Objectives The study was part of the Global Burden of Disease 2010 study and aimed to quantify the burden arising from low back pain (LBP) due to occupational exposure to ergonomic risk factors.
Methods Exposure prevalence was based on occupation distribution; estimates of relative risk came from a meta-analysis of relevant published literature. The work-related burden was estimated as disability-adjusted life years (DALYs). Estimates were made for each of 21 world regions and 187 countries, separately for 1990 and 2010 using consistent methods.
Results Worldwide, LBP arising from ergonomic exposures at work was estimated to cause 21.7 million DALYs in 2010. The overall population attributable fraction was 26%, varying considerably with age, sex and region. 62% of LBP DALYs were in males—the largest numbers were in persons aged 35–55 years. The highest relative risk (3.7) was in the agricultural sector. The largest number of DALYs occurred in East Asia and South Asia, but on a per capita basis the biggest burden was in Oceania. There was a 22% increase in overall LBP DALYs arising from occupational exposures between 1990 and 2010 due to population growth; rates dropped by 14% over the same period.
Conclusions LBP arising from ergonomic exposures at work is an important cause of disability. There is a need for improved information on exposure distributions and relative risks, particularly in developing countries.
Statistics from Altmetric.com
Back pain is a common experience for adult workers. Cross-sectional studies of back pain from developed countries report a point prevalence range of 15–30%, a 1-year prevalence as high as 50% and a lifetime prevalence between 60% and 80%.1 Data are more scarce for developing countries, but rates are similar in Africa, with an estimated point prevalence of 32%, 1-year prevalence of 50% and lifetime prevalence of 62%.2 Back pain is a major cause of disability, socioeconomic problems and loss of quality of life in developed countries,3 but its consequences have rarely been studied in other regions.
Occupational conditions that cause back pain have been studied widely. Known risk factors include rapid pace of work and repetitive motion; insufficient recovery time; vibration; physical loading involving heavy lifting, bending and twisting; and sustained non-neutral postures,1 ,4–11 although the role of these factors has been questioned by some authors.12 ,13 The causes of low back pain (LBP) are acknowledged to be multifactorial.14 ,15 Any of these risk factors in combination with each other or with negative aspects of the psychosocial work environment such as high pressure and minimal autonomy can contribute to back pain.4
The Global Burden of Disease Study 2010 (GBD 2010) is an international collaborative effort led by the Institute for Health Metrics and Evaluation to quantify the absolute and relative burden of ill health and to assess the contribution of major risk factors to this ill health. One aspect of this study involved examining risk factors and their associated outcomes arising from employment, with ergonomic risk factors causing LBP being one of the specific risk factors. The aim of the study presented here was to quantify the burden arising from LBP due to occupational exposure to ergonomic risk factors. This paper is one of 10 papers published in Annals of Rheumatic Diseases that arose from the GBD 2010 study and that address various aspects of the burden of musculoskeletal disorders.16–24
The overall methods for the GBD 2010 study are described in detail elsewhere25–28 and are summarised only briefly here, along with the specific methods for the work-related LBP analysis.
GBD 2010 methods
The GBD 2010 study sought to quantify all health loss due to all conditions in all countries (which were grouped into 21 regions) for 1990, 2005 and 2010 (only the 1990 and 2010 results are presented here). The main measure of disability was disability-adjusted life years (DALYs), calculated from estimates of deaths, years of life lost (YLLs) due to premature mortality and years lived with disability (YLDs). Since back pain is not a fatal condition, only YLDs contribute to the DALY estimates relevant to the burden arising from occupational ergonomic risk factors for LBP.
The burden (estimated as the population attributable fraction (PAF)) attributable to categorical exposures, which was the basis of the analysis presented here, was calculated by comparing exposure categories with a reference category for each age, sex, year and cause according to the following formula:
where RRi is the RR for exposure category i, Pi is the fraction of the population in exposure category i and n is the number of exposure categories.
This calculation required an estimate of the ‘counterfactual’ exposure, which for GBD 2010 was the theoretical-minimum-risk exposure, defined as the optimal exposure distribution in terms of effect on population health.25 For the occupational LBP study presented here, the theoretical-minimum-risk exposure distribution was that all individuals have the ergonomic factors of clerical and related workers (see below).
Definition of outcome
The definition of LBP is based on the work of Hoy and coworkers,29 being “...back pain (±pain referred into one or both lower limbs) that lasts for at least one day. The ‘low back’ is defined as the area on the posterior aspect of the body from the lower margin of the twelfth ribs to the lower gluteal folds”. There was no specific requirement that the pain be activity limiting.
Occupational exposures assumed to be aetiologically relevant causes of LBP were lifting, forceful movements, awkward postures and vibration. These have not been measured or characterised sufficiently for each country and region to allow direct measures to be the basis of exposure (and therefore the prediction of risk arising from these exposures) on the global scale required for the study. Instead, occupation was used as a proxy for ergonomic factors associated with occupation. Occupations were grouped into broad risk categories, determined in part by economic sector, judged to have similar physical and psychosocial exposures. They were also designed to be consistent with the occupational categories in the studies that provided risk information.
Estimates of the proportion of the working population employed in each occupation group in each country for 1990 and 2010 were based on data from the International Labour Organization (ILO) Labour Force. This information was available on the ILO website data search engine LABORSTA Internet.30 The proportion of the population which was working (the Economically Active Population (EAP)) was also obtained from the ILO LABORSTA Internet site. All information was available separately for males and females, and the EAP data were also available for each GBD 2010 age group. The ILO data came primarily from national censuses and secondarily from national Labour Force Surveys.
The LABORSTA database contains 8849 data points for at least one of the seven occupational categories for 133 of the 187 GBD countries. To get a complete time series for all countries, we used a three-step regression process, applied separately for each sex and each of the seven ISCO 1988 occupational categories. We also corrected for potential bias that arises due to the different ISCO classifications that were used to classify occupation types by including a dummy variable indicating whether ISCO Revision 1988 was used using the reference classification of ISCO Revision 1968. The first step involved a linear regression on the logit-transformed percentage and a set of covariates. The second step smoothed residuals across space and time to incorporate information for nearby countries and time. Finally, we used a Gaussian process regression that incorporated the variance of the data and model predictions. The estimates of the percentage of workers in each occupational category were then rescaled so that they summed to one by dividing each estimate by the sum of all the estimates. This was then multiplied by the EAP proportion to determine population-level exposure.
Information on risk estimates was obtained by conducting a systematic review of international literature and a meta-analysis of relevant results (manuscript in preparation). In brief, all relevant epidemiological studies of LBP risk by occupational group, published between 1990 and 2005, were sought. After initial screening on the basis of title and abstract, removal of duplicates and exclusion of those for whom a full paper in English was not obtainable, 161 papers remained. One hundred and fifty-two of these were excluded on the basis of relevance and methodological quality (eg, significant potential for selection or measurement bias or incomplete control of confounding), leaving nine relevant studies that provided data in a form suitable for calculation of the relative risk (or OR) for the outcome of interest in at least one of the exposure groups.31–38
The exposure category used to calculate the background rate was clerical and related workers because that occupation group was used as the referent group, on the basis of little to no exposure to the occupational factors noted above, in the majority of studies selected in the review. Where required and appropriate, the published data in the studies were regrouped or reanalysed to allow calculation of risk estimates that could be combined. Relative risk, OR from case–control sampling and prevalence OR were considered equivalent for the purposes of the meta-analysis. Meta-analysis was conducted using the MetaXL program.39 The same risk estimates were used for males and females and for all age groups (table 1). Further details are available from the authors.
The approach used to calculate uncertainty for the estimates is as described in the relevant capstone GBD 2010 paper. A simulation analysis was used in which 1000 draws were taken from the posterior distribution of exposure, RR, and each relevant outcome for each age, sex, country and year combination. The mean DALYs attributable to each risk factor and risk factor cluster from the 1000 draws were then computed. The 95% uncertainty intervals (95% UI) were calculated as the 2.5th and 97.5thcentiles of the 1000 draws.25
There were 21.8 million (95% UI 14.5–30.5 million) DALYs worldwide from LBP arising from ergonomic exposures at work (35% of all DALYs arising from occupational risk factors) in 2010. Of these, 13.5 million (95% UI 9.0–18.9 million) (61.9%) of the DALYs were in males and 8.3 million (95% UI 5.5–11.6 million) in females. The absolute burden was greatest in persons aged 35–55 years, with the per capita rates highest in persons aged 35–65 years (figure 1). Relative risk was about 3.7 in agriculture and related work, more than double that estimated in all other occupational groups (table 1).
The largest number of DALYs occurred in the regions with high populations—Asia East, Asia South, Asia Southeast and North Africa/Middle East. On a per capita basis of persons 15 years and over, the rate of DALYs per 100 000 persons varied considerably across regions, from a low of 208 (95% UI 134–297) in North America to a high of 540 (95% UI 317–836) in Central Sub-Saharan Africa. The highest rates were in areas where the workforce had much higher proportions of workers employed in agriculture and labouring occupations (tables 2 and 3).
The DALYs arising from occupational exposures were 28% of all DALYs from LBP for persons 15 years and over (ie, the PAF was 28%). The PAF varied considerably with age (highest in middle age and lowest in the elderly), sex (33% in males; 23% in females) and region. The region-specific PAFs were lowest in the high-income regions (14–17%) and highest in Central Sub-Saharan Africa (44%), Eastern Sub-Saharan Africa (42%) and Oceania (40%) (table 4).
There was a 22% increase in overall LBP DALYs arising from occupational exposures between 1990 and 2010, reflecting the increase in population over the two decades. When calculated on a per capita basis, there was a 14% fall over that time, on average. Most regions showed an average fall in per capita rates between 1990 and 2010, although the change varied from a decrease of 26% to an increase in 34%.
Low back pain arising from occupational ergonomic factors accounts for about one-third of all disability arising from the occupational risk factors included in the GBD 2010 project. The burden was considerable in both sexes, all age groups and all regions. The absolute burden increased considerably from 1990 to 2010 in line with population increases and an aging population, but decreased an average of 14% when calculated on a per capita basis.
The changes over time primarily reflect changes in population size and employment distribution since the relative risks used to calculate the PAF measure were the same for 1990 and 2010. These changes in employment distribution were commonly a proportionate decrease in employment in high-risk occupations (agriculture work, machine operation and labourers) that generally result in substantial exposure to ergonomic factors linked to LBP occurrence (lifting, forceful movements, awkward postures and vibration).
The results are qualitatively similar to those found in the previous comparative risk analysis (CRA) project, which looked at the burden arising from occupational risk factors in 2000.40 The similarity is not surprising given that the general approach to analysis was similar. The PAF estimate of 26% for 2010 is less than the estimate of 37% from the CRA 2000 study, but still substantial (the estimate of the PAF for 1990 was 31%). The difference arises from two main areas. First, different relative risks, based on a different referent group, were used to estimate the PAF. For the analysis presented here, the relative risk estimates were based on a meta-analysis of useable measures in the published literature, and the referent group was clerical and related workers, whereas in the CRA 2000 the referent group was managers and professionals. The change in referent group was made for the current analysis because clerical and related workers were the referent group used in most of the studies able to be included in the meta-analysis. Nevertheless, prolonged sitting with low latitude to vary tasks and postures during the work shift, especially in chairs with poor lumbar support and other ergonomic features, has also been identified as a potential LBP risk factor, so the referent group may also not be an optimal one and might have produced some bias towards the null. Second, there have been changes in employment distribution in the 10 years between the studies (eg, decreased proportions employed in agriculture in most regions).
There was a huge difference in absolute numbers of DALYs estimated to arise from LBP, with 818 000 estimated for the CRA 2000 project and 21.8 million estimated for the current project, about a 25-fold increase. A small component of this difference is population increases, but primarily the difference arises from improvements in the methodologies used to estimate the DALYs from LBP in the GBD 2010 project compared with the CRA 2000 project, including better estimates of disease prevalence and different weights applied to the YLD calculations.
Even when considering the substantial uncertainty, the results provide strong support for the need to identify, develop and implement effective interventions, primarily by decreasing exposure to the causative ergonomic risk factors.41 ,42 Note that the policy implication of the findings is not to shift all workers to clerical occupations but rather to remove or limit the LBP risks that are associated with the other occupational categories. When selecting and implementing such interventions, it is important to keep in mind that the relationships and interactions between various risk factors and LBP incidence may not be straightforward43 and many current interventions—especially training and other administrative measures—do not appear to be supported by a strong evidence base.44 ,45
The analysis presented here only addresses LBP. This is because there were few studies that provided useable estimates of relative risk for episodes of pain in the thoracic or cervical spine. Also, no attempt was made to include other possible musculoskeletal pathology arising from occupational exposures. This was because of one or more of the criteria of evidence regarding these being considered insufficient to meet the stringent requirements of the GBD 2010 study, the required relative risk measures being insufficient and/or the available exposure measures being insufficient. The burden estimates provided here are therefore an underestimate of the total burden due to musculoskeletal problems arising from occupational exposures.
There are a number of limitations to the analysis presented here, some of which arise out of the requirements of the GBD 2010 project and some of which are relatively specific to the LBP study. The relative risks used in the estimate of the PAFs were based on a meta-analysis of studies that used somewhat different methodologies. The definition of the various employment groups and LBP outcome measures was slightly different or unclear, and the researchers did not necessarily specify or take explicitly into account possible control measures. The definition of LBP commonly varies somewhat between relevant studies and/or is often not defined clearly. The definition used in the current study was that adopted for use in the GBD 2010 study and was reasonably consistent with that used in the studies that contributed to the relative risk measures.
Most of the studies that contributed to the relative risk measures were based in developed countries rather than in developing countries, and the same relative risks were applied to all persons in a given occupational group, regardless of the region in which they lived. This was because there was insufficient information to allow potential differences in exposure in different regions to be taken into account. It is reasonable to accept that ergonomic stressors (note that for the purposes of this paper ergonomic risk factors primarily refers to occupational physical activity rather than to design aspects of the workplace) on the back are broadly similar in the same occupation group regardless of region, but differences could arise due to varying use of mechanisation, for example, or the use of various control measures.40 If these differences were of similar relative magnitude in all or most occupations in a region, then the relative risks would be similar across regions regardless of differences in absolute risk between regions. However, some variation between occupations in adoption of mechanisation or other preventive ergonomic interventions would not be unexpected. Similar arguments apply to using the same relative risks for 1990 and 2010—the studies that contributed data to the meta-analysis mainly analysed data collected in the late 1980s and so presumably are more directly relevant to the 1990 estimates. It is likely that there have been improvements in machinery technology, health and safety standards and ergonomic awareness that have reduced the risk of LBP due to occupational exposures. However, the effect of such possible changes on the study results is impossible to quantify worldwide.
Occupation was used as a proxy for exposure to the work-related ergonomic factors suggested in the literature as affecting the risk of developing LBP. Ideally, measures of the exposure themselves would be available to be combined with appropriate estimates of the relative risk of developing LBP as a result of those exposures. Unfortunately, national information on prevalence of occupational exposures by job group is currently not available, even in developed countries. On the other hand, an advantage of the approach used is that it avoids the question of unmeasured interaction among exposures, as each occupation represents the combined effect of all exposures involved, both physical and psychosocial.40 Additional potential sources of error include using the same relative risk measures across ages and genders; not including members of the informal workforce, which can be considerable, especially in developing countries; and not including child workers. Finally, the approach assumes an instantaneous effect from exposure, rather than the accumulation of LBP risk over time from occupational and non-occupational sources. In reality, this accumulated risk will not be removed instantaneously once exposure is changed. Suggestions for future work in this field and implications for future disease burden studies are summarised in online supplementary table S1.
This study has shown that LBP arising from occupationally related ergonomic exposures is an important cause of disability in working persons, both as a proportion of all work-related burden and as a proportion of community health arising from LBP. This burden varies across age groups, sexes and regions, largely reflecting the different occupation distributions. There has been some apparent reduction from 1990 to 2010, primarily due to changes in occupation distribution. Focus on interventions to reduce exposure to the causative ergonomic risk factors, and work to improve the available exposure and risk information, especially in developing countries, would be appropriate responses to these findings.
The authors would like to thank Dr Kevin McGeechan for assistance with consideration of statistical issues.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Handling editor Tore K Kvien
Contributors TD undertook the systematic review work that underpinned the risk measures used in the analysis, conducted the meta-analysis that produced the final risk measures, worked with the core GBD team to produce the estimates and wrote the first draft and revision of the paper. GJ, JO and EP undertook the systematic review work that underpinned the analysis. TV, GF and SL were members of the core GBD team who undertook the central analysis of the burden of disease estimates presented in the manuscript. LP contributed to the early conceptual stage of the project and had a significant input into the interpretation of the results. All authors have read and revised the paper for important intellectual content and given final approval of the version to be published.
Competing interests TD was supported in part by funding from the National Occupational Health and Safety Commission (now Safework Australia).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The data provided in this paper arose from the GBD 2010 project. Data from GBD 2010 are available from the IHME website (see http://www.healthmetricsandevaluation.org/gbd). Some more detailed data may be available from the authors upon request.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.