest a possible role of long working hours, which merits further investigation.