The Spine Journal 10 (2010) 252–261

Review Article

Causal assessment of occupational sitting and low back

pain: results of a systematic review
Darren M. Roffey, PhDa, Eugene K. Wai, MD, MSc, CIP, FRCSCa,b,*,
Paul Bishop, DC, MD, PhDc,d,e,f,
Brian K. Kwon, MD, PhD, FRCSCd,e,f, Simon Dagenais, DC, PhDa,b,g
a
Clinical Epidemiology Program, Ottawa Hospital Research Institute, Ottawa, Ontario, Canada
b
Division of Orthopaedic Surgery, Department of Surgery, University of Ottawa, The Ottawa Hospital, Ottawa, Ontario, Canada
c
Department of Orthopaedics, University of British Columbia, Vancouver, British Columbia, Canada
d
UBC Combined Neurosurgical and Orthopaedic Spine Program, Vancouver, British Columbia, Canada
e
Vancouver Hospital Spine Program and Acute Spinal Cord Injury Unit, Vancouver, British Columbia, Canada
f
International Collaboration on Repair Discoveries, University of British Columbia, Vancouver, British Columbia, Canada
g
Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, Ontario, Canada
Received 29 July 2009; accepted 2 December 2009

Abstract BACKGROUND CONTEXT: Low back pain (LBP) is a common and disabling musculoskeletal
disorder that often occurs in a working-age population. Determining the precise causation of LBP
remains difficult. Any attempt to implicate a specific occupational activity in the genesis of LBP
requires a methodologically rigorous approach.
PURPOSE: To conduct a systematic review of the scientific literature focused on evaluating the
causal relationship between occupational sitting and LBP.
STUDY DESIGN: Systematic review of the literature using Medline, EMBASE, CINAHL,
Cochrane Library, Occupational Safety and Health database, grey literature, hand-searching occu-
pational health journals, reference lists of included studies, and content experts. Evaluation of study
quality using a modified version of the Newcastle-Ottawa Scale. Summary levels of evidence sup-
porting Bradford-Hill criteria for different categories of sitting and types of LBP.
SAMPLES: Studies reporting an association between occupational sitting and LBP.
OUTCOME MEASURES: Numerical association between different levels of exposure to occupa-
tional sitting and the presence or severity of LBP.
METHODS: A systematic review was performed to identify, evaluate, and summarize the litera-
ture related to establishing a causal relationship, according to Bradford-Hill criteria, between occu-
pational sitting and LBP.
RESULTS: This search yielded 2,766 citations. Twenty-four studies met the inclusion/exclu-
sion criteria and five were high-quality studies, including two case-controls and three prospec-
tive cohorts. Strong, consistent evidence was found for no association between occupational
sitting and LBP. A moderate level of evidence was found for the absence of any dose-response
trend. Risk estimates evaluating temporality were not statistically significant. Biological plau-
sibility was not discussed in these studies. No evidence was available to assess the experiment
criterion.
CONCLUSIONS: This review failed to uncover high-quality studies to support any of the Brad-
ford-Hill criteria to establish causality between occupational sitting and LBP. Strong and consistent
evidence did not support criteria for association, temporality, and dose response. Based on these

FDA device/drug status: not applicable.
Author disclosures: BKK (consulting, Medtronic); SD (consulting and
member scientific advisory board, Palladian Health).
This study was funded by a peer-review grant provided to Drs Wai,
Bishop, Kwon, and Dagenais by the Workers Compensation Board of
British Columbia.
* Corresponding author. Division of Orthopaedic Surgery, Department
of Surgery, University of Ottawa, The Ottawa Hospital, 1053 Carling Ave.,
Ottawa, Ontario K1Y 4E9, Canada. Tel.: (613) 798-5555 ext 19138; fax:
(613) 761-4944.
E-mail address: ewai@ottawahospital.on.ca (E.K. Wai)

1529-9430/10/$ – see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.spinee.2009.12.005
 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261
results, it is unlikely that occupational sitting is independently causative of LBP in the populations
of workers studied. Ó 2010 Elsevier Inc. All rights reserved.
Keywords: Occupational health; Low back pain; Sitting; Causality; Etiology; Systematic review
Introduction
Low back pain (LBP) is a musculoskeletal disorder
thought to result from a combination of chronic overuse
and acute injury to the muscles, ligaments, tendons, inter-
vertebral discs, nerves, or vertebrae of the lumbar spine
[1]. LBP is one of the most common musculoskeletal disor-
ders, with a reported lifetime prevalence of up to 90% [2,3].
In the working-age population, the medical costs associated
with LBP exceed those of coronary artery disease, respira-
tory infections, and diabetes [4]. More importantly, the in-
direct costs of LBP because of lost productivity and early
retirement likely surpass the substantial medical costs [5].
Given its occurrence in working-age adults, LBP is as
a major occupational health concern. To reduce the inci-
dence and impact of work-related LBP, it is essential that
we improve our understanding of its etiology as it relates
to modifiable risk factors, such as specific occupational ac-
tivities. Sitting is a specific physical activity (or inactivity)
common to many of today’s service-oriented occupations.
A recent study reported that adults in sedentary occupa-
tions spent an average of 597 minutes (10 hours) in a sitting
position during a 24-hour period [6]. Previous studies have
suggested that extended sitting decreases lumbar lordosis
and increases muscle activity, intradiscal pressure, and
pressure on the ischium [7,8]. Prolonged flexion during sit-
ting has also been shown to cause a redistribution of the nu-
cleus within the annulus [9]. Together, these factors may
culminate in disc degeneration, herniation, or rupture,
potentially leading to LBP [7,8].
Biomechanical studies [10–12] and a small number of
cross-sectional studies [13–16] as well as literature reviews
[17,18] have previously reported a weak-to-moderate associ-
ation between occupational sitting and LBP. However, other
studies have reported that sitting is not associated with the de-
velopment of occupational LBP [19–21]. To reconcile this
conflicting information, a more thorough comprehension of
the causal association between occupational sitting and
LBP is required. This knowledge is needed to help establish
parameters for exposure to sitting in occupational health
guidelines and also to provide guidance to stakeholders
involved in the adjudication of work-related LBP claims.
Establishing causal links between specific risk factors and
LBP from single studies has proven complex and unreliable
in the past because of the limitations imposed by specific re-
search questions, study designs, study populations, study
methodological quality, and specific types of statistical anal-
yses [22]. In these situations, a systematic review can help to
establish causation between isolated risk factors and LBP by
summarizing all available evidence in light of the many cri-
teria that have been proposed to determine causation [22,23].
253
During this process, the methodological quality of the stud-
ies can also be assessed to determine the importance that
should be accorded to their results in light of their propensity
for bias and confounding [22,24,25].
The aim of this study was to identify, evaluate, and sum-
marize the best-available scientific evidence related to
occupational sitting and LBP using the Bradford-Hill [26]
criteria for causation.
Methods
Literature search
An electronic search of Medline (1966 to November
2007; updated in August 2008), EMBASE (1980 to No-
vember 2007), and CINAHL (1982 to November 2007)
was executed using a broad strategy with three main com-
ponents: setting (ie, work related), etiology (ie, sitting), and
outcome (ie, LBP). Both indexed search terms and free text
were used. (Note: search strategy and results are available
on request). A hand search was also performed of the three
occupational health journals with the highest impact factor
ranking (Occupational and Environmental Medicine, Scan-
dinavian Journal of Work Environment and Health, and
Journal of Occupational and Environmental Medicine)
from January 1997 to April 2008. The search included grey
literature and reviewed the following sources of informa-
tion: conference proceedings from the International Society
for the Study of the Lumbar Spine and the North American
Spine Society, Web sites of members of the International
Network of Agencies for Health Technologies Assessment,
the Occupational Safety and Health Read only Memory
database, the National Institute for Occupational Safety
and Health database, and a general Internet search for re-
lated materials. Additional references were sought from
previous systematic reviews on similar topics and by con-
tacting content experts. References of included studies were
screened for relevance.
Eligibility criteria
The criteria required for inclusion were published in En-
glish or French and related to occupational exposure, LBP,
etiology or causation, and occupational sitting.
The criteria required for exclusion were no specific pop-
ulation, exposure, and outcome (eg, too broad); nonscien-
tific studies (eg, commentaries, letters to the editor);
literature reviews; related only to treatment of LBP (eg,
does not address a specific risk factor); health services
research only (eg, costs of injuries); basic sciences, biome-
chanics studies, cadaver studies; less than 30 exposed
254 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261
subjects; whole body vibration, psychosocial, or environ-
mental risk factors only; neck pain, thoracic pain, whole
spine pain, or other nonspecific back pain.
Screening process
Search results were imported into Systematic Review
Software (SRS), version 3.0 (TrialStat, Ottawa, Ontario,
Canada) and screened by two reviewers after a calibration
and training process. Disagreements among reviewers were
resolved by discussion until consensus was reached on in-
clusion or exclusion. Two levels of screening were used.
The first evaluated all available information returned by
the electronic search (eg, abstract, title, keywords). The
second evaluated full-text reports for studies deemed poten-
tially eligible after the first level of screening or for which
insufficient information was available to determine eligibil-
ity (eg, no abstract).
Methodological quality assessment
The methodological quality of studies was assessed in-
dependently by two reviewers using a modified version of
the Newcastle-Ottawa Scale case-control, cross-sectional,
and cohort studies [24]. Disagreements among reviewers
were resolved by discussion until consensus was reached.
To be considered of high methodological quality, studies
needed a score of five or higher on the Newcastle-Ottawa
Scale (maximum score59) and appropriate statistical anal-
ysis. Multivariate analysis or other acceptable methods of
adjusting for confounding were required to minimize the
possibility of reporting biased results that failed to account
for the multiple risk factors for LBP.
Data abstraction
Data pertaining to the following elements were obtained
from all studies deemed by one reviewer and verified inde-
pendently by the other reviewer. Disagreements were re-
solved by discussion until consensus was reached.
1. Study design (cross-sectional, case-control, prospec-
tive cohort)
2. Study population and setting (country, employer,
industry, occupation)
3. Categories of occupational sitting (definition,
measurement, level of exposure)
4. Type of LBP outcome (definition, type, severity,
assessment period, health care use, sick leave)
5. Measurement and controlling for known LBP con-
founders (psychosocial work factors, other physical
factors)
6. Type of analysis (statistical methods, univariate/mul-
tivariate, adjusting for confounders)
7. Measures of association (odds ratio, relative risk)
with confidence intervals, or raw data necessary to
calculate these measures of association
8. Study funding source and reported author conflicts of
interest
Subgroup analyses
Separate analyses were conducted for each combination
of category of occupational sitting and type of LBP out-
come. The two categories of sitting considered were not
specified or not reported and harmful sitting (as reported
in the primary studies). The two types of LBP outcomes
were considered: LBP or injury (any) and LBP or injury
(severe). LBP outcomes were classified as ‘‘any’’ if the
study used a minimal level of severity or duration or failed
to define the level of severity or duration. LBP outcomes
were defined as ‘‘severe’’ based on LBP lasting more than
2 weeks [27], a Functional Rating Index score greater than
30% [28], or more severe LBP compared with no or minor
LBP at baseline [29].
Analysis
These Bradford-Hill [26] criteria for causation were
evaluated for each subgroup analysis: association (includ-
ing strength of significant associations), dose response,
experiment, temporality, and biological plausibility.
The parameters used to determine whether each criterion
was satisfactorily met are summarized in Table 1. When stud-
ies reported multiple risk estimates (eg, separate results for dif-
ferent populations within the study), each risk estimate was
analyzed independently to determine if it satisfied each of
the Bradford-Hill criteria. If most of the risk estimates in
a study satisfied a specific Bradford-Hill [26] criterion, the re-
sults of the study were considered supportive. Other Bradford-
Hill [26] criteria for causality were not considered in the
analysis because they did not apply or could not be assessed.
Level of evidence
The results from each study were then summarized to
determine the overall level of evidence supporting each cri-
terion for causality for each subgroup (ie, category of sit-
ting and type of LBP outcome). The levels of evidence
were developed based on previous methodologies to com-
bine results from different study designs (eg, Agency for
Health Care Policy and Research [34], Oxford Center for
Evidence Based Medicine [25]). Summary results are pre-
sented in Table 2.
Results
The search strategy yielded a total of 2,766 citations. A
total of 275 were deemed potentially relevant at the first
level of screening. On further review, 24 satisfied the eligi-
bility criteria. The 24 studies enrolled a total of 75,103 par-
ticipants (mean 3,129; standard deviation 5,173). The mean
prevalence of LBP reported across all studies was 42.2%
 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261 255

Table 1
Statistical assessment for specific Bradford-Hill criteria for causation
Criteria Statistical assessment Qualification of strength of relationship* Reference
Association and experiment Odds ratio Protective: !1.0 [30]
Weak: 1.0–2.4
Moderate: 2.5–3.9
Strong: O4.0
Relative risk Protective: !1.0 [31]
Hazard ratio Weak: 1.0–1.9
Prevalence ratio Moderate: 2.0–2.9
Incidence rate ratio Strong: O3.0
t test Clinically Significant: O10% change in effect [32]
Consistency Sackett’s strength of evidence Strong: O75% of studies (at least 2 high quality) [25]
Dose response Pearson’s correlation Protective: !0.0 [33]
Weak: 0.1–0.29
Moderate: 0.3–0.49
Strong: O0.5
Logistical regression Protective: !0.0 [33]
Weak: 0.1–0.29
Moderate: 0.3–0.49
Strong: O0.5
Confident intervals on estimates Significant: nonoverlapping
Trend: overlapping confidence interval
* Strength at the risk estimate level refers to how strong a relationship is for the observed unique risk estimate or comparison. In contrast, strength at the
evidence level (Table 2) refers to how strong the evidence supporting a conclusion is.

(standard deviation 19.4%). These studies were conducted
in 12 countries, most commonly from the United States
(n55), Sweden (n54), and Denmark (n53). A total of 36
occupations were represented by these studies, including
multiple occupations (n510), nurses (n54), laborers
(n54), administrators (n54), and retail salesperson
(n53). There were 5 prospective cohort studies, 16 cross-
sectional studies, and 3 case-control studies.
Of the 24 studies identified, 19 (79%) were considered
of low methodological quality and five (21%) were of high
methodological quality. The mean Newcastle-Ottawa Scale
score was 3.5, with a standard deviation of 1.5. Figure 1
summarizes the retrieval, screening, abstraction, and analy-
sis process. The characteristics of the included high-quality
(Table 3) and low-quality (Table 4) studies are summarized
below.
the occupational sitting and LBP [27,29,35–37]. Two
were case-control studies [27,35] and three were prospec-
tive cohort studies [29,36,37]. One study was in nurses
[35], one study was in construction workers [37], and
three studies were in multiple occupations [27,29,36]. A
total of 1,591 participants were analyzed in these five
studies.
None of the multivariate risk estimates for occupational
sitting and LBP in these five high-quality studies were sta-
tistically significant. As such, results were consistent for no
association across all the studies. Only one of the five high-
quality studies (20%) assessed multiple doses of occupa-
tional sitting; results did not support any dose-response
trend [36]. Most high-quality studies (80%) were able to
assess temporality [29,36,37] but none demonstrated any
significant associations. There was no available evidence

Overall association of occupational sitting with LBP Table 2

Requirements for levels of evidence relating to the Bradford-Hill criteria
Together, these 24 studies reported a total of 108 sepa- Evidence* Requirements
rate estimates of the association between specific categories
Strong 2 or more high-quality studies with consistent
of occupational sitting and specific types of LBP outcomes. multivariate results
Of these 108 estimates, 17 (16%) were reported to be sta- Moderate 1 high-quality study or 2 low-quality studies with
tistically significant (eg, p!.05). Of these 17 statistically consistent multivariate results
significant estimates, 3 (18%) were classified as ‘‘weak’’ Limited 1 low-quality study or unadjusted results (Note:
and 14 (82%) were classified as ‘‘protective.’’ There was these studies were not considered in the causation
assessment)
a marked difference in the proportion of estimates consid- Conflicting Inconsistent studies of same quality (consistent high
ered statistically significant for high-quality (0%) versus quality O inconsistent low quality)
low-quality (100%) studies. There was strong evidence
* Strength at an evidence level refers to how strong the evidence
for consistency of findings within the subgroups. supporting a conclusion is. In contrast, strength at the risk estimate level
A total of five high-quality (score 5–7) studies using (Table 1) refers to how strong a relationship is for the observed unique risk
multivariate analysis reported on an association between estimate or comparison.
256 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261

Fig. 1. Study flow diagram.

to evaluate the experiment criterion. As there were no Subgroup analyses

significant associations reported, there was no discussion
of the biological plausibility of a causal link between Based on the above classifications, there were seven
multivariate subgroup analyses in five high-quality studies
occupational sitting and LBP.
 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261 257

Table 3
Characteristics of high-quality studies
Occupation(s)
Author, year (reference) Country Study design (FU) studied (industry) Mean age N NOS score
Yip, 2004 [35] China Case-control Nurses (district hospitals [6 31.1 144 6
hospitals])
Harkness et al., 2003 [36] United Kingdom Pros. cohort (2 y) Retail salespersons; general 23 625 6
laborers; childcare
providers; administrators;
firefighters; police officers;
military personnel;
shipbuilders; nurses;
podiatrists; forestry
workers; postal workers
(multiple [12 occupation
groups])
Latza et al., 2000 [37] Germany Pros. cohort (3 y) Construction workers 32.5 230 7
(construction)
Yip et al., 2004 [27] China Case-control Multiple (general population NR 418 5
and patients from family
practice unit)
Andersen et al., 2007 [29] Denmark Pros. Cohort (2 y) Multiple (multiple [39 27.8 174 6
workplaces])
FU, follow-up; N, number analyzed; NOS, Newcastle-Ottawa Scale; NR, not reported; Pros. cohort, prospective cohort.

presented in Table 5. None of these categories of sitting had
any high-quality evidence to satisfy any of the criteria for
causation.
Discussion
Previous studies [13–15] have suggested that working in
a sitting position for extended periods may result in LBP.
However, results from the current systematic review indi-
cate that sitting did not meet any of the standard criteria re-
quired to establish causation for LBP. When applying the
Bradford-Hill criteria for causation [26], this systematic re-
view revealed strong and consistent evidence suggesting no
association and no temporality between occupational sitting
and LBP, and there was a moderate level of evidence sug-
gesting the absence of a dose-response trend. In the
reviewed literature, the criteria of biological plausibility
and experiment were not discussed. Among the subgroups,
none had any evidence to satisfy the criteria for causation.
Sitting (not reported) and LBP (any) had a strong level of
evidence suggesting no association and no temporality,
and a moderate level of evidence for no dose-response
trend.
Perhaps more noteworthy was our finding that 14 of 17
(82%) statistically significant estimates of the association
between occupational sitting and LBP suggested a poten-
tially ‘‘protective’’ effect. Admittedly, these estimates were
reported only in low-quality studies, and none of the high-
quality studies produced any similarly statistically signifi-
cant estimates to validate this finding. However, these
results suggest the possibility that rather than occupational
sitting causing LBP, it may in fact help prevent it. Perhaps
these results are more easily interpreted when compared
with the possibility of a causal relationship existing
between LBP and other potentially harmful occupational
activities, such as twisting, bending, kneeling, lifting, carry-
ing, standing, and manual handling.
The questions raised by these results must be considered
in the broader context of numerous other studies in recent
decades that have attempted to establish a definitive causal
link between isolated occupational risk factors and LBP.
Common LBP is likely to have multiple etiologies, which
would dilute the potential impact of any one isolated risk
factor. In the case of sitting, there are a few examples of
studies that have examined sitting in combination with
awkward postures or whole body vibration [15–18]. In such
instances, it becomes difficult to isolate a causal relation-
ship between sitting and LBP or whether sitting is acting
as a moderating effect for these risk factors.
In the current literature review, none of the high-quality
studies discussed the biological plausibility of this potential as-
sociation. Although biomechanical and physiological studies
have evaluated possible mechanisms by which occupational
sitting could theoretically cause injury to lumbar tissues
[7,8], it is clear that the biological plausibility of this theory re-
quires further evidence from more high-quality studies.
There are numerous explanations available for not finding
any positive moderate-to-strong levels of evidence for causa-
tion. It is admittedly challenging to establish causality accord-
ing to all of the Bradford-Hill [26] criteria. For example, no
studies met the experiment criterion. Although challenging,
it may be possible to do this by measuring the level of exposure
to a certain risk factor (eg, sitting more than 8 hours) and the
outcome of interest (eg, LBP) both before and after an inter-
vention aimed at reducing the exposure (eg, worker education,
258 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261

Table 4
Characteristics of low-quality studies
Author, year [reference] Country Study design (FU) Occupations studied (industry) Mean age N NOS score
Bos et al., 2007 [38] Netherlands Cross-sectional Nurses; medical radiation 38.0 3,169 3
technologists (university
hospitals [8 hospitals])
Ghaffari et al., 2006 [39] Iran Cross-sectional Multiple (car manufacturing) NR 14,384 3
Sanya and Ogwumike, 2005 [40] Nigeria Cross-sectional Plastic products laborers; general NR 604 1
office clerks; machine operators
(manufacturing)
Schneider et al., 2005 [41] Germany Cross-sectional Multiple (general population) 40.0 3,488 3
van Vuuren et al., 2005 [28] South Africa Cross-sectional Metal fabrication laborers (steel 31.8 366 3
plant)
Masset, 1994 [42] Belgium Cross-sectional Steel workers (steel industry [2 NR 618 3
industries])
Svensson, 1989 [43] Sweden Cross-sectional Multiple (general population) NR 1,410 4
Frymoyer, 1980 [44] USA Cross-sectional Multiple, (general population 31.5 3,920 2
patients from family practice
unit)
Magora, 1972 [12] USA Case-control Retails salespersons; bus drivers; NR 3,316 3
police officers; nurses; farmers;
light industrial workers; heavy
industrial workers (multiple [8
occupations])
Xu, 1997 [45] USA Cross-sectional Multiple (general population) NR 8,664 4
Skov, 1996 [46] Denmark Cross-sectional Retails salespersons (Association 41.4 1,306 3
of Danish salesperson)
Holmstrom, 1992 [47] Sweden Cross-sectional Bricklayers; carpenters; concrete 39.5 1,773 3
workers; plumbers; roofers;
scaffolders; insulators; machine
operators; crane operators
(construction [trade union])
Walsh, 1991 [48] United Kingdom Cross-sectional Multiple, population (patients from NR 2,667 3
family medicine units [6 towns])
Burdorf, 1990 [49] Netherlands Cross-sectional Crane operators; maintenance 42.0 30 2
workers (steel company)
Linton, 1990 [50] Sweden Cross-sectional Multiple (multiple population) 42.0 22,180 3
Svensson, 1983 [51] Sweden Cross-sectional Multiple (general population) 44.0 714 3
Macfarlane, 1997 [52] United Kingdom Pros. cohort (1 y) Multiple, population (patients from 38.0 847 3
family medicine unit [2 units])
(general population patients
from family medicine unit
[2 units])
Wickstrom and Pentti, 1998 [53] Finland Pros. cohort (2 y) Administrators; plumbers; 39.0 306 2
sheetmetal workers; welders
(shipyard and ventilation
company)
Spyropoulos et al., 2007 [54] Greece Cross-sectional Administrators (government 44.5 648 4
offices [4 offices])
FU, follow-up; N, number analyzed; NOS, Newcastle-Ottawa Scale; NR, not reported; Pros. cohort, prospective cohort.

or manipulation of posture). The Bradford-Hill [26] criteria for
causality are used in epidemiologic research to minimize the
possibility that important public health decisions are made
on the basis of incomplete or flawed evidence. Given the socio-
economic burden of work-related LBP, it would appear that
better information is in fact necessary to inform decision mak-
ing in an evidence-based manner.
Most of the high-quality studies reported risk estimates as
‘‘nonsignificant’’ without reporting actual values, making
statistical pooling of results impossible. Therefore, this sys-
tematic review did not rely on statistical pooling in the assess-
ment of causation across studies. Although statistical pooling
across studies may have increased the power between studies
to detect a difference, it is likely that the studies reviewed had
sufficient power to detect an effect. Using an assumption of
LBP prevalence as 35%, an alpha of 0.05, power of 0.80,
and equal distribution of risk factor, 30 subjects with the ex-
posed risk factor would have been sufficient to demonstrate
a moderate relative risk; all 24 low- and high-quality studies
included in this review had more than 30 subjects.
There were two main sources of limitations with this
present study. The first were weaknesses encountered in
the primary studies. The reporting quality of primary stud-
ies was often poor, making consolidation of incomplete
 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261 259

results difficult. Commonly noted reporting weaknesses in-

* Strength at an evidence level refers to how strong the evidence supporting a conclusion is. In contrast, strength at the risk estimate level (Table 1) refers to how strong a relationship is for the observed
Biological plausibility
cluded failure to adopt common operational definitions of
LBP, failure to report basic data about the study population
(eg, age, sex), failure to describe the type of statistical
methods used (eg, univariate vs. multivariate), failure to ad-
just for known confounders, and a failure to disclose which
NA
NA
NA
NA
NA

NA
NA
NA
variables were adjusted for in multivariate analyses.
The second were the limitations inherent in the system-
Temporality

No (strong)

No (mod.)
atic review process itself, which requires a uniform applica-
tion of criteria based on reported information. It is possible
NA

NA
No
No
No

No
that meaningful studies that did not meet our criteria were
overlooked. However, the screening process was transpar-
Experiment

ent and confirmed independently to ensure that only the

most relevant studies were included. Heterogeneity was
NA
NA
NA
NA
NA

NA
NA
NA
noted in some of the categories of sitting among the in-
cluded studies; forcing them into specific categories may
Dose response

have resulted in misclassification. However, this classifica-

No (mod.)

tion process was undertaken before the analysis with two

independent reviewers to minimize bias.
NA

NA
NA

NA
NA
NA
No

Most of the studies uncovered were cross-sectional;

unique risk estimate or comparison. LBP, low back pain; NR, not reported; Mod, moderate; NA, not available; NS, not significant.

a design that is insufficient to ascertain causation because

Strength of association*

both the risk factor and the outcome are measured simulta-
neously. Case-control studies are also subject to other limita-
tions (eg, recall bias). If possible, future studies examining
occupational causes of LBP should use a prospective cohort
design. Many studies reported only a dichotomous exposure
variable (eg, sitting yes/no), making it impossible to deter-
NS
NS
NS
NS

NS
NS

mine whether a dose-response relationship was present. This

could be corrected by measuring the exposure numerically
Association

No (strong)

No (strong)

(eg, percentage of work time spent sitting), which could then

be categorized into dose groups.
No
No
No
No

No
No

Previous cross-sectional studies and systematic reviews

examining sitting and LBP have methodological limitations
Estimates per study

which reduce their potential impact. Such limitations

include failing to use explicit criteria to consolidate high-
quality adjusted results, the inclusion of co-exposure fac-
tors, such as whole body vibration and/or awkward postures
[15–18], the examination of a sedentary lifestyle rather than
1
2
1
1

1
1

sitting per se [19], and a literature search date range (1985–

Andersen et al., 2007 [29]
Harkness et al., 2003 [36]

1997) [21] that does not include studies conducted in the

Author, Year [Reference]

Latza et al., 2000 [37]

modern-day sedentary environment.

Yip et al., 2004 [27]

Yip et al., 2004 [27]

Yip, 2004 [35]

Conclusion
Results in high-quality studies for sitting

A systematic review was unable to uncover any evidence

supporting the Bradford-Hill criteria required to establish
Level of evidence across studies

Level of evidence across studies

a causal relationship between occupational sitting and LBP.

Categories of posture and LBP

Based on the results of this study, occupational sitting does

Sitting (NR)-LBP or injury

Sitting (NR)-LBP or injury

not appear to be independently causative of LBP in workers.

The strength of evidence suggesting no association was con-
sistent and rated as strong, and only one study demonstrated
(any or NR)

a trend toward a nonstatistically significant dose response.

(severe)

This indicates that if a causal relationship were to exist, it

Table 5

would be a very weak one. Results suggest that working in

a sitting position for a prolonged duration may, in fact, have
260 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261
a protective effect; this scenario is only theoretical, however,
because of the lack of strong, high-quality evidence. Future
studies examining occupational sitting and LBP should at-
tempt to examine biological plausibility, and to fully and
clearly report their results to avoid the common methodolog-
ical weaknesses uncovered in this review.
References
[1] Chou R, Qaseem A, Snow V, et al. Diagnosis and treatment of low
back pain: a joint clinical practice guideline from the American Col-
lege of Physicians and the American Pain Society. Ann Intern Med
2007;147:478–91.
[2] Frymoyer JW, Pope MH, Clements JH, et al. Risk factors in low-back
pain. An epidemiological survey. J Bone Joint Surg Am 1983;65:213–8.
[3] Papageorgiou AC, Croft PR, Ferry S, et al. Estimating the prevalence
of low back pain in the general population. Evidence from the South
Manchester Back Pain Survey. Spine (Phila Pa 1976) 1995;20:
1889–94.
[4] Hanney WJ, Kolber MJ, Beekhuizen KS. Implications for physical
activity in the population with low back pain. Am J Lifestyle Med
2009;3:63–70.
[5] 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.
[6] McCrady SK, Levine JA. Sedentariness at work: how much do we
really sit? Obesity (Silver Spring) 2009;17:2103–5.
[7] Claus A, Hides J, Moseley GL, Hodges P. Sitting versus standing:
does the intradiscal pressure cause disc degeneration or low back
pain? J Electromyogr Kinesiol 2008;18:550–8.
[8] Makhsous M, Lin F, Bankard J, et al. Biomechanical effects of sitting
with adjustable ischial and lumbar support on occupational low back
pain: evaluation of sitting load and back muscle activity. BMC Mus-
culoskelet Disord 2009;10:17.
[9] McKenzie RA. Prophylaxis in recurrent low back pain. N Z Med J
1979;89:22–3.
[10] Nachemson A. The load on lumbar disks in different positions of the
body. Clin Orthop Relat Res 1966;45:107–22.
[11] Nachemson A, Elfstrom G. Intravital dynamic pressure measure-
ments in lumbar discs. A study of common movements, maneuvers
and exercises. Scand J Rehabil Med Suppl 1970;1:1–40.
[12] Magora A. Investigation of the relation between low back pain and
occupation. 3. Physical requirements: sitting, standing and weight
lifting. IMS Ind Med Surg 1972;41:5–9.
[13] Mork PJ, Westgaard RH. Back posture and low back muscle activity
in female computer workers: a field study. Clin Biomech (Bristol,
Avon) 2009;24:169–75.
[14] Chiou WK, Wong MK, Lee YH. Epidemiology of low back pain in
Chinese nurses. Int J Nurs Stud 1994;31:361–8.
[15] Chen JC, Chang WR, Chang W, Christiani D. Occupational factors
associated with low back pain in urban taxi drivers. Occup Med
2005;55:535–40.
[16] Miyamoto M, Shirai Y, Nakayama Y, et al. An epidemiologic study
of occupational low back pain in truck drivers. J Nippon Med Sch
2000;67:186–90.
[17] Lyons J. Factors contributing to low back pain among professional
drivers: a review of current literature and possible ergonomic con-
trols. Work 2002;19:95–102.
[18] Pelham TW, White H, Holt LE, Lee SW. The etiology of low back
pain in military helicopter aviators: prevention and treatment. Work
2005;24:101–10.
[19] Chen SM, Liu MF, Cook J, et al. Sedentary lifestyle as a risk factor
for low back pain: a systematic review. Int Arch Occup Environ
Health 2009;82:797–806.
[20] Lis AM, Black KM, Korn H, Nordin M. Association between sitting
and occupational LBP. Eur Spine J 2007;16:283–98.
[21] Hartvigsen J, Leboeuf-Yde C, Lings S, Corder EH. Is sitting-while-
at-work associated with low back pain? A systematic, critical litera-
ture review. Scand J Public Health 2000;28:230–9.
[22] Bombardier C, Kerr MS, Shannon HS, Frank JW. A guide to inter-
preting epidemiologic studies on the etiology of back pain. Spine
(Phila Pa 1976) 1994;19:2047S–56S.
[23] Weed DL. Interpreting epidemiological evidence: how meta-analysis
and causal inference methods are related. Int J Epidemiol 2000;29:
387–90.
[24] Wells G, Shea B, O’Connell D, et al. The Newcastle-Ottawa Scale
(NOS) for assessing the quality of nonrandomised studies in meta-
analyses. Available at: http://www.ohri.ca/programs/clinical_epide
miology/oxford.htm 2006. Accessed August 7, 2008.
[25] Sackett DL, Straus SE, Richardson WS, et al. Evidence-based med-
icine: how to practice and teach EBM. Edinburgh, UK: Churchill Liv-
ingstone, 2000.
[26] Hill AB. The environment and disease: association or causation? Proc
R Soc Med 1965;58:295–300.
[27] Yip YB, Ho SC, Chan SG. Identifying risk factors for low back pain
(LBP) in Chinese middle-aged women: a case-control study. Health
Care Women Int 2004;25:358–69.
[28] van Vuuren BJ, Becker PJ, van Heerden HJ, et al. Lower back prob-
lems and occupational risk factors in a South African steel industry.
Am J Ind Med 2005;47:451–7.
[29] 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.
[30] Rosenthal J. Qualitative descriptors of strength of association and ef-
fect size. J Soc Serv Res 1996;21:37–59.
[31] Milloy S. 2-4-6-8 What can we associate! Science without sense: the
risky business of public health research. Washington, DC: Cato Institute,
1995.
[32] Feinstein A. Multivariate analysis: an introduction. New Haven, CT:
Yale University Press, 1996.
[33] Cohen J, Cohen P. Applied multiple regression/correlation analysis
for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum, 2002.
[34] Bigos SJ, Bowyer O, Braen G. Acute low back problems in adults.
Clinical Practice Guidelines No. 14. Rockville: Agency for Health
Care Policy and Research Publications; 1994.
[35] Yip YB. New low back pain in nurses: work activities, work stress
and sedentary lifestyle. J Adv Nurs 2004;46:430–40.
[36] Harkness EF, Macfarlane GJ, Nahit ES, et al. Risk factors for new-
onset low back pain amongst cohorts of newly employed workers.
Rheumatology (Oxford) 2003;42:959–68.
[37] Latza U, Karmaus W, Sturmer T, et al. Cohort study of occupational
risk factors of low back pain in construction workers. Occup Environ
Med 2000;57:28–34.
[38] Bos E, Krol B, van der Star L, Groothoff J. Risk factors and muscu-
loskeletal complaints in non-specialized nurses, IC nurses, operation
room nurses, and X-ray technologists. Int Arch Occup Environ
Health 2007;80:198–206.
[39] Ghaffari M, Alipour A, Jensen I, et al. Low back pain among Iranian
industrial workers. Occup Med (Lond) 2006;56:455–60.
[40] Sanya AO, Ogwumike OO. Low back pain prevalence amongst in-
dustrial workers in the private sector in Oyo State, Nigeria. Afr J
Med Med Sci 2005;34:245–9.
[41] Schneider S, Schmitt H, Zoller S, Schiltenwolf M. Workplace stress,
lifestyle and social factors as correlates of back pain: a representative
study of the German working population. Int Arch Occup Environ
Health 2005;78:253–69.
[42] Masset D, Malchaire J. Low back pain. Epidemiologic aspects and
work-related factors in the steel industry. Spine (Phila Pa 1976)
1994;19:143–6.
[43] Svensson HO, Andersson GB. The relationship of low-back pain, work
history, work environment, and stress. A retrospective cross-sectional
 D.M. Roffey et al. / The Spine Journal 10 (2010) 252–261
study of 38- to 64-year-old women. Spine (Phila Pa 1976) 1989;14:
517–22.
[44] Frymoyer JW, Pope MH, Costanza MC, et al. Epidemiologic studies
of low-back pain. Spine (Phila Pa 1976) 1980;5:419–23.
[45] Xu Y, Bach E, Orhede E. Work environment and low back pain: the in-
fluence of occupational activities. Occup Environ Med 1997;55:741–5.
[46] Skov T, Borg V, Orhede E. Psychosocial and physical risk factors for
musculoskeletal disorders of the neck, shoulders, and lower back in
sales people. Occup Environ Med 1996;53:351–6.
[47] Holmstrom EB. Low back and neck/shoulder pain in construction
workers: occupational workload and psychosocial risk factors. Part
1: Relationship to low back pain. Spine (Phila Pa 1976) 1992;17:
663–71.
[48] Walsh KC. Interaction of height and mechanical loading of the spine
in the development of low back pain. Scand J Work Environ Health
1991;17:420–4.
261
[49] Burdorf A, Zondervan H. An epidemiological study of low-back pain
in crane operators. Ergonomics 1990;33:981–7.
[50] Linton SJ. Risk factors for neck and back pain in a working popula-
tion in Sweden. Work and Stress 1990;4:41–9.
[51] Svensson HO, Andersson GB. Low-back pain in 40- to 47-year-old
men: work history and work environment factors. Spine (Phila Pa
1976) 1983;8:272–6.
[52] Macfarlane GJ, Thomas E, Papageorgiou AC, et al. Employment and
physical work activities as predictors of future low back pain. Spine
(Phila Pa 1976) 1997;22:1143–9.
[53] Wickstrom GJ, Pentti J. Occupational factors affecting sick leave at-
tributed to low back pain. Scand J Work Environ Health 1998;24:
145–52.
[54] Spyropoulos P, Papathanasiou G, Georgoudis G, et al. Prevalence of
low back pain in Greek public office workers. Pain Physician
2007;10:651–9.