Physical and psychosocial ergonomic risk factors for low back pain in automobile manufacturing workers
- 1Department of Public Health, Temple University, Philadelphia, Pennsylvania, USA
- 2Department of Family and Community Medicine, School of Nursing, University of Pennsylvania, Philadelphia, Pennsylvania, USA
- 3Department of Work Environment, University of Massachusetts Lowell, Lowell, Massachusetts, USA
- Correspondence to Jonathan L Vandergrift, National Comprehensive Cancer Network, 275 Commerce Drive, Suite 300, Fort Washington, PA 19034, USA;
Contributors All authors contributed to the conception and design, acquisition of data or analysis and interpretation of data, and to the drafting or critical revision of the article for important intellectual content. All authors gave final approval for the version published.
- Accepted 14 April 2011
- Published Online First 17 May 2011
Objectives To examine the association between ergonomic physical and psychosocial exposures and the risk of prevalent and incident low back pain (LBP) in a longitudinal cohort of automobile manufacturing workers.
Methods Ergonomic exposure intensity and LBP presence were determined through questionnaires at baseline (n=1181) and to workers in the same job 1 year later (n=505). Models were constructed using log-binomial regression with special attention to interactions between ergonomic exposures.
Results Awkward back posture (prevalence ratio (PR) 1.12, 95% CI 1.07 to 1.17), hand force (PR 1.06, 95% CI 1.02 to 1.10), physical effort (PR 1.10, 95% CI 1.04 to 1.16) and whole body vibration (PR 1.04, 95% CI 1.01 to 1.08) were each associated cross-sectionally with LBP. Awkward back posture (RR 1.13, 95% CI 0.98 to 1.31) and hand force (RR 1.07, 95% CI 0.93 to 1.22) also predicted incident LBP, although estimates were statistically less precise. Neither job control, psychological demands, nor job strain was independently related to risk of incident LBP. Among participants reporting high physical ergonomic exposures and moderate to high job control, increasing job demands was associated with a reduced LBP risk (RR 0.72, 95% CI 0.52 to 1.00). Among participants reporting high physical exposures and low job control, psychological demands was associated with an increased LBP risk (RR 1.30, 95% CI 1.02 to 1.66).
Conclusions Psychosocial workplace interventions for LBP should prioritise jobs in which there are high physical ergonomic exposures. Future studies of LBP should examine the interactions between physical ergonomic risk factors.
What this paper adds
Physical and (less consistently) psychosocial risk factors have been associated with low back pain (LBP); the interaction between these two categories of exposures has not been resolved epidemiologically.
These data support previous findings that self-reported occupational physical exposures including awkward back postures, hand force, physical effort and whole body vibration are associated with LBP.
Psychosocial job strain (high demands, low control) was associated with new LBP only in those with high baseline physical exposures.
Low back pain (LBP) is one of the most common ailments in the USA and at any one time affects 12–30% of the population.1 Many of the factors that are important in LBP aetiology are associated with the workplace and as much as 30% of all LBP can be attributed to occupational exposures.2 The occupational physical ergonomic exposures associated with LBP risk include awkward back postures,3–5 physical effort, including manual exertion related to handling objects or people,5 6 and exposure to whole body vibration (WBV).5 7
In addition to the separate effects of each of these physical ergonomic risk factors, experimental evidence and biomechanical theory suggest that they may interact, for example, producing a higher risk of LBP when hand forces8 or physical effort9 10 are exerted in combination with awkward back posture. Biomechanical theory also predicts that local low-back muscle fatigue resulting from job-related physical effort9 may increase the risk of LBP associated with lifting.11 Lastly, evidence from epidemiological studies of LBP suggests an interaction between awkward posture and WBV.7
Exposure to job-related psychosocial stress has also been implicated in the aetiology of work-related LBP. Job strain is defined in the job demand–control model as a combination of high psychological job demands and low decision-making authority at work (‘job control’).12 Job strain has been associated with LBP in two prospective13 14 and two cross-sectional studies.15 16 LBP has also been separately associated with both job demands and job control.3 17
Physical and psychosocial ergonomic risk factors are often correlated with one another in the workplace,18 suggesting a more complex aetiology than is characterised by treating them as covariates in separate statistical models. In addition, evidence from both experimental19 and epidemiological studies20 21 suggests that psychosocial and physical exposures may interact synergistically to cause musculoskeletal disorders.
In the current study, we examined the association between occupational physical and psychosocial ergonomic risk factors and LBP in a 1-year longitudinal cohort of automobile manufacturing workers. The primary aims of the study were to examine: (1) the separate associations between physical and psychosocial ergonomic risk factors and LBP risk; (2) selected interactions between physical ergonomic risk factors for risk of LBP; and (3) the interaction between physical and psychosocial ergonomic risk factors in their association with LBP.
The study was conducted among a cohort of automobile manufacturing workers (n=1550) from an automotive stamping plant and an engine assembly plant located in Detroit, Michigan.22 Overall, 85% (n=1315) of workers in the targeted departments were enrolled in the study. Individuals were excluded from the study for poor baseline data quality or cooperation, inability to participate in a physical examination of the musculoskeletal system due to injury (such as an amputation) or if they were employed in non-production work duties (eg, union officers) (n=34). Participants were excluded from these analyses if they reported mechanical back problems (spondylitis, spondylolisthesis or ankylosing spondylitis), a ruptured disc in the neck or back, or a history of back surgery (n=100). Only participants who were LBP-free at baseline (figure 1) and remained in the same job during the study period were eligible for inclusion in the analysis of incident LBP approximately 1 year later.
Assessment of exposures and outcome
Ergonomic exposures, demographics and LBP status were assessed during structured interviews conducted during work time. The five physical exposures used in the current analysis included exposure to awkward back postures, WBV, physical effort, hand forces related to handling tools or materials, and work pace. The Borg CR-10 scale23 was used to grade the intensity of the physical exposures on a 0–10-point scale. A composite physical exposure metric was computed by summing the five individual psychophysical (Borg CR-10) exposure scores. Participants were divided into low and high physical exposure categories based on the median of the composite physical score. These physical exposure measures were obtained at baseline (T0) and at follow-up (T1).
The psychosocial work environment was assessed using the Job Content Questionnaire (JCQ)12 only in the follow-up interview. The standard JCQ algorithm was used to score each participant's exposure to job demands and job control. Job strain was defined as a dichotomous variable where job demands scores were ≥30.67 and job control scores were ≤65.92, based on mean values for the US male working population.12
LBP was assessed by a question on the presence of any musculoskeletal symptoms experienced more than three times or lasting more than 1 week during the previous 12 months.24 Participants were asked to locate the source of their symptoms using a body map. Those individuals who located the source of pain in the lower back were designated as having reported LBP.
The analysis of LBP was conducted in two stages. First, the cross-sectional association between physical exposures and prevalent LBP at T0 was examined in all subjects. Second, the risk of 1-year ‘incident’ LBP (present at T1 among those pain-free at T0) among participants who remained in the same job during the study period, was examined in relation to physical risk factors measured at baseline and psychosocial factors measured at T1.
Demographic and occupational factors examined included age, seniority in the company, body mass index (BMI), height, weight and gender. Differences between means were assessed with the Student t test or Satterthwaite's approximate t test (ts) if there were significantly different variances between groups. χ2 Tests were used to assess differences between proportions. All analyses were conducted using SAS v. 9.1 (SAS Institute). A p value of ≤0.05 denoted statistical significance.
Univariate and multivariable log-binomial regression models25 were used to compute prevalence ratios (PR) in the cross-sectional analysis of prevalent LBP and RRs in the longitudinal analysis of incident LBP. Physical and psychosocial ergonomic risk factors were entered as interval data into the models predicting LBP.
Confounding was defined as a change of 20% or more in the computed risk estimate. No confounding effects were observed among the covariates examined.
A number of exposures were hypothesised to interact with one another in their association with LBP. To examine these conditional relationships, the association between LBP and one exposure was stratified on the second exposure. Participant exposure scores were divided into tertiles (low, medium and high) based on the distribution of the data in the sample, to give roughly one-third of the study population in each stratum. Risk estimates were computed within each strata of the suspected effect modifier. A noteworthy interaction was defined as a >100% change in the calculated risk estimates among strata.
In the analysis of the interaction between physical exposure and psychosocial job characteristics, there were relatively few incidents of reported LBP among participants with a high physical workload and medium or high job control. Therefore, the medium and high job control strata were combined to ensure convergence of the log-binomial model.
Baseline cohort characteristics
In total, 1181 participants were included in the cross-sectional baseline analysis of prevalent LBP (figure 1). This population was mostly male and on average overweight (mean BMI 27 kg/m2, SD 4.84), with a mean age of 46 years (SD 8.21) and a mean of 20 years (SD 6.60) of seniority at the automobile manufacturing company (table 1). The overall age range of the cohort was 20–73 years with the middle 50% of the population between 41 and 52 years of age (inter-quartile range of 11 years). The highest reported exposure rating was for job pace (mean 7.12; table 1), followed by physical effort (mean 6.80). The lowest perceived exposure intensity was reported for WBV (mean 2.66).
At baseline, 20% of participants (n=232) reported having had LBP in the previous 12 months (figure 1). Participants reporting LBP at baseline had 8–9 months higher seniority (ts (489)=2.22, p<0.05) than participants free of LBP (table 1). No other demographic variables examined were associated with prevalent LBP and none of these confounded the effects of the occupational exposures examined.
Awkward back posture (PR 1.12, 95% CI 1.07 to 1.17), physical effort (PR 1.10, 95% CI 1.04 to 1.16), WBV (PR 1.04, 95% CI 1.01 to 1.08) and hand force (PR 1.06, 95% CI 1.02 to 1.10) were each significantly associated with LBP at baseline (figure 2). No effect was observed for job pace on prevalent LBP (PR 1.02, 95% CI 0.97 to 1.08). There was no effect modification between awkward back posture and either physical effort or hand force. Exposure to WBV did not appear to affect the association between awkward back posture and prevalent LBP.
Follow-up cohort and new LBP
Of the participants without LBP at baseline (n=949), 598 (63%) were assessed at T1 and 84% of these (505) reported being in the same job (figure 1). A total of 25 cases of incident LBP (5%) were identified among these 505 participants. No significant demographic differences were observed between participants with and without incident LBP at follow-up.
Participants lost to follow-up had slightly higher BMI at baseline (mean 28.0 kg/m2, SD 5.41) than those assessed at follow-up (mean 27.2 kg/m2, SD 7.2) (ts (621)=2.20, p<0.05). There were no significant differences between the two groups in age, seniority or gender. Among the physical exposures measured at baseline, only the hand force rating differed between those assessed at T1 (mean 5.63, SD 3.17) and those lost to follow-up (mean 5.02, SD 3.40) (p=0.01).
For incident LBP, awkward back posture and hand force at baseline had similar coefficients for risk per unit exposure rating to those obtained in the cross-sectional data, although with wider confidence intervals (figure 2). Exposure to physical effort, WBV and job pace were unrelated to risk of incident LBP.
The risk due to awkward back posture did not change by level of exposure to physical effort or hand force. Similarly, there was no interaction between physical effort and hand force. However, the risk of incident LBP associated with awkward back postures did increase when combined with high exposure to WBV. At low (Borg CR-10 score 0, RR 1.10, 95% CI 0.94 to 1.32) and medium (Borg CR-10 score 0.5–4, RR 1.11, 95% CI 0.81 to 1.53) levels of WBV, risk estimates associated with awkward back postures were similar to those associated with awkward back posture overall. At higher exposures to WBV (Borg CR-10 score 5–10), the effect of awkward back postures was larger although not statistically significant (RR 1.66, 95% CI 0.91 to 3.03).
Psychosocial risk factors and new LBP
There was no association between incident LBP and psychological job demands (RR 1.01, 95% CI 0.90 to 1.12) or job control (RR 0.98, 95% CI 0.95 to 1.03) assessed at T1 for the cohort as a whole. In addition, being in a high strain job at T1 was unrelated to incident LBP (RR 0.96, 95% CI 0.34 to 2.76). The risk of incident LBP due to high psychological job demands did not change by job control tertile (table 2).
In contrast, there was an interaction between demands and control after further stratification on physical workload. Among participants with both high physical exposures and low job control, job demands was associated with a significantly increased risk of incident LBP (RR 1.30, 95% CI 1.02 to 1.66) (table 3). Among those with high physical exposure and medium to high job control, increasing job demand was protective against risk of incident LBP (RR 0.72, 95% CI 0.52 to 1.00). When physical exposure was low, job demand was unrelated to LBP, regardless of level of job control.
A moderate correlation was observed between physical workload and psychological job demands (r=0.33, p<0.001). This correlation was consistent across job control tertiles (low job control: r=0.33, p<0.001; medium job control: r=0.38, p<0.001; high job control: r=0.33, p<0.001). No correlation was observed between job control and physical exposures (r=−0.03, p=0.55).
Awkward back posture and hand force were associated with an increased risk of both prevalent and 1-year incident LBP in a cohort of automobile manufacturing workers. Neither psychological job demands nor job control alone was associated with incident LBP for the cohort as a whole. Among participants with high physical exposure at baseline and low job control, job demand was associated with an increased risk of incident LBP during the 1-year follow-up period.
One in five participants enrolled at baseline reported LBP in the prior year. This is similar to the 23% prevalence rate observed in a previous study of autoworkers15 and is within the 12–30% annual point prevalence of LBP in the US adult population.1 Among participants who did not report prevalent LBP at baseline, approximately 5% reported an episode of incident LBP during the 1-year follow-up period. This is greater than the 1-year incidence rate of 2% observed in a cohort of Iranian autoworkers.15
The date of the first occurrence of LBP was not ascertained in this cohort. Hence, workers without back symptoms at baseline may have experienced prior LBP that had resolved. Therefore, it is possible that LBP ‘incident’ cases (defined as those without back symptoms at baseline, and with back symptoms at follow-up) may include both re-occurring and new LBP cases. Because of this limitation, the observed rate of incident LBP may overestimate the true rate in the study population.
Strengths and limitations of the study
One of the primary strengths of the current study is the large cohort of workers (n=1181) enrolled, representing 85% of the autoworkers in the targeted departments at baseline. Despite the high enrolment rate, about 40% of participants were lost to follow-up. However, many participants probably simply aged out of the workforce. The population lost at T1 was similar to the population assessed at follow-up in terms of demographics, physical exposures and baseline reports of LBP. This suggests that attrition did not likely result in selection bias.
Another advantage of the current study is that it included both a cross-sectional and longitudinal component. The longitudinal analysis ensured that the physical exposures being examined occurred prior to the outcome of new LBP, providing support for a causal association. In addition, validated measures were used for grading participant's perceived physical (Borg CR-10)23 and psychosocial exposures (JCQ).12
Despite the large cohort of participants enrolled at baseline, the analysis of incident LBP is limited by statistical power, specifically the small number of new cases. In part, this may be a function of the maturity and seniority of the cohort, representing a ‘survivor’ group. The current cohort of automobile manufacturing workers was on average employed in the company for over 20 years. If musculoskeletal pain leads to earlier departure from the workplace, it is possible that those workers still employed after 20 years, and subsequently enrolled in the current study, would have a lower risk of work-related pain associated with ergonomic exposures than an employee who had recently started working (the ‘healthy worker effect’).26 For example, Miranda et al27 observed a stronger relationship between exposure to physical ergonomic risk factors and incident LBP among younger than older workers. If the population included in our study was less susceptible to developing work-related LBP, the effects observed could underestimate the true risk that a new employee would face from similar exposures.
One limitation in the analysis of psychosocial factors is that these variables were only assessed at follow-up. Although subjects analysed at follow-up were in the same jobs as at baseline, the study is unable to address definitively the directionality of the association between LBP and psychological job demands or job control. It has been suggested that the development of LBP may lead individuals to perceive a poorer psychosocial work environment,28 and it cannot be conclusively determined that LBP did not affect the participant's reporting of their psychosocial work environment.
Physical exposures were assessed through self-report, which introduces the potential for information bias. An analysis of self-reported versus direct measures of exposure in this cohort found no evidence of a differential bias that might lead to a spurious association with musculoskeletal symptoms.29 This analysis examined upper extremity exposures, and, as such, examined all of the physical exposures analysed in the current study with the exception of self-reported awkward back postures. Hence, there is still potential for a bias in this self-reported exposure. But the nature of this bias (whether differential or non-differential) is unknown, as the published literature is inconclusive with regard to how those with back pain estimate their physical exposures.30–32
Lastly, only automobile manufacturing workers were included in the study suggesting there may be limited variability in ergonomic exposures. However, the two plants included in the study had different degrees of automation and were selected to enrol a cohort of autoworkers with variability in physical ergonomic exposures. Nonetheless, at baseline, almost all participants were working in highly routinised jobs, such as on machine paced assembly lines, with the majority focused on a single cyclical task.33 Therefore, there may be limited variation in psychosocial ergonomic exposures, particularly with regards to job control.
Awkward back posture, hand force, physical effort and WBV were associated with an increased prevalence of LBP at baseline. In the analysis of incident LBP, only exposure to awkward back postures and hand force predicted an LBP episode, although with low precision.
Awkward back postures have been previously associated with LBP in a number of prospective,3 4 27 case–control5 24 and cross-sectional studies.34 35 The LBP risk estimates associated with awkward back postures observed in the current study are smaller than the majority of the risk estimates previously reported. However, a direct comparison of effect sizes is problematic due to differences in exposure assessment. Similarly, the risk of LBP associated with hand force observed in the current study is smaller than the associations observed in previous studies.6
WBV was associated with a small, but statistically significant, increased prevalence of LBP at baseline. A number of epidemiological studies have examined the association between WBV and LBP.5 7 27 36 Much of the epidemiological literature examining the effect of WBV on LBP has been conducted in drivers or in heavy machine operators exposed through the seat of a vehicle. In the current study, exposure occurred as a steady state vibration through the feet and legs when workers were in contact with the vibrating floor in the stamping plant. Biomechanically, the stiffness of the coupling (in the current study, whether or not the supporting legs are flexed) may have had a substantial effect on the transmission of vibration to the spinal column.37 Since this factor was not accounted for, there may have been some further misclassification of exposure leading to dilution of effect.
There was limited evidence of an increased risk of incident LBP in participants exposed to both WBV and awkward back postures, compared to those exposed only to awkward back postures. An interaction between WBV and awkward postures in bus drivers was previously reported by Okunribido et al.7 The physiological mechanism for this interaction is unknown, although laboratory studies suggest possible mechanisms including increased muscle fatigue and disc compression.37
Incident LBP was higher in a high demand–low control psychosocial work environment for workers also heavily exposed to physical ergonomic factors. No effect modification was observed between job control and demands in the incident LBP analysis for the cohort as a whole. However, there was a three-way interaction among physical exposures, psychological demands and job control. The association between job demands and incident LBP was only observed among participants with low job control who were also highly exposed to physical stressors. Surprisingly, the risk of LBP decreased with increasing job demands in participants with medium to high job control and high physical exposures. It was anticipated that risk of LBP would not be associated with job demands in these participants. Potentially, this may represent a protective effect of the psychosocial work environment for musculoskeletal disorders associated with increased active learning (eg, high demand, high control jobs). Active jobs have been associated with other protective health effects.38 Further research is needed to confirm a protective effect of active learning on LBP.
A moderate correlation between psychological demands and physical exposures was observed suggesting that the findings in table 3 are due to psychological demands over and above physical demands. In addition, this correlation was consistent across job control tertiles. The observed interaction between job demands, job control and physical exposures suggests a more complex relationship between psychosocial and physical ergonomic exposure than one of simple confounding.
The association between the physical and psychosocial risk factors in the aetiology of LBP has been explored in previous studies.20 21 39 40 Waters et al21 identified an interaction between work stress and heavy lifting. Huang et al20 found effect modification between biomechanical exposures and low ‘participatory management’, which may be analogous to greater job control. Devereux et al40 examined the interaction between psychosocial and physical exposures and LBP in a cross-sectional survey using additive risk models. A large proportion of risk (65%) was attributed to interaction effects for recent LBP in mixed-sex models using proportional prevalence ratios. However, there were no female participants in the high physical exposure group; when the model calculating proportional prevalence ratios was examined in only male participants, a minimal deviation from additivity was observed (12%). The impact of job control modifying the risk of LBP (or any other musculoskeletal disorder) due to psychological demands in the presence of high, but not low, physical exposures has not been previously reported.
The exact mechanism by which psychosocial stress may lead to an increased risk of musculoskeletal disorders is not well understood. One hypothesis is that job strain leads to increased muscle tension that magnifies the impact of physical stressors on the lower back.5 Alternatively, observed increased spinal loading when a lifting task was combined with simultaneous mental processing was attributed to an over-reaction of the musculoskeletal system characterised by less controlled movements and increases in muscle co-activation.19
Exposure to awkward back postures and hand force exertion in automotive manufacturing increased the risk of LBP at both baseline and 1-year follow-up. Job demands were associated with the development of incident LBP, although only in workers with low job control and high reported baseline exposure to physical risk factors. Results suggestive of an interaction between awkward postures and WBV were also observed.
The observed relationship between psychosocial and physical exposures may help explain some of the inconsistency observed between studies examining the impact of the psychosocial work environment on the risk of LBP. The current results suggest that if the association between the psychosocial work environment and LBP risk were examined in a population with an overall low exposure to physical risk factors, no association would be detected. It is possible that physical stress is one component of the pathway through which psychosocial factors increase the risk of LBP. Alternatively, it may be that job stress does not cause LBP directly but instead aggravates the impact of physical stressors on the lower back.
It will be important to examine the interaction between the physical and psychosocial risk factors in additional populations and using alternative methods of exposure assessment. A number of hypothesised relationships between physical risk factors were not observed in the current study. However, many of these relationships have not been previously explored outside of the laboratory. These should be examined using epidemiological methods in other occupational settings to conclusively determine their impact on LBP.
The current study, in addition to identifying and quantifying specific LBP risk factors, illustrates the complex multi-factorial nature of LBP aetiology. While it is widely recognised that many different factors may be the cause of an episode of LBP, it is also important to consider that these factors are not experienced in isolation. Identifying interactions between LBP risk factors is particularly important when designing control measures. For example, an intervention designed to address the psychosocial work environment may be most effective if it is focused on increasing job control and/or reducing job demands for workers who are exposed to a high physical ergonomic load. When designing a prevention program or evaluating a workplace risk factor, recognising that some workers or departments are more at risk will help to maximise the impact of any intervention efforts.
The authors thank Dr Deborah Nelson for her guidance during the development of this paper. We also thank the many individuals at the United Automobile Workers and the manufacturing company who assisted with data collection. Participation of the individual workers is gratefully acknowledged.
Funding This research was supported by the National Joint Committee on Health and Safety jointly sponsored by the manufacturing company involved and the United Automobile Workers. This manuscript is solely the responsibility of the authors and does not necessarily represent the official views of any other agency or institution.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.