Biomechanical constraints remain major risk factors for low back pain. Results from a prospective cohort study in French male employees

(1)

Clinical Study

Biomechanical constraints remain major risk factors for low back pain.

Results from a prospective cohort study in French male employees

Aline Ramond-Roquin, MD

^a,b,

* , Julie Bodin, MSc

^a

, C eline Serazin, MSc

^a

, Elsa Parot-Schinkel, MD

^a,c

, Catherine Ha, MD

^d

, Isabelle Richard, PhD

^a,e

, Audrey Petit Le Manach, MD

^a,f

, Natacha Fouquet, MSc

^a,d

, Yves Roquelaure, PhD

^a,f

aLaboratory of Ergonomics and Epidemiology in Occupational Health, Faculty of Medicine, University of Angers, Rue Haute de Reculee, 49045 Angers Cedex 01, France

bDepartment of General Practice, Faculty of Medicine, University of Angers, Rue Haute de Reculee, 49045 Angers Cedex 01, France

cClinical Research Centre, University Hospital of Angers, 4, rue Larrey, 49933 Angers Cedex, France

dDepartment of Occupational Health, French Institute for Public Health Surveillance, 12 rue du Val d’Osne, 94415 Saint-Maurice Cedex, France

eRegional Centre for Rehabilitation of Angers, rue des Capucins, B.P. 40329, 49103 Angers Cedex 02, France

fDepartment of Occupational Health, Faculty of Medicine, University Hospital of Angers, 4, rue Larrey, 49933 Angers Cedex, France Received 4 July 2011; revised 12 February 2013; accepted 22 May 2013

Abstract BACKGROUND CONTEXT: Low back pain (LBP) is a major public health problem, with a considerable impact on workers.

PURPOSE: To model the risk of LBP in the male general working population.

STUDY DESIGN/SETTING: Repeated cross-sectional surveys in a wide occupational setting.

PATIENT SAMPLE: A random sample of 2,161 men working in various occupations in a French region participated in a first survey in 2002, and 1,313 of these (60.8%) participated in a second survey in 2007.

OUTCOME MEASURE: The self-reported prevalence of LBP during the previous week in the second survey.

METHODS: Twenty-one biomechanical, organizational, psychosocial, and individual factors were assessed in the first survey. The association between these potential risk factors and the prevalence of later LBP (in the second survey) was studied, using multistep logistic regression models.

RESULTS: Three hundred ninety-four men reported LBP in the second survey (prevalence 30.0%). The final multivariate model highlighted four risk factors: frequent bending (odds ratio [OR], 1.45, 95% confidence interval [CI], 1.07–1.97 for bending forward only; and OR, 2.13, 95% CI, 1.52–3.00 for bending both forward and sideways), driving industrial vehicles (OR, 1.35; 95% CI, 1.00–1.81), working more hours than officially planned (OR, 1.38; 95% CI, 1.05–

1.81), and reported low support from supervisors (OR, 1.35; 95% CI, 1.02–1.79).

CONCLUSIONS: These results emphasize that biomechanical factors remain worth considering, even when psychosocial factors are taken into account, and provide a significant contribution to pre- ventive strategies. Ó2015 Elsevier Inc. All rights reserved.

Keywords: Low back pain; Occupational health; Risk factors; Cohort studies; Logistic models

Introduction

Low back pain (LBP) is a major public health problem.

Although this condition can reveal specific and severe

diseases, nonspecific LBP is much more frequent[1]. The annual prevalence of LBP is very high in Western coun- tries; it was estimated to be 54% in men and 57% in women

FDA drug/device approval status: Not applicable.

Author disclosures: AR-R:Fellowship Support: CNAM-TS: French National Health Insurance Fund (B, Paid directly to institution).JB:Noth- ing to disclose.CS:Nothing to disclose.EP-S:Nothing to disclose.CH:

Nothing to disclose.IR:Nothing to disclose.APLM:Nothing to disclose.

NF:Nothing to disclose.YR:Nothing to disclose.

The disclosure key can be found on the Table of Contents and atwww.

TheSpineJournalOnline.com.

* Corresponding author. Laboratoire d’Ergonomie et d’Epidemiologie en Sante au Travail (LEEST), Faculte de Medecine Rue Haute de Reculee 49045 ANGERS Cedex 01, France. Tel.: (33) 24-173-5930; fax: (33) 24- 173-5881.

http://dx.doi.org/10.1016/j.spinee.2013.05.040

(2)

in the French general population from 30 to 59 years in 2002[2]. Low back pain results in significant levels of individual pain and disability and is a frequent reason for seeking care [3,4]. The working population is particularly affected by this condition, leading to work absenteeism, resulting loss of productivity and hence considerable socioeconomic costs for society [4–7].

Many studies have been undertaken to identify risk factors for LBP in the last 20 years, and certain types of occupational biomechanical exposures have been causally linked (heavy load lifting, ‘‘whole-body’’ vibrations, bending and twisting, and so forth)[8–13]. However, the respec- tive role of each of them in the incidence of LBP is still a matter of debate. First, the definition of LBP is not homo- geneous in the literature, and this sometimes gives rise to conflicting results[14,15]. Second, the very high frequency of LBP in the general population and its recurrent evolution make the concept of incidence very difficult to apply or even unsuitable[16]. Third, most studies about risk factors for LBP are cross-sectional or case-control studies, which prevent any causal interpretation [12,13]. Fourth, some studies have focused on specific occupational groups (nurses, automobile industry workers, bus drivers, and so forth), and their external validity is therefore insufficient for the generalizability of their conclusions [11,12]. Fifth, the studies often assess a limited number of risk factors si- multaneously, although LBP is known to be multifactorial.

Last, certain factors that have been more recently highlighted and that might be worthy of interest such as psychosocial and individual factors are seldom integrated into the research, and their assessment raises methodological issues [17–19].

The current research on LBP focuses mainly on the prevention of transition from acute to chronic LBP or on the

reduction of the consequences of chronic LBP because the latter is responsible for most of the costs associated with LBP [20,21]. However, these goals are not easy to achieve, and reduction of the prevalence of LBP, whatever its duration and its characteristics, might be the best way to limit the human, medical, and socioeconomic costs involved.

From 2002, a network of occupational physicians in a French region (Pays de la Loire) has recruited a vast regional cohort of employees and conducted an extensive assessment of their occupational conditions. These workers have been the subject of a subsequent evaluation, concerning their health status in terms of potential musculoskeletal disorders, including LBP, about 5 years later [22].

The main objective was to assess the relative impact of biomechanical, organizational, psychosocial, and individual factors on the risk of later LBP in a cohort of male workers exposed to various levels of work constraints.

The second objective was to compare the risk models of LBP in the workers with and without previous LBP.

Methods

Design and population

This prospective study was based on two successive surveys of a large sample of workers in the French Pays de la Loire region. The economic structure of this region (5% of the French working population) is diversified and similar to that of most French regions[23].

First survey

In 2002, all the occupational physicians working in the Pays de la Loire region (n5460) were asked to participate in this study. It consisted in randomly selecting a sample of male and female employees, aged from 20 to 59 years, whatever their economic sector, from the workers undergoing a mandatory scheduled annual health examination (in France, all workers, including temporary and part-time workers, are the subjects of such medical surveillance, except for self- employed people). Eighty-three occupational physicians vol- unteered to participate. They were representative of the 460 occupational physicians of the region in terms of medical practice, working time, geography, and economic sectors covered. Their participation rates were quite similar in the different types of companies surveyed (private company 18.4%, public services 17.2%, hospitals 14.8%, and agricul- tural sector 14.3%; the occupational physicians working in private companies corresponding to 88% of our sample).

Workers were selected at random after a two-stage sampling procedure: first, 15 to 45 half-days of scheduled examinations for each physician were chosen for sampling by the investiga- tors (from 15 for those working part-time, to 45 for those working full-time); next, using random sampling tables, each Context

Many biological, mechanical, and psychosocial factors are felt to impact low back pain.

Contribution

In this repeated survey with a five-year gap, the authors found that frequent bending, driving industrial vehicles, working longer hours than expected, and poor support from supervisors were associated with low back pain.

Implications

While the study design does not allow for discussion of causation and there may well be undetected causa- tive factors for low back pain that were not assessed, the findings are in line with several other studies in the field.

—The Editors

(3)

physician included 1 out of 10 workers from the schedule on the half-days of worker examinations under consideration.

The workers selected completed a self-administered questionnaire to assess their working conditions.

Second survey

In 2007, another self-administered questionnaire was mailed to all the subjects, some questions of which aimed at screening musculoskeletal disorders, including LBP. If they had moved, their new address was sought, sometimes communicated either by their occupational physician or by the French public postal services. In cases of nonresponse, they were sent two successive reminder letters with the self- administered questionnaire.

Each subject provided informed written consent to participate in this study, which was granted authorization by the National Commission on Data Protection (Commission Na- tionale Informatique et Libertes), first in 2001, then in 2006.

Outcome

The second self-administered questionnaire, sent out in 2007, included a ‘‘Nordic’’ questionnaire that had been de- veloped and validated to screen workers for the main musculoskeletal disorders [24]. In this questionnaire, the subjects were asked if they had experienced any aching, discomfort, pain, or numbness in the low back (located with a body mannequin, including the lumbar and the gluteal regions), hereafter designated as LBP, during the previous week. The outcome of our study was ‘‘the prevalence of LBP in the second survey,’’ the prevalence being defined by the report of symptoms during the previous week.

Risk factors

Four categories of factors were considered to be potential risk factors for subsequent LBP, the first three being related to occupational conditions. They were all assessed by the first self-administered questionnaire (2002–2005).

1. Biomechanical factors: Frequent bending (O2 hours daily), forward or sideways or both;high physical demands, that is, score 14 or more on the Borg perceived exertion rating scale graduated from 6 to 20 [25]; driving industrial vehicles (tractors or forklift trucks, with exposure to whole-body vibrations) at least occasionally; driving nonindustrial vehicles (cars, buses, or trucks, without exposure to whole- body vibrations) more than 4 hours daily; andheavy lifting, composite criteria, more than 25 kg at least occasionally or between 10 and 25 kg more than 2 hours daily.

2. Organizational factors:Working more hoursthan officially planned; not knowing the daily tasks at the beginning of the day;having several occupational positions; havingvariable pay; working withcolleagues

on fixed-term contracts; havingwork rate constraints, that is, production time limits or paced work or both;

having market constraints (work rate influenced by public or customers); and not being able or allowed tostop or change the taskfor 10 minutes every hour.

3. Psychosocial factors: the five subscales of Karasek job content questionnaire[26,27]were used to assess work-related psychological factors: low job decision authority,low skill discretion,high psychological demands,low support from coworkers, andlow support from supervisors. For each dimension independently, subjects in the quartile with the worst scores in the cohort of men were considered to be exposed to the factor.

4. Individual factors:Agein four groups (20–29, 30–39, 40–49, and 50–59 years),heightin the tallest quartile of the cohort of men, and overweight or obesity according to the World Health Organization definition (body mass index$ 25 kg/m²).

As described previously, frequent bending and having work rate constraints were qualitative variables with four modalities. In fact, each of these variables derived from the combination of two different binary questions in the first survey: frequent bending forward and frequent bending sideways on the one hand and having production time limits and having paced work on the other hand.

The distributions of their responses were too strongly linked to be considered independently. These questions were therefore combined together into variables with four modalities.

Statistical analysis

The frequency of LBP reported by the male workers in the second survey was first calculated overall and then according to their initial occupational characteristics.

The modeling process was then based on a multistep binary logistic model, chosen to limit the number of potential risk factors in the final multivariate model. The effects of potential risk factors were estimated by means of odds ra- tios (ORs) and their 95% confidence intervals (CIs).

In the first step, univariate binary logistic regression models were used to estimate the unadjusted links between each potential risk factor and the outcome. The factors linked to the outcome in these models with a p value lower than .20 were selected. In the second step, the factors previously selected within each group of factors (biomechanical, organizational, psychosocial, and individual) were introduced in ‘‘within-group’’ multivariate models (four different models were then constructed). In the third step, the factors selected in these within-group models were en- tered in a global multivariate model. In the second and third steps, selection of the variables was based on the method from Hosmer and Lemeshow [28], and selection was performed manually. Factors were selected either because they

(4)

were associated with the outcome (p!.10 in the second step and then p!.05 in the third step) or because they were considered to be important confounding factors (factors whose deletion modified the estimated coefficient of another variable in the model by at least 10%). In the fourth step, whether age had been selected as a significant risk factor in the third step, interactions between each selected factor and age were introduced systematically one by one and submitted to the selection process. The goodness of fit of the models was assessed by the statistic of Hosmer and Le- meshow and by individual regression diagnostics.

For all variables except one (heavy lifting), few (!5%) observations had missing values and were deleted from the models comprising these variables. For the variable heavy lifting, 13% of the observations had missing values.

People who had not answered this question presented certain characteristics (high proportion of men older than 50 years, with short job tenure, precarious employment contracts, various tasks in their jobs, and so forth) that may have subjected (at least some of) them to irregular and/or hardly quantifiable exposure to heavy lifting, and this could have led to missing answers. Because these missing values might have made sense, the variable heavy lifting was considered as a qualitative variable with three modalities: exposed, not exposed, or missing value.

Finally, the whole modeling process was performed again considering a ‘‘new’’ study population, restricted to the male workers without LBP reported in the first survey.

The aim was to reduce the proportion of chronic or recurrent LBP in the prevalence of LBP reported in the second survey. Adding the potential reporting of LBP in the first survey as a covariate in the model was a possible alternative but might have led to an overadjustment. The strategy chosen is usual for such frequent and recurrent symptoms [18,29,30].

The threshold of statistical significance was set at .05 unless otherwise indicated.

As men and women present major differences in their job tasks and occupational exposure, they were studied sep- arately. Stratifying analyses by sex is a strategy favored in the literature about musculoskeletal disorders [31]. Only the results for men are presented here.

Data were analyzed using SAS version 9.2 (SAS Insti- tute, Cary, NC, USA).

Results Population

Fewer than 10% of the selected workers failed to participate, with 2% having refused inclusion in the study and 5%

not been seen by their occupational physician as planned in the protocol. Finally 3,710 employees: 2,161 men and 1,549 women were included between 2002 and 2005.

The 2,161 men corresponded to 0.375% of the male employees aged from 20 to 59 in the region. These men worked mainly in the service industries (50.0%), in the meat and manufacturing industries (39.8%) and, more rarely, in the construction industry (8.8%) and agriculture (1.4%). When compared with the male working population of the region, according to the French 1999 census, the sample overrepre- sented moderately young and blue-collar workers[32]. How- ever, the distribution of occupations was close to that of the regional male workforce, except for the rare occupations not surveyed by occupational physicians, for example, shop- keepers and self-employed craftsmen (data not shown).

Of the 2,161 men who had participated in the first survey, 1,313 (60.8%) completed the second self-administered questionnaire between 2007 and 2009 (Figure). In comparison with the nonrespondents in the second survey, the respondents were (at the moment of the first survey) moderately older (mean 3 years), with longer job tenure, and were more often executives and less often blue-collar workers. However, there was no difference between the

Figure. Flow diagram of the male participants in the study.

(5)

respondents and the nonrespondents regarding the prevalence of LBP reported in the first survey (Table 1).

Risk factors for LBP

The prevalence of LBP reported by the men in the second survey was 30.0% (n5394). There was a significant difference between sufferers and nonsufferers in the distribution of socio-occupational groups in the first survey, with 6.4% executives, 26.9% technicians and intermediate occupations, 7.6% lower grade white-collar workers, and 59.1%

blue-collar workers among sufferers and 13.5%, 27.1%, 8.0%, and 51.3% among nonsufferers, respectively (p5.0013). However, no difference could be found regarding their length of job tenure and type of employment contract (data not shown).

The univariate analysis showed that frequent bending, some work rate constraints, and reported low support from supervisors were the factors most strongly linked to the probability of subsequent LBP (p!.01). Nine other factors were associated with the outcome (p!.20) and therefore selected for the multivariate analyses (Table 2).

The final multivariate model comprised four independently significant variables. Bending for more than 2 hours every day was the factor most strongly linked to the outcome, with ORs (CIs) of 1.45 (1.07–1.97) for bending forward only and 2.13 (1.52–3.00) for bending both forward and sideways. Three other factors were associated with the outcome, that is, driving industrial vehicles, working more hours than officially planned, and reported low support from supervisors. They presented somewhat similar ORs (CIs) for the risk of subsequent LBP, 1.35 (1.00–

1.81), 1.38 (1.05–1.81), and 1.35 (1.02–1.79), respectively.

A low level of skill discretion behaved as a confounding factor and had to be kept in the final model to allow adjust- ment for the other factors (Table 2).

Out of the 1,313 men who participated in the 2 successive surveys, 381 had reported LBP in the first survey and 1 had not answered this question. Finally, 931 men reported no LBP in the first survey and participated in the second (Figure). The prevalence of LBP reported in the second survey in this subgroup was 21.2%. The univariate analysis showed that frequent bending and working more hours than officially planned were the factors that most strongly linked to the outcome (p!.01). Seven other factors were selected for the subsequent multivariate analysis. The final multivariate model related to this subgroup comprised two significant variables, that is, frequent bending and working more hours than officially planned. Furthermore, it had to be adjusted for the work rate constraints because of a significant confounding effect (Table 3).

Compared with the previous results regarding the whole cohort, the two significant factors in this subgroup model presented a stronger multivariate association with the outcome, with ORs (CIs) of 1.66 (1.10–2.52) for bending forward, 2.67 (1.70–4.20) for bending both forward and sideways, and 1.69 (1.16–2.47) for working more hours than officially planned. Driving industrial vehicles and reported low support from supervisors presented no multivariate association with the outcome in this subgroup (Table 3).

Discussion

Several variables were found to be independently linked to the risk of subsequent LBP in this large and

Table 1

Comparison between the male respondents and nonrespondents in the second survey

Characteristics in the first survey Respondents Nonrespondents p*

Age, mean (standard deviation) 39.9 (10.0) 36.3 (10.6) !.0001

Length of job tenure, n (%) !1 y 125 (9.6) 145 (17.3) .0019

1 to 2 y 200 (15.4) 134 (16.0)

3 to 10 y 429 (33.1) 296 (35.2)

O10 y 544 (41.9) 265 (31.5)

Total 1,298 (100.0) 840 (100.0)

Insecure employment contract, n (%)^y 97 (7.4) 139 (16.5) !.0001

Socio-occupational group, n (%) Executives 144 (11.0) 66 (7.9) .0004

Technicians and intermediate occupations 355 (27.1) 185 (22.1)

Lower grade white-collar workers 105 (8.0) 82 (9.8)

Blue-collar workers 705 (53.9) 504 (60.2)

Total 1,309 (100.0) 837 (100.0)

Economic sector, n (%) Service industries 658 (50.2) 419 (49.8) .67

Meat and manufacturing industries 513 (39.2) 340 (40.4)

Construction 116 (8.9) 73 (8.7)

Agriculture 22 (1.7) 9 (1.1)

Total 1,309 (100.0) 841 (100.0)

Prevalence of LBP in first survey, n (%) 381 (29.2) 233 (27.5) .47

LBP, low back pain.

Bold entries signify statistical significance.

* p Values from chi-square tests except for age (ttest) and length of job tenure (Wilcoxon test for ordinal values).

y Insecure contracts: fixed-term contracts, temporary employment, and trainees (vs. secure contracts: civil servants and open-ended contracts).

(6)

diverse cohort of male employees. These variables be- long to three of the four groups of variables taken into account in this study, that is, biomechanical, organizational, and psychosocial variables. Bending for more than 2 hours every day was the factor most strongly associated with the outcome, with a high excess risk for those bending forward only and an even higher excess risk for those bending both forward and sideways. Be- sides, men who reported driving industrial vehicles, working more hours than officially planned, or low support from supervisors at baseline were also at elevated risk of LBP at follow-up.

The last analyses suggested that frequent bending and working more hours than planned could be more strongly linked to the risk of subsequent LBP in subjects without prior LBP, whereas driving industrial vehicles and reported low support from supervisors may be more associated with

the risk of subsequent LBP among those who have already experienced LBP.

These results are based on a large cohort of male employees, fairly representative of the diverse economic sectors and occupations of the male workforce of a French region. The percentage of respondents to the second survey might be considered moderate (60.8%), but it is fairly sat- isfactory in view of the time interval between the two surveys and above all the obstacles to following-up samples of healthy workers. There were slight socioeconomic differences between the respondents and the nonrespondents, with lower percentages of young and blue-collar workers among the respondents. Such differences are frequently observed in studies based on the voluntary participation of healthy subjects [33]. These categories of workers were also the most exposed to insecure employment contracts, which may have contributed to limiting their follow-up.

Table 2

Univariate and multivariate prospective modeling of the risk of LBP in the cohort of 1,313 male employees in the French Pays de la Loire region Univariate analyses* Multivariate analyses^y

Risk factors in the first survey OR (95% CI) p OR (95% CI) p

Biomechanical factors Frequent bending No (reference) 1.00 !.0001^z 1.00 .0001

Forward 1.61 (1.20–2.16) 1.45 (1.07–1.97)

Sideways 2.26 (0.98–5.23) 1.78 (0.74–4.27)

Both 2.10 (1.51–2.93) 2.13 (1.52–3.00)

High physical demands 1.28 (0.99–1.65) .060^z

Driving industrial vehicles^x 1.41 (1.06–1.86) .017^z 1.35 (1.00–1.81) .047

Driving nonindustrial vehicles^x 1.28 (0.89–1.85) .19^z

Heavy lifting No (reference) 1.00 .060^z

Missing 1.31 (0.92–1.86)

Yes 1.35 (1.03–1.77)

Organizational factors Working more hours 1.25 (0.96–1.61) .092^z 1.38 (1.05–1.81) .021

Not knowing the daily tasks 1.21 (0.86–1.70) .27

Several occupational positions 0.97 (0.76–1.25) .82

Variable pay 1.34 (1.02–1.76) .039^z

Colleagues in fixed-term contract 1.02 (0.78–1.32) .90

Work rate constraints No (reference) 1.00 .0078^z

Production time limits 1.26 (0.97–1.65)

Paced work 1.17 (0.64–2.17)

Both 1.90 (1.31–2.75)

Market constraints 1.00 (0.78–1.27) .97

Not stopping or changing task 1.08 (0.84–1.39) .55

Psychosocial factors Low job decision authority^{ 1.09 (0.77–1.55) .63

Low skill discretion^{ 1.38 (1.05–1.81) .022^z 1.28 (0.95–1.72) .11

High psychological demands^{ 0.86 (0.64–1.16) .32

Low support from coworkers^{ 1.30 (0.96–1.74) .087^z

Low support from supervisors^{ 1.48 (1.13–1.94) .0043^z 1.35 (1.02–1.79) .038

Individual factors Age 20 to 29 y (reference) 1.00 .21

30 to 39 y 0.98 (0.69–1.40)

40 to 49 y 0.87 (0.61–1.24)

50 to 59 y 1.24 (0.85–1.81)

HeightO179 cm 1.03 (0.78–1.35) .85

Overweight/obesity 1.18 (0.93–1.50) .18^z

LBP, low back pain; OR, odds ratio; CI, confidence interval.

* Crude ORs, 95% CIs, and p values from univariate binary logistic regression models.

y Adjusted ORs, 95% CIs, and p values from the final multivariate binary logistic regression model (n51,255).

z p!.20 in univariate analysis.

x Industrial vehicles: tractors or forklift trucks; nonindustrial vehicles: cars, buses, or trucks.

{ Score belonging to the quartile with the worst scores, in the cohort of the 1,313 men, in the first survey (scores for the five dimensions:!32,!32,O24,

!12,!11, respectively).

(7)

Indeed, the second survey period coincided with a major economic crisis in the region between 2008 and 2009, during which the regional salaried workforce declined by 3.4%, and even by 33.7% in temporary employment agencies, according to the French Economic Institute [34]. Even if blue-collar workers presented differences in their expositions to psychosocial risk factors, in comparison with other categories, there were only minor differences between respondents and nonrespondents regarding these factors, with a higher rate of low job decision authority among the nonrespondents but no difference for the other psychosocial factors. It seems therefore unlikely that the psychosocial profile of the workers could have specifically influenced their participation in the second survey and that the role of psychosocial factors could have been underesti- mated because of it. Finally, the absence of difference in

terms of economic sector and of prevalence of LBP reported in the first survey between respondents and nonrespondents was another reassuring element in terms of nonresponse bias.

A selection bias linked to the ‘‘healthy worker effect’’

cannot be excluded, particularly at inclusion. This bias often makes it difficult to extrapolate the results from occupational samples to more general populations. However, great efforts were made to contact as many subjects as possible for the second survey, including those who had ceased working. This might have contributed to the reduction of the healthy worker effect.

The risk factors were evaluated by a self-administered questionnaire, which might have led to classification bias.

However, the assessment methods considered as reference methods (such as direct observation) are known to present

Table 3

Univariate and multivariate prospective modeling of the risk of LBP in the cohort of 931 male employees in the French Pays de la Loire region without initial LBP

Univariate analyses* Multivariate analyses^y

Risk factors in the first survey OR (95% CI) p OR (95% CI) p

Biomechanical factors Frequent bending No (reference) 1.00 .0001^z 1.00 .0002

Forward 1.57 (1.06–2.34) 1.66 (1.10–2.52)

Sideways 2.40 (0.80–7.15) 1.87 (0.56–6.18)

Both 2.56 (1.65–3.96) 2.67 (1.70–4.20)

High physical demands 1.15 (0.81–1.63) .44

Driving industrial vehicles^x 1.25 (0.85–1.82) .25

Driving nonindustrial vehicles^x 1.44 (0.88–2.34) .15^z

Heavy lifting No (reference) 1.00 .35

Missing 1.30 (0.82–2.06)

Yes 1.25 (0.87–1.79)

Organizational factors Working more hours 1.60 (1.12–2.29) .0099^z 1.69 (1.16–2.47) .0063

Not knowing the daily tasks 1.25 (0.80–1.95) .33

Several occupational positions 0.73 (0.52–1.04) .079^z

Variable pay 1.34 (0.94–1.93) .11^z

Colleagues in fixed-term contract 1.06 (0.75–1.49) .74

Work rate constraints No (reference) 1.00 .012^z 1.00 .20

Production time limits 1.39 (0.97–1.99) 1.19 (0.83–1.72)

Paced work 0.79 (0.30–2.10) 0.80 (0.29–2.17)

Both 2.13 (1.31–3.46) 1.69 (1.01–2.84)

Market constraints 1.09 (0.79–1.50) .60

Not stopping or changing task 1.16 (0.84–1.62) .37

Psychosocial factors Low job decision authority^{ 1.04 (0.65–1.67) .87

Low skill discretion^{ 1.30 (0.90–1.88) .16^z

High psychological demands^{ 0.78 (0.52–1.19) .26

Low support from coworkers^{ 0.92 (0.60–1.41) .70

Low support from supervisors^{ 1.31 (0.91–1.89) .15^z

Individual factors Age 20 to 29 y (reference) 1.00 .30

30 to 39 y 0.82 (0.53–1.29)

40 to 49 y 0.73 (0.46–1.16)

50 to 59 y 1.09 (0.67–1.76)

HeightO179 cm 0.98 (0.67–1.42) .91

Overweight/obesity 1.27 (0.92–1.74) .14^z

LBP, low back pain; OR, odds ratio; CI, confidence interval.

* Crude ORs, 95% CIs, and p values from univariate binary logistic regression models.

y Adjusted ORs, 95% CIs, and p values from the final multivariate binary logistic regression model (n5889).

z p!.20 in univariate analysis.

x Industrial vehicles: tractors or forklift trucks; nonindustrial vehicles: cars, buses, or trucks.

{ Score belonging to the quartile with the worst scores, in the cohort of the 1,313 men, in the first survey (scores for the five dimensions:!32,!32,O24,

!12,!11, respectively).

(8)

certain important limitations, in particular insufficient consideration for intrasubject (over time) and intersubject variabilities [35]. Moreover, only self-administered ques- tionnaires have been validated in the assessment of some variables such as the psychosocial factors[17].

We chose not to treat certain important occupational characteristics such as length of job tenure and socio- occupational group as potential risk factors in the risk models. They are indeed very strongly associated with other factors that were taken into account, for example, length of job tenure with age and socio-occupational groups with biomechanical factors [36]. Adding them to the models would have led to problems of collinearity.

All the potential predictors included in these analyses have been suggested as increasing the risk of subsequent LBP in the literature. Given the high number of potential risk factors, a selection procedure was undertaken to make the final model more concise and readable. When no strong hypothesis can lead the variable selection, an exploratory strategy such as a stepwise procedure is often preferred.

This was later conducted cautiously, with manual selection based both on p values and on variation of the estimated co- efficients resulting from variable deletion. Besides, the number of candidate predictors before the stepwise selection was lower than the maximum generally recommended to avoid overfitting[37].

However, this method presents some limits. The selection procedure between numerous risk factors can lead to over- estimation of the role of the factors selected [37]. Besides, the stepwise procedure may be influenced by collinearity, which characterizes a situation in which several ‘‘indepen- dent’’ variables of a regression analysis are strongly associated. In this case, the selection might be highly dependent of the own configuration of the study sample[37]. In our study, some occupational expositions were moderately associated.

However, no variance inflation could be observed, with CIs remaining reasonably large throughout the modeling procedure, what is a reassuring element regarding the risk of problematic collinearity. Moreover, several alternative multivariate models, adding one by one the variables not selected by the stepwise procedure, were fitted. They did not provide any decisive argument suggesting that a factor could or should be added, deleted, or replaced by another one. Notwithstanding, some factors may have been selected in the model rather than others, for other reasons than a stronger causal link with the outcome, for example, because of a more accurate measurement method.

Most studies on LBP in workers are cross sectional;

therefore, the prospective design is one of the strengths of this study. Longitudinal studies (with continuous data collection) are very difficult to carry out in occupational settings, as in the general population, and repeated cross- sectional studies on a closed cohort are the most often chosen alternative to obtain prospective data[30,36,38,39].

The diversity and the great number of the risk factors taken into account constituted another main strength of this

study because it is a necessary condition to obtain adjusted estimates of excess risk. It might nevertheless have been worth assessing and testing other individual factors such as smoking, extra-occupational physical activities, and individual psychosocial factors (anxiety, depression, family support, traumatic life events, social contact, social participation, and so forth) for their potential influence on the risk of subsequent LBP. A recent review concluded that current and former smokers had a higher prevalence of LBP than those who had never smoked but that the association was modest [40]. The results regarding extra- occupational physical activities are inconsistent in the literature, with most studies showing no association with LBP and certain others showing a favorable effect of physical activity but an unfavorable effect of some vigorous sports [41,42]. A review found insufficient evidence of an effect of psychosocial factors in private life on the risk of subsequent LBP[19]. The influence of all these factors is thus probably only modest, and their absence from our models is unlikely to have led to major bias in our estimates.

The high percentage of missing values for the variable

‘‘heavy lifting’’ was probably enhanced by the difficulty in answering the related question, particularly when the exposure was not regular (the worker was asked to evaluate his exposure to heavy lifting ‘‘during a typical day of work throughout the last 12 months’’). Despite our efforts to take this into account, it could have contributed to concealing a moderate impact of this factor, which is traditionally causally related to LBP [8,42]. However, heavy lifting may increase the risk of LBP in specific categories of workers (highly exposed) rather than influence the overall risk on a cohort scale [13]. A recent review of literature found only moderate evidence of an association for specific types of lifting and LBP and suggested further research in specific subcategories of lifting[43]. Other studies have hy- pothesized that the effect of lifting could be modulated by age or that it could increase only specific types of LBP (with radiating pain)[44,45].

Among the different groups of variables studied, biomechanical factors seem to have the strongest link with the risk of subsequently experiencing LBP in a cohort of nonspecific male workers. Frequent bending constitutes flexing (forward) and twisting (sideways) movements of the spine of large amplitude. Driving industrial vehicles, that is, tractors or forklift trucks, involves being subjected to whole- body vibrations. Both types of exposures are known to be harmful to the spine, and our results are in agreement with previous literature and with biomechanical hypotheses [8,9,11,13,42,46–48]. Our results emphasize that primary prevention of LBP in occupational settings needs first to take into account the biomechanical constraints on workers, besides psychosocial factors that have to be considered too.

Moreover, they suggest that this attention to biomechanical constraints should not be limited not only to previously affected workers but also to the unaffected: the association

(9)

between frequent bending and the risk of subsequent LBP was found to be greater in the latter.

However, other factors related to the work organization seem to be independently associated with a higher risk of LBP. The variable ‘‘working more hours than officially planned’’ had not, to our knowledge, been studied to date, and its interpretation remains somewhat uncertain. It might correspond to a marker of intensification of work constraints, leading people to feel that they continuously have to do more, better, and faster, with fewer self-perceived re- sources. It would therefore involve objective organizational constraints (such as work rate constraints) and more subjec- tive and individual responses to these constraints. This result could be compared with that of another French study, in which the workers who had reported ‘‘not having the re- sources in terms of time, equipment, and information to do their job with quality’’ presented a higher prevalence of LBP[49].

Men who reported low support from supervisors were found to have a significantly and independently higher risk of subsequent LBP. This was already observed for LBP and for other musculoskeletal disorders, particularly in mixed populations of workers such as our cohort [18,19,50]. A low skill discretion may increase moderately the risk of LBP too, even if the association was not statistically significant in our study. This latter result is quite similar to that found in the same cohort of male workers, regarding the risk of rotator cuff syndrome [51]. Different mechanisms may explain these results. The internal biomechanical loads might increase with higher levels of psychosocial constraints, even in stable conditions of external exposure [50]. Other pathophysiological hypotheses have been pro- posed, such as muscle tensions modifications, reduced blood flow, and decreased ability of tissues to recover because of hormonal processes [17,50]. Psychosocial constraints could also have an impact on the work style, which represent the behavioral, cognitive, and physiologi- cal coping strategies of the worker in response to these constraints[50,52]. Some of these strategies could increase the risk of musculoskeletal disorders, for example, taking fewer breaks, working through pain, anticipating the possible negative reactions of colleagues, and making high demands on one’s own performances at work. Finally, the work-related psychosocial environment may reduce the workers’ pain thresholds or even modify their report of symptoms, with more symptoms being perceived and/or reported in demanding jobs than in more favorable environ- ments[17,50].

No significant association between these psychosocial factors and subsequent LBP was found in the subgroup of men without LBP in the first survey. The smaller sample size may have lead to this result because of some lack of power. However, the univariate estimation of the association between a low support from supervisors and later LBP was a bit higher in the whole sample than in the

subgroup of men without LBP in the first survey. The hypothesis of a higher impact of psychosocial factors on the risk of recurrence or persistence than on the risk of incidence has already been raised in relation to other musculoskeletal disorders[50].

Despite all the methodological limits raised previously, these results suggest at least that some factors from different natures (biomechanical, organizational, and psychosocial) do influence independently the risk of later LBP.

Within each group of variables, the selection of the factors may have been sensitive to the configuration of the sample, because of a moderate degree of collinearity, or to the ac- curacy of the measurement methods. However, the risk factors highlighted in this study are consistent with preexisting scientific hypothesis and international literature, what in- creases the validity of its results.

This study was based on data from the workforce in the occupational setting, independently of the workers’ com- pensation system. Besides, most of the companies related to the study are confronted by the European and world economic market laws. The specificities of the French social support system are therefore unlikely to have had a large influence on the results, which seem to be generalizable out- side of France to a great extent.

In our cohort of workers, the relative impact of biomechanical factors was higher than those of organizational and psychosocial factors on the risk of later LBP. This pre- dominance of biomechanical constraints was even stronger among the male workers without previous LBP. These results emphasize that biomechanical factors remain worth considering in the prevention of LBP, even when psychosocial factors are taken into account. This study constitutes a significant and original contribution toward future public health policies aimed at primary prevention for large and nonspecific populations of workers.

Some of its results suggest significant differences between workers (with or without prior LBP). Future research should explore this issue in greater depth, taking histories of LBP more carefully into account and extending the assessment of histories of other musculoskeletal disorders and even other symptoms or diseases.

References

[1] Henschke N, Maher CG, Refshauge KM, et al. Prevalence of and screening for serious spinal pathology in patients presenting to primary care settings with acute low back pain. Arthritis Rheum 2009;60:3072–80.

[2] Gourmelen J, Chastang JF, Ozguler A, et al. Frequency of low back pain among men and women aged 30 to 64 years in France. Results of two national surveys. Ann Readapt Med Phys 2007;50:640–4. 33–39.

[3] Hart LG, Deyo RA, Cherkin DC. Physician office visits for low back pain. Frequency, clinical evaluation, and treatment patterns from a U.S. national survey. Spine 1995;20:11–9.

[4] Vingard E, Mortimer M, Wiktorin C, et al. Seeking care for low back pain in the general population: a two-year follow-up study: results from the MUSIC-Norrtalje Study. Spine 2002;27:2159–65.

(10)

[5] Walker BF, Muller R, Grant WD. Low back pain in Australian adults: the economic burden. Asia Pac J Public Health 2003;15:

79–87.

[6] Dagenais S, Caro J, Haldeman S. A systematic review of low back pain cost of illness studies in the United States and internationally.

Spine J 2008;8:8–20.

[7] Ivanova JI, Birnbaum HG, Schiller M, et al. Real-world practice patterns, health-care utilization, and costs in patients with low back pain: the long road to guideline-concordant care. Spine J 2011;11:

622–32.

[8] Burdorf A, Sorock G. Positive and negative evidence of risk factors for back disorders. Scand J Work Environ Health 1997;23:243–56.

[9] Lotters F, Burdorf A, Kuiper J, Miedema H. Model for the work- relatedness of low-back pain. Scand J Work Environ Health 2003;29:431–40.

[10] Hoogendoorn WE, Bongers PM, de Vet HC, et al. Flexion and rota- tion of the trunk and lifting at work are risk factors for low back pain: results of a prospective cohort study. Spine 2000;25:3087–92.

[11] Josephson M, Vingard E. Workplace factors and care seeking for low-back pain among female nursing personnel. MUSIC-Norrtalje Study Group. Scand J Work Environ Health 1998;24:465–72.

[12] Kerr MS, Frank JW, Shannon HS, et al. Biomechanical and psychosocial risk factors for low back pain at work. Am J Public Health 2001;91:1069–75.

[13] Vingard E, Alfredsson L, Hagberg M, et al. To what extent do current and past physical and psychosocial occupational factors explain care-seeking for low back pain in a working population? Results from the Musculoskeletal Intervention Center-Norrtalje Study.

Spine 2000;25:493–500.

[14] Ozguler A, Leclerc A, Landre MF, et al. Individual and occupational determinants of low back pain according to various definitions of low back pain. J Epidemiol Community Health 2000;54:215–20.

[15] Dionne CE, Dunn KM, Croft PR, et al. A consensus approach toward the standardization of back pain definitions for use in prevalence studies. Spine 2008;33:95–103.

[16] Eisen EA. Methodology for analyzing episodic events. Scand J Work Environ Health 1999;25(4 Suppl):36–42.

[17] Davis KG, Heaney CA. The relationship between psychosocial work characteristics and low back pain: underlying methodological issues. Clin Biomech (Bristol, Avon) 2000;15:389–406.

[18] Hoogendoorn WE, Bongers PM, de Vet HC, et al. Psychosocial work characteristics and psychological strain in relation to low- back pain. Scand J Work Environ Health 2001;27:258–67.

[19] Hoogendoorn WE, van Poppel MN, Bongers PM, et al. Systematic review of psychosocial factors at work and private life as risk factors for back pain. Spine 2000;25:2114–25.

[20] Henschke N, Maher CG, Refshauge KM, et al. Prognosis in patients with recent onset low back pain in Australian primary care: incep- tion cohort study. BMJ 2008;337:a171.

[21] Lambeek LC, van Mechelen W, Knol DL, et al. Randomised con- trolled trial of integrated care to reduce disability from chronic low back pain in working and private life. BMJ 2010;340:c1035.

[22] Ha C, Roquelaure Y. Reseau experimental de surveillance

epidemiologique des troubles musculo-squelettiques dans les Pays de la Loire—Protocole de la surveillance dans les entreprises (2002-2004). Saint-Maurice, France: Institut de veille sanitaire, 200784.

[23] Ha C, Roquelaure Y, Leclerc A, et al. The French Musculoskeletal Disorders Surveillance Program: Pays de la Loire network. Occup Environ Med 2009;66:471–9.

[24] Kuorinka I, Jonsson B, Kilbom A, et al. Standardised Nordic ques- tionnaires for the analysis of musculoskeletal symptoms. Appl Er- gon 1987;18:233–7.

[25] Borg G. Borg’s perceived exertion and pain scales. Human Kinetics, 1998.

[26] Niedhammer I. Psychometric properties of the French version of the Karasek Job Content Questionnaire: a study of the scales of decision

latitude, psychological demands, social support, and physical demands in the GAZEL cohort. Int Arch Occup Environ Health 2002;75:129–44.

[27]Larocque B, Brisson C, Blanchette C. [Internal consistency, factorial validity and discriminant validity of the French version of the psychological demands and decision latitude scales of the Karasek ‘‘Job Content Questionnaire’’]. [in French]. Rev Epidemiol Sante Publique 1998;46:371–81.

[28]Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York, NY: Wiley Interscience, 200092–104.

[29]Jacob T. Low back pain incident episodes: a community-based study. Spine J 2006;6:306–10.

[30]Andersen JH, Haahr JP, Frost P. Risk factors for more severe regional musculoskeletal symptoms: a two-year prospective study of a general working population. Arthritis Rheum 2007;56:1355–64.

[31]Messing K, Stock SR, Tissot F. Should studies of risk factors for musculoskeletal disorders be stratified by gender? Lessons from the 1998 Quebec Health and Social Survey. Scand J Work Environ Health 2009;35:96–112.

[32] INSEE. Les resultats du recensement, 1999. Available at: http://

www.insee.fr. Accessed June 16, 2013.

[33]Goldberg M, Luce D. [Selection effects in epidemiological cohorts:

nature, causes and consequences]. [in French]. Rev Epidemiol Sante Publique 2001;49:477–92.

[34] Hautbois L. Crise en Pays de la Loire: industrie et interim dans la tourmente. INSEE: 2010;88:1–4. Available at:http://www.insee.fr/

fr/themes/document.asp?reg_id53&ref_id516261. Accessed June 13, 2013.

[35]Stock SR, Fernandes R, Delisle A, Vezina N. Reproducibility and validity of workers’ self-reports of physical work demands. Scand J Work Environ Health 2005;31:409–37.

[36]Plouvier S, Leclerc A, Chastang JF, et al. Socioeconomic position and low-back pain—the role of biomechanical strains and psychosocial work factors in the GAZEL cohort. Scand J Work Environ Health 2009;35:429–36.

[37]Harrell FE. Regression modeling strategies. With applications to lin- ear models, logistics regression, and survival analysis. New-York, NY: Springer-Verlag, 2001.

[38]Leclerc A, Landre MF, Chastang JF, et al. Upper-limb disorders in repetitive work. Scand J Work Environ Health 2001;27:

268–78.

[39]Waxman R, Tennant A, Helliwell P. Community survey of factors associated with consultation for low back pain. BMJ 1998;317:

1564–7.

[40]Shiri R, Karppinen J, Leino-Arjas P, et al. The association between smoking and low back pain: a meta-analysis. Am J Med 2010;123:

87.e7–87.e35.

[41]Hildebrandt VH, Bongers PM, Dul J, et al. The relationship between leisure time, physical activities and musculoskeletal symptoms and disability in worker populations. Int Arch Occup Environ Health 2000;73:507–18.

[42]Hoogendoorn WE, van Poppel MN, Bongers PM, et al. Physical load during work and leisure time as risk factors for back pain.

Scand J Work Environ Health 1999;25:387–403.

[43]Wai EK, Roffey DM, Bishop P, et al. Causal assessment of occupational lifting and low back pain: results of a systematic review.

Spine J 2010;10:554–66.

[44]Miranda H, Viikari-Juntura E, Punnett L, Riihimaki H. Occupa- tional loading, health behavior and sleep disturbance as predictors of low-back pain. Scand J Work Environ Health 2008;34:

411–9.

[45]Plouvier S, Renahy E, Chastang JF, et al. Biomechanical strains and low back disorders: quantifying the effects of the number of years of exposure on various types of pain. Occup Environ Med 2008;65:

268–74.

[46]Kumar S, ed. Biomechanics in ergonomics. London, UK: Taylor &

Francis, 2008.

(11)

[47]Tiemessen IJ, Hulshof CT, Frings-Dresen MH. Low back pain in drivers exposed to whole body vibration: analysis of a dose- response pattern. Occup Environ Med 2008;65:667–75.

[48]Kuijer PP, Frings-Dresen MH, Gouttebarge V, et al. Low back pain: we cannot afford ignoring work. Spine J 2011;11:164. author reply 5–6.

[49]Derriennic F, Touranchet A, Volkoff S. Age, travail, sante. Etude sur les salaries^ages de 37a 52 ans. Enqu^ete ESTEV 1990. In: INSERM Le, ed 1996.

[50]Bongers PM, Ijmker S, van den Heuvel S, Blatter BM. Epidemiol- ogy of work related neck and upper limb problems: psychosocial

and personal risk factors (part I) and effective interventions from a bio behavioural perspective (part II). J Occup Rehabil 2006;16:

279–302.

[51] Roquelaure Y, Bodin J, Ha C, et al. Personal, biomechanical, and psychosocial risk factors for rotator cuff syndrome in a - working population. Scand J Work Environ Health 2011;37:

502–11.

[52] Feuerstein M, Nicholas RA, Huang GD, et al. Workstyle: develop- ment of a measure of response to work in those with upper extremity pain. J Occup Rehabil 2005;15:87–104.