ArticlePDF Available

National Institute for Working Life Ergonomic Expert Committee Document No 1Visual Display Unit Work and Upper Extremity Musculoskeletal DisordersA Review of Epidemiological Findings

Authors:

Figures

Content may be subject to copyright.
arbete och hälsa vetenskaplig skriftserie
ISBN 91–7045–436–1 ISSN 0346–7821
1997:16
National Institute for Working Life – Ergonomic Expert
Committee Document No 1
Visual Display Unit Work and Upper
Extremity Musculoskeletal Disorders
A Review of Epidemiological Findings
Laura Punnett
Ulf Bergqvist
National Institute for Working Life
ARBETE OCH HÄLSA
Redaktör: Anders Kjellberg
Redaktionskommitté: Anders Colmsjö,
Elisabeth Lagerlöf och Ewa Wigaeus Hjelm
© Arbetslivsinstitutet & författarna 1997
Arbetslivsinstitutet,
171 84 Solna, Sverige
ISBN 91–7045–436–1
ISSN 0346-7821
Tryckt hos CM Gruppen
National Institute for Working Life
The National Institute for Working Life is Sweden's
center for research and development on labour market,
working life and work environment. Diffusion of infor-
mation, training and teaching, local development and
international collaboration are other important issues for
the Institute.
The R&D competence will be found in the following
areas: Labour market and labour legislation, work
organization and production technology, psychosocial
working conditions, occupational medicine, allergy,
effects on the nervous system, ergonomics, work
environment technology and musculoskeletal disorders,
chemical hazards and toxicology.
A total of about 470 people work at the Institute, around
370 with research and development. The Institute’s staff
includes 32 professors and in total 122 persons with a
postdoctoral degree.
The National Institute for Working Life has a large
international collaboration in R&D, including a number
of projects within the EC Framework Programme for
Research and Technology Development.
Förord 1
TCO anförde i en skrivelse till Arbetslivsinstitutet (Arbetslivsinstitutets diarie-
nummer 732/95) att den forskning som idag bedrivs avseende belastningsskador
har ingen eller mycket liten relevans för tjänstemän. Erfarenheterna vid TCO-
förbunden och vid TCO:s arbetsskadeenhet visar dock att belastningsskador
förekommer i stor utsträckning på tjänstemannasidan, särskilt hos kvinnor.
TCO begärde en kartläggning beträffande belastningsskador och skaderisk
framför allt hos kvinniga tjänstemän. Därutöver angavs att även psykosociala
faktorers påverkan när det gäller dessa skador skulle närmare undersökas.
Arbetslivsinsitutet avsatte ekonomiska medel för att handlägga begäran från
TCO (Arbetslivsinstitutet E 51/96). En projektgrupp bildades bestående av
chefsjurist Stig Gustafsson och förbundsjurist Lill Dahlberg från TCO:s
arbetsskadeenhet, professor Ulf Lundberg, Psykologiska institutionen, Stock-
holms universitet, professor Francesco Gamberale och professor Mats Hagberg
på Arbetslivsinstitutet. Diskussioner har också förts med professor Niklas Bruun
och jur lic Lottie Ryberg, Arbetslivsinstitutet.
Projektgruppen initierade tre kartläggningar:
”Arbetsskadeförsäkringen – bedömningen i domstol av belastningsskador hos
kontorister och sjuksköterskor” av jur kand Maria Sundström, Arbete och
Hälsa 1997:17.
”Visual Display Unit Work and Upper Extremity Musculosceletal Disorders:
A Reveiw of Epidemiological Findings” av professor Laura Punnett och dr
med sci Ulf Bergqvist, Arbete och Hälsa 1997:16.
”Ländryggbesvär vid sjukvårdsarbete” av fil dr Monica Lagerström, professor
Tommy Hansson och professor Mats Hagberg, Arbete och Hälsa 1997:xx.
Dessa arbeten kan utgöra en bas för fortsatta prioriteringar för fortsatt forsk-
ning inom det angelägna forskningsområde som TCO pekat ut. Ett varmt
tack riktas till alla de som bidragit till framtagningen av de tre dokumenten.
Rapporterna ”Visual Display Unit Work and Upper Extremity Musculo-
skeletal Disorders: A Review of Epidemiological Findings” och ”Ländryggs-
besvär vid sjukvårdsarbete” har godkänts av Arbetslivsinstitutets Ergonomiska
Expert Kommitté, se nedan.
Solna i september 1997
Mats Hagberg
Koordinator
Förord 2
Arbetslivsinstitutets Ergonomiska Expertkommitté har granskat och godkänt
slutsatserna i detta dokument. Kommitténs ordinarie ledamöter är professorer
inom enheten för ergonomi och psykologi.
Francesco Gamberale, professor i arbetspsykologi
Mats Hagberg, professor i arbets- och miljöfysiologi
Åsa Kilbom, professor i arbetsfysiologi
Anders Kjellberg, professor i arbetspsykologi
Jörgen Winkel, professor i tillämpad arbetsfysiologi
Detta dokument har författats av professor Laura Punnett, Lowell University
of Massachusetts, och med sci Ulf Bergqvist, Arbetslivsinstitutet.
Målsättningen för arbetet har varit att med stöd av en genomgång och värde-
ring av föreliggande litteratur om möjligt komma fram till ett dos-respons-
och dos-effektsamband. Fastställande av dos-respons- och dos-effektsamband
är i de flesta fall inte möjligt och då blir uppgiften att i samma förebyggande
syfte utvärdera den litteratur som finns. Det insamlade materialet har värde-
rats och ett dokumentsförslag utarbetats av författarna på uppdrag av kommit-
tén. Förslaget har diskuterats och bearbetats innan det blivit antaget som ett
kommittédokument.
Endast vetenskaplig litteratur som bedöms vara pålitlig och ha betydelse för
just denna diskussion åberopas i dokumentet.
Detta dokumentsförslag har diskuterats med kommittén, bearbetats och vid
kommitténs möte 1997-09-09 antagits som dess dokument.
För Arbetslivsinstitutets expert kommitté för arbetshälso- och ergonomiska
frågor.
Solna i september 1997
Mats Hagberg
Ordförande
Preface
The National Institute for Working Life – Ergonomic Expert Committee has
reviewed and approved the conclusions in this document. The members of the
committee are professors at the Department of Ergonomics.
Francesco Gamberale, professor of work psychology
Mats Hagberg, professor of work and environment physiology
Åsa Kilbom, professor of work physiology
Anders Kjellberg, professor of work psychology
Jörgen Winkel, professor of applied work physiology
For this document the authors, professor Laura Punnett, University of
Lowell, Massachusetts, USA, and Ulf Bergqvist, Ph D, from the National
Institute for Working Life, were appointed by the expert committee as
authors. The authors searched for litterature in different data-bases, such as
Medline and NIOSHTIC. Evaluation was made of all relevant scientific
original literature found. In exceptional cases information from documents
difficult to access were used. The draft document was discussed within the
expert committee and was finally accepted as the group’s document Septem-
ber 9, 1997.
The document aims at establishing a dose-response/dose-effect-relationship
and the effect is based only on the scientific literature. The task is not to give
a proposal for numerical occupational exposure limit value.
The topic of this document was initiated and discussed in a project group
initiated by the Swedish union ”TCO” . Advise to the authors has also been
provided by the TCO-project group, chief lawyer Stig Gustavsson (TCO) and
lawyer Lill Dahlberg (TCO), professor Ulf Lundberg, Department of
Psychology, University of Stockholm, Ronnie Eklund, Stockholm University
Law School, professor Francesco Gamberale, National Institute for Working
Life and professor Mats Hagberg, National Institute for Working Life.
Solna in September 1997
Mats Hagberg
Chairman
Executive Summary 1
Svensk sammanfattning 3
Introduction 5
Musculoskeletal disorders 5
Occupational risk factors for musculoskeletal disorders 6
Individual risk factors 8
Musculoskeletal risk factors in visual display unit operation - an overview 9
Workstation dimensions and input devices 10
Visual demands 11
VDU work load, task design and work organisation 11
Psychological and social factors and stress-mediated effects 12
Interaction between VDU work and gender 13
Exposure categories and effect modification 14
Methods and review of methodology 16
Aim and organisation of this review 16
Definitions of common epidemiological methods and measures 19
Criteria for causality 20
Acquisition of data 21
Exposure contrasts 22
VDU work per se 26
VDU operation in general, compared to non-VDU work 26
Full-time vs. part-time VDU operators 34
Hours of VDU operation per day 38
Duration (years) of exposure to VDU or keyboard operation 44
VDU work in general, compared to low-exposed industrial jobs 48
Physical ergonomic factors in VDU work situations 50
Workstation dimensions and postural stress 50
Keyboard model 58
Use of a mouse or other input devices 59
Visual demands, corrective lenses and monitor placement 60
Types of VDU work and work organisational factors 64
VDU task types 64
Work load and work demand 72
Repetitiveness of keyboard work 77
Rest break patterns and duration of work tasks 81
Monitoring and supervision 84
Psychological and social factors 86
WorkerÕs control and decision latitude 86
Social support and co-operation 89
Fear and insecurity or job dissatisfaction 92
Stress reactions 93
Gender 94
Gender as a risk factor for musculoskeletal disorders 94
Gender as an effect modifier 99
Methodologic considerations and interpretation of findings in the reviewed
epidemiological studies 101
Characterisation of ergonomic exposures 101
Characterisation of work organisational and psychological and social exposures 102
Characterisation of upper extremity disorders 102
Temporal sequence of cause and effect 105
Potential confounding 105
Potential selection bias 106
Intervention studies 108
Summary and conclusions 115
VDU use per se 115
Ergonomic considerations of the VDU work stations 118
Type and organisation of VDU work 118
Psychological and social factors in VDU work 119
Gender as a risk factor among VDU workers 120
Conclusions 120
Recommendations for research 122
Recommendations regarding prevention and compensability 124
References 126
APPENDIX I. Methodological summary of reviewed studies 139
APPENDIX II. Epidemiological studies not included in the review. 155
APPENDIX III. Methodological strength of studies 156
1
Executive Summary
Laura Punnett and Ulf Bergqvist. Visual display unit work and upper extremity
musculoskeletal disorders: A review of epidemiological findings. Arbete och
HŠlsa 1997;16:1-161.
A review has been undertaken of the epidemiological literature on work with
visual display units (VDUs) and neck or upper extremity musculoskeletal
problems among office workers. The questions to be answered are whether there
is an increased risk of such disorders among VDU users, compared with people
working in other types of jobs, and - if so - which specific feature(s) of the VDU
device or the work environment are responsible. In addition, the question of
whether women using VDUs are at even greater risk than men, and if so, why, has
also been investigated.
Comprehensive and up-to-date reviews summarising current epidemiologic
knowledge about such factors for musculoskeletal problems among VDU users
are scarce. This report therefore summarises current epidemiological evidence
concerning increased risks of upper extremity and neck disorders among VDU
users, what specific features of the VDU device or work environment that may be
involved; and whether women using VDUs are at even greater risk than men.
A total of 72 relevant reports from 56 epidemiologic studies have been
identified. Most were published in peer-reviewed journals, although a few studies
have been included from peer-reviewed conference proceedings or technical
reports from national research institutions. Only papers written in English or
Swedish have been reviewed. These studies cover a range of health endpoints,
from non-specific discomfort to median nerve conduction velocity through the
carpal tunnel, and a range of exposures from VDU work per se to specific task
types, rest break patterns, and keyboard configurations. There is also a great
variety in study designs and populations; some studies have compared VDU
operators to other employed groups, such as non-VDU clerical workers, and
others have used internal comparisons, within groups of VDU users, to examine
the effects of many specific features of the VDU work environment. In general,
however, inadequate attention has been paid to the consequences of the choice of
study population and the range of exposures within which statistical associations
or dose-response relationships can be examined.
Some general conclusions regarding VDU work and musculoskeletal didsorders
emerge from this review. These conclusions are supported both by studies of
questionnaire-reported symptoms and studies utilising objective findings from
physical examinations or diagnoses. For disorders of the hand and wrist, we found
evidence that the use of the VDU or the keyboard was a direct causative agent,
mediated primarily through repetitive finger motion and sustained muscle loading
across the forearm and wrist. The odds for such disorders among VDU users with
at least 4 hours of keyboard work per day appear to be about twice that of those
with little or no keyboard work. For neck and shoulder problems, the role of the
2
VDU or the VDU work seems to be less that of a single source of mechanical
stresses and more the central feature around which modern office work is
organised. This increasingly common pattern of VDU work organisation, with
both physical (repetitive finger motions) and work organisational features (task
fragmentation), is perhaps less typical of a number of other work environments.
Thus, VDU work is often used as a proxy variable, that is, as a symbol which
represents this particular combination of exposures.
Although not all specific factors have been adequately studied, either singly or
in combination with each other, there is convincing evidence regarding some.
Strong evidence exists for elevated risks of upper extremity disorders with data
entry and similar intensive keying tasks, and for hand and wrist disorders, at least,
with hours of keying per day. High work demand and postural stress resulting
from poor work-station design and layout also increase the risk of upper extremity
disorders. Thus, there is - in our opinion - a scientific basis that justifies
ergonomic and work organisation interventions to improve work situations
characterised by these conditions.
Among VDU workers, upper extremity and neck musculoskeletal disorders are
more common among women. Specific reasons for this gender difference have
not been fully identified, although they could include differences in job types,
housework and childcare, body size and strength, hormonal or other physiological
conditions. When men and women were compared within fairly homogenous job
groups, they reported similar rates of MSDs. However, few studies have
unfortunately been able to compare women and men doing similar VDU tasks.
3
Svensk sammanfattning
Laura Punnett and Ulf Bergqvist. Visual display unit work and upper extremity
musculoskeletal disorders: A review of epidemiological findings. Arbete och
HŠlsa 1997;16:1-161.
En šversikt har tagits fram šver den epidemiologiska litteratur som inriktats pŒ
bildskŠrmsarbete och muskuloskeletala problem i nacke och švre extremiteter
bland kontorsarbetare. FrŒgestŠllningarna Šr om det fšreligger en škad risk fšr
sŒdana sjukdomar bland bildskŠrmsarbetare jŠmfšrt med individer med andra
arbeten, och - i sŒ fall - vilka specifika fšrhŒllanden kring bildskŠrmen eller
arbetsfšrhŒllanden kring skŠrmen som Šr fšrknippade med detta. Dessutom har
šversikten ocksŒ inriktats pŒ frŒgan om kvinnor vid bildskŠrmsarbete lšper en
hšgre risk Šn mŠn.
Det saknas fšr nŠrvarande fšrdjupade och uppdaterade šversikter šver
nuvarande epidemiologisk kunskap om sŒdana faktorers inverkan pŒ
muskuloskeletala problem bland bildskŠrmsarbetare. Denna rapport sammanfattar
dŠrfšr vŒra nuvarande epidemiologiska erfarenhet i dessa frŒgor.
Totalt identifierades 72 rapporter frŒn 56 epidemiologiska studier. De flesta var
publicerade i tidskrifter med vetenskapligt granskningsfšrfarande, Šven om nŒgra
studier har erhŒllits frŒn konferenssammanstŠllningar med vetenskapligt
granskningsfšrfarande eller tekniska rapporter frŒn nationella forskningsinstitut.
Endast rapporter skrivna pΠengelska eller svenska har tagits med. Dessa studier
tŠcker en rad olika hŠlsoeffekter, frŒn ospecificerade besvŠr till mŠtning av
ledningshastigheten i mediannerven i karpaltunneln, och olika expone-
ringssituationer frŒn bildskŠrmsarbete i sig till speciella arbetsuppgifter,
pausmšnster och tangentbordutformning. Det finns ocksŒ en stor variation i
studiedesign och val av studerade grupper; vissa studier har jŠmfšrt bildskŠrms-
arbetare med andra grupper av anstŠllda, t. ex. kontorsanstŠllda utan
bildskŠrmsarbete, medan andra studier har anvŠnt sig av interna jŠmfšrelser
mellan olika grupper av bildskŠrmsarbetare fšr att utršna effekten av specifika
faktorer i arbetsmiljšn kring bildskŠrmen. MŒnga studier har inte tagit tillrŠcklig
hŠnsyn till hur val av studiegrupper och variation av exponeringar kan pŒverka
analysen och dess tolkning.
Vissa allmŠnna slutsatser betrŠffande bildskŠrmsarbete och belastnings-
sjukdomar kan dras ur denna šversikt. Dessa slutsatser stšds bŒde av studier som
bygger pŒ enkŠtbaserade symtom, och studier som utnyttjat objektiva fynd eller
diagnoser. Fšr sjukdomar i hand och handled fann vi belŠgg fšr att anvŠndning av
bildskŠrm eller tangentbord var en orsaksfaktor, frŠmst erhŒllen genom upprepade
fingerršrelser och vidmakthŒllen belastning i underarm och handled. Oddsen fšr
sŒdana sjukdomar bland bildskŠrmsarbetare med Œtminstone 4 timmars tangent-
bordsarbete per dag tycks vara ungefŠr det dubbla jŠmfšrt med de som har lite
eller inget tangentbordsarbete. Fšr nack- och skuldraproblem tycks bildskŠrmen
eller bildskŠrmsarbetet spela en mindre roll som en direkt orsak till mekanisk
4
belastning, och mer en roll som den centrala punkt kring vilken modernt
kontorsarbete Šr uppbyggt. Detta alltmer vanliga mšnster av bildskŠrmsarbetets
organisation, med bŒde fysisk (repetitiva fingerršrelser) och arbetsorganisatoriska
inslag (fragmentering av arbetsuppgifter) Šr kanske mindre vanligt fšrekommande
i vissa andra arbetsmiljšer. PŒ sŒ sŠtt har bildskŠrmsarbete ibland utnyttjats som
en ÒstŠllfšretrŠdandeÓ variabel, d.v.s. en symbol som representerar denna
kombination av faktorer.
€ven om inte alla specifika faktorer har studerats tillrŠckligt, varken var fšr sig
eller i kombination, sŒ fann vi švertygande belŠgg avseende vissa faktorers
pŒverkan pŒ muskuloskeletala sjukdomar i bildskŠrms-sammanhang. Starka
belŠgg fšreligger avseende škade risker vid inmatningsarbete och arbetsuppgifter
med liknande intensiv anvŠndning av tangentbordet, och, Œtminstone fšr hand-
och handledsproblem, med antalet timmar i sŒdant arbete per dag. Hšga krav i
arbetet och olŠmpliga kroppsstŠllningar som ett resultat av dŒlig utformning av
arbetsplatserna škar ocksŒ risken fšr belastningssjukdomar i švre kroppen. Detta
innebŠr att det - enligt vŒr uppfattning - finns en vetenskaplig grund som
motiverar ergonomiska och arbetsorganisatoriska ŒtgŠrder fšr att fšrbŠttra sŒdana
arbetssituationer.
Bland bildskŠrmsarbetare tycks belastningssjukdomar i nacken och de švre
extremiteterna vara vanligare bland kvinnor Šn bland mŠn. De konkreta orsakerna
till detta har dock inte klarlagts; de innefattar sannolikt bŒde skillnader i
arbetsuppgifter, hemarbete, kroppsstorlek och muskelstyrka eller andra
fysiologiska skillnader. NŠr mŠn och kvinnor med mer likartade arbetsuppgifter
jŠmfšrdes, sŒ var skillnaderna mellan deras rapportering av belstningsbesvŠr
mindre. TyvŠrr finns det fŒ studier dŠr man kan jŠmfšra mŠn och kvinnor med
lika arbetsuppgifter.
5
Introduction
Ever since a more widespread introduction of Visual Display Units (VDUs) in
many workplaces began in the middle seventies, concerns about adverse health
problems occurring among VDU users have been voiced. Some concerns have
been directed towards the VDU unit as such (often in terms of ÒradiationÓ or
electromagnetic fields), while others have been related to changes in work life
that is being brought about by extensive computerisation of the work. Ergonomic
factors have formed a central part of such concern from the beginning, discussed
especially in terms of visual fatigue and musculoskeletal problems. These latter
health issues are probably the most prevalent concerns among VDU users, and by
now, their relationships to work with VDUs appear to be fairly well documented.
This relationship can be phrased in the following way: ÔSome features of VDU
work may cause adverse health or discomfort, such as musculoskeletal problemsÕ.
In order to enable effective interventions or preventive actions, knowledge is
required about what specific factor(s) in the VDU work situation is responsible,
however.
Comprehensive and up-to-date reviews summarising current epidemiologic
knowledge about such factors for musculoskeletal problems among VDU users
are scarce. This report was written in an attempt to fill this need. The literature
review presented here seeks to answer the questions of 1) whether there is an
increased risk of upper extremity and neck disorders among VDU users,
compared with people working in other types of jobs; 2) if so, which specific
feature(s) of the VDU device or work environment are responsible; and 3)
whether women using VDUs are at greater risk than men, and if so, why.
Musculoskeletal disorders
Musculoskeletal and related soft-tissue disorders (MSDs) may affect any part of
the neck or the upper extremity, from the shoulder out to the fingers. They include
a variety of clinical syndromes such as nerve compression or entrapment
disorders, tendon inflammations and related conditions, muscle inflammations,
and degenerative joint disease. They also include less well standardised
conditions such as myositis, fibromyalgia, and focal dystonia, as well as regional,
sometimes poorly localised pain and paresthesia not attributable to other
pathologies (29, 49, 64, 121, 148, 156, 160, 206). The most widely known
peripheral nerve entrapment is carpal tunnel syndrome, which involves
compression of the median nerve where it passes through the wrists or carpal
tunnel; it produces numbness, tingling, pain, and eventually loss of muscle
function in the thumb and first two and one-half fingers of the hand. Other
peripheral nerve compressions may occur in the ulnar tunnel of the wrist, the
forearm, and the thoracic outlet. Tendon inflammatory conditions are generally
known as tendinitis or tenosynovitis; at various locations of the upper extremity
6
they are named for the specific point of inflammation, such as epicondylitis or
shoulder bursitis. Tension neck syndrome (TNS) is the collective term given
several non-articulate syndromes of the neck region. Generally, this syndrome is
characterised by pain and a feeling of tiredness and stiffness in the neck, as well
as tenderness over the descending part of the trapezius muscle (77, 206). Cervical
disorders is also a collective term used (here) to describe cervical syndrome and
cervical degenerative disease or spondylosis, where the former is based on
anamnestic and physical examinations, while the latter would (normally) require
radiographic evidence of disc degeneration (77). However, few of the reviewed
studies appear to have included x-ray examinations.
These disorders are often discussed collectively, in part because they are not
always well diagnosed and therefore distinguished from each other, and in part
because they share several epidemiologic features, including particularly that they
often result from a common group of risk factors or ergonomic stressors and that
they tend to develop after months or years of exposure, rather than as point-in-
time injuries. While they are not rare in the general population, there is evidence
that various occupational groups are at higher risk because of the physical
demands of their work (e.g., (58, 155, 177, 178, 192)).
Occupational risk factors for musculoskeletal disorders
The physical features of work that are frequently cited as risk factors for MSDs
include rapid work pace and stereotyped repetition of motion patterns; insufficient
recovery time; forceful manual exertions; anatomically non-neutral body postures
(either dynamic or static); mechanical stress concentrations (direct pressure of
hard surfaces or sharp edges on soft tissues); vibration; and low temperature.
The literature on soft tissue physiology and biomechanics demonstrates several
plausible pathomechanisms by which these ergonomic factors may injure the soft
tissues of the musculoskeletal system (see the review by Armstrong and
colleagues (4). For example, viscous strain and deformation of tendons
accumulate as a function of work pace (frequency and duration of loading), level
of muscular effort, and recovery time between exertions, indicating the
importance of these dimensions of physical loading for cumulative tissue damage
(5, 35, 67). Symptoms and tolerance of physical work demands for repetitive
wrist motion are similarly determined by hand posture, frequency and force of
wrist bending, and task duration (184). EMG measurements show that the
shoulder muscles fatigue quickly when the arm is elevated at or above shoulder
height, especially if the exposure is prolonged or repeated frequently (73, 76, 194,
201).
The epidemiologic studies of MSDs in workers occupationally exposed to these
generic ergonomic factors are too numerous to list here. Many reviewers of this
literature have concluded that there is substantial epidemiological evidence of the
etiologic importance of occupational ergonomic stressors for neck and upper
extremity MSDs (e.g. (4, 7, 75, 77, 185, 193, 211)), although some authors still
7
dispute the importance of these factors, especially relative to non-occupational
causes (23, 72, 127, 135, 137, 198).
The ergonomic effects of work in seated postures have been studied, in general,
with two distinct questions in mind. One of these is the effect of prolonged
sedentary work or specific seating designs on low back morbidity; since we
restrict this review to upper extremity and neck disorders, we do not discuss that
literature here. The other issue is the nature of the restrictions on postural mobility
that result from the seated position, with the trade-off of greater postural stability
in order to support precise manual and visual work such as are required in
computer use tasks (161). Here we assume that most individuals work at
computers in the seated position, and the importance of workstation dimensions
and other postural constraints in VDU work is considered in that context of
limited mobility.
Another type of stressor that has received increasing interest with respect to
MSDs is that of psychological and social factors. These often occur as
consequences of objectively definable features of the work organisation such as
task allocation, incentive wages, and work pace (see below). Psychological and
social factors have already been demonstrated epidemiologically to have etiologic
importance for cardiovascular disease (97, 106, 173, 195). On the hypothesised
etiologic pathway from psychological and social job characteristics to MSD
development lie several well-known physiological mechanisms which could
explain the associations observed. Among these are adverse circulatory patterns
(173); high levels of sympathetic nervous symptom arousal with general central
nervous system consequences as well as endocrine system impacts on circulating
hormones (12, 30, 79, 107, 209); tonic activation or ÓpsychogenicÓ muscular
tension (9, 202, 207); and interference with normal muscle and tendon repair
processes. The occupational psychological and social stressor most consistently
associated to date with musculoskeletal disorders is decision latitude or autonomy
(24). While understanding of the role of psychological and social factors in the
development of MSDs is still evolving, it is important to assess these work
environment factors and include them in an overall evaluation of job stressors.
Psychological and social variables used in the studies reviewed here are
typically based on both the worker«s own perceptions and his/her judgement
about the situation - i.e., cognitive, perceptual, and affective processes may all be
involved. Psychological and social dimensions of the work environment are
commonly divided into the following subcategories, mainly based on the work by
Karasek and colleagues (96):
· The psychological demands of the job, including the amount of work and the
time available to complete it.
· The worker«s opportunities to exercise control in the job, defined variously as
influence, job control and decision latitude.
· The degree of social support by supervisors or workmates.
8
· Job insecurity, the worker«s fear of being replaced, or perceived opportunities
for future employment.
Individual risk factors
The most important non-occupational risk factors are age, gender, socio-economic
status, certain systemic diseases, and for carpal tunnel syndrome possibly female
endocrine conditions; other factors of possible importance for upper extremity and
neck MSDs include race or ethnicity, smoking, alcohol use, obesity, and
recreational sports (25, 37, 43, 155, 177, 178, 192, 199).
The possible impact of gender as a risk factor or an effect modifier for MSDs is
- by request - specifically considered in this review. Otherwise, the main emphasis
of this review is on occupational factors, in particular, specific aspects of VDU
work. Socio-economic factors may be closely linked to occupational demands in
general population studies, and could thus cause some confounding. Likewise,
individual factors such as age may also cause considerable confounding. Apart
from such considerations this review has not attempted to present a com-
prehensive overview of possible associations between non-occupational factors
and MSDs.
9
Musculoskeletal risk factors in visual
display unit operation - an overview
Work at a visual display unit (VDU) represents a complex, multifaceted physical
work environment, with interactions among the various dimensions of the work
station and equipment, speed of data entry, position and lighting of visual targets
(screen and documents), and job content. As in the classic Óperson-machine
environmentÓ model, these features of the work interact as a dynamic system and
together determine the presence and intensity of ergonomic exposures, both in
physical and psychological or social terms. The widespread use of VDUs makes it
important, for public health reasons, to understand these different components and
their interactions in terms of adverse health effects. For some further discussion,
see e.g. Gerr (65), Hendrick (80) or Smith and Carayon (181).
Musculoskeletal
health
Work organisation
and work tasks
Individual
factors
Physical ergo-
nomic factors
Psychological
and social
conditions
Figure 1. Some interrelationships between different aspects of VDU work, that are
important for the aetiology of musculoskeletal pains.
A conceptual model of various features relevant to the development of
musculoskeletal problems among VDU users is presented in figure 1. The model
provides a structure used within this review for ordering the information. Among
physical ergonomic factors, considerable emphasis is given in a number of studies
to certain postural determinants, especially workstation design features, input
devices and visual demands - as these features are often major sources of physical
strain in VDU work. Likewise, some work organisation and work task descriptors
are also typical of many work situations utilising VDUs. Psychological and social
factors encompass a broad range of influences. We are here primarily concerned
with factors which are prevalent in and have been studied in VDU jobs. Some of
these may be specific for VDU work while other may be prevalent also in other
types of work. Of individual factors, apart from gender, which is treated here in
10
some detail, special consideration is given in some studies to the use of corrective
glasses.
While separating the various risk factors in accordance with figure 1 may help
organise the available information, it must not be taken as a statement that these
factors can a priori be seen as independent. For example, when describing a
situation where visual demands may influence posture, it is necessary to include a
large number of factors from different parts of figure 1, such as the workstation
layout, the specific task performed, the familiarity of the operator with the
keyboard (need to look at it or not), use and type of corrective glasses, presence of
glare, work load and stress and pre-existing neck pain.
A further discussion of some of these risk factors is found below.
Workstation dimensions and input devices
The keyboard is the primary mode of data input to the VDU. Its location, height
and slope, in combination with other workstation dimensions, determines the
angles of the wrist, elbow, and shoulder joints and the magnitude of static muscle
loading as the operator maintains the arm positioned over the keyboard (8, 11, 12,
14, 34, 68, 69, 119, 126, 129, 175, 186). For example, placement of a keyboard on
a work surface that is too high causes shoulder elevation, elbow flexion and wrist
extension. In experimental sessions, a poor workstation layout resulted in
markedly greater upper extremity pain than a workstation adjusted to optimise the
operatorÕs posture (38), while ergonomic workstation modifications improved
postures (51). In addition, keyboard height and other dimensions may lead to
contact pressure at the base of the wrist with subsequent bone and nerve damage
(e.g. (167)).
Keyboard operation requires inherently repetitive hand motion (finger flexion
and extension) in order to depress the keys. Recent laboratory studies show
increases in EMG levels and pressures within the carpal tunnel as a consequence
of finger keying motions, as well as in ulnar deviation, flexion and extension of
the wrist (157, 158, 159, 209). Considerable attention has been given to the pos-
sibility of alternative designs, such as the split keyboard or keyboard with
different key layouts (i.e. other than the standard QWERTY layout), in order to
reduce muscle loading (13, 31, 51, 53, 63, 107, 134, 162, 209).
Non-keyboard computer input devices, such as the increasingly used computer
Ómouse,Ó may also be related to additional ergonomic strain. Sustained pinching
required to hold and move the mouse may cause increased tendon tension and
sustained finger muscle activity (74, 91, 92). Disadvantageous upper extremity
postures such as shoulder rotation and flexion and ulnar deviation of the wrist
may also occur as a result of positioning the body so that all components of the
computer work station can be utilised simultaneously (99). Fernstršm and co-
workers (52) measured shoulder and forearm muscular load in a laboratory study
comparing the use of mouse and trackpoint (small joy-stick located in the center
of the keyboard) pointing devices in word processing tasks. They found that work
with the mouse produced higher loading in the shoulder muscles. Shoulder load
11
could be reduced by either use of the trackpoint device or a movable arm support
while operating the mouse, but both of these alternatives increased the muscle
load in the hand and forearm. It should be kept in mind that for someone using
both a keyboard and a mouse, the mouse work represents an added stressor that
acts in combination with the keyboard operation.
Visual demands
The other main mode of interaction is the visual contact with the computer screen,
which may have consequences in terms of postural demands. Several components
of the visual work situations can contribute to this including reading at a
downward angle, use of corrective lenses, screen glare, and the need for visual
contact with several task objects (screen, document, keyboard etc.).
A classical problem is that of bifocals not designed for the VDU working
situation causing a backwards position of the neck in order to utilise the lower,
ÓnearÓ segment of the glasses. Experimental studies have indicated increased
EMG signals and muscle discomforts associated with incorrect glasses and visual
work (113, 114). Furthermore, multifocal lenses have been shown to induce
higher neck muscle tension than monofocal glasses (85, 86), whereas monofocal
glasses may cause increased visual problems among older individuals (17).
Another area of potential conflict between visual and neck comfort may arise in
relation to screen height; De Wall and co-workers (40) indicated that a VDU
placed at eye level or higher would result in a better sitting posture, while
observations of preferred postures suggested a lower VDU screen position (69,
143) - where presumably also the visual demands of the individuals were taken
into account. Some experimental and epidemiological investigations also suggest
that a VDU placed high would increase visual discomforts (2, 17, 112, 149, 196,
200, 208).
VDU work load, task design and work organisation
Different types of VDU work exhibit obvious differences in physical and mental
demands, such as the amount of data input performed or of visual information
processing. The categories proposed originally by the NRC Panel on Impact of
Video Viewing on Vision of Workers (140) have often been used as an
approximate characterisation of different task designs; Ódata entryÓ, Ódata
acquisitionÓ, Óinteractive communicationÓ, Óword processingÓ and Óprogramming,
computer-assisted design and computer-assisted manufacturingÓ. These categories
represent not only differences in work content or objectives, but also potential
differences in ergonomic exposures: speed of keying, lack of variability in motion
patterns, opportunities for rest breaks, and decision latitude, among others.
Task design has been shown to have at least short-term consequences for the
operator in VDU work. For example, rapid speed of typing on a keyboard (as is
often found in intensive data entry jobs) increased muscle loading (measured by
EMG) as well as perceived exertion and discomfort levels (62). Incentive pay,
12
compared with non-incentive pay, significantly increased operatorsÕ tension in
data entry tasks; both upper extremity discomfort and tension were also found to
increase linearly as a function of the hours per day spent performing the task,
regardless of the pay system (171). The frequency and nature of rest breaks
further influence both the discomfort experienced (171) and the total measured
load on the neck and shoulder muscles (76, 194).
Psychological and social factors and stress-mediated effects
Worker«s control and decision latitude, social support and co-operation and fear
and insecurity are - in this review - to be found under this heading. In addition,
stress symptoms as an intermediate descriptor is also included here. Some other
factors closely linked to this area are found in the section of work organisation
(above). The distinction between work organisational and social or psychological
parameters is not always obvious and the nomenclature varies among authors. The
terminology used in this text is for convenience and an attempt to adhere to
common usage in much of the literature reviewed. In principle, a distinction
should be possible between observable job organisation features in the external
work environment and intermediate effects that are experienced subjectively by
the individual. However, in epidemiologic studies the distinction may be less
clear-cut, since the intermediate psychological effects are often at least partially
consequences of the objective environmental conditions, and furthermore because
the dimensions actually measured by the investigators may be a mix of objective
and subjective variables.
For example, work pace can be quantified objectively, such as in terms of
keystrokes per hour, yet the effect may actually be more dependent on the
workerÕs experience of time pressure. Monotonous work has both physical
consequences (repetitive loading on the same soft tissues) and psychosocial
(boredom, low decision latitude). Bergqvist et al. (see below) obtained several
different characterisations of work pace and stereotypy by questionnaire from the
members of their cohort, and found an elevated risk of neck/shoulder symptoms
with the combinations of data entry work with limited opportunity to take rest
breaks and VDU use for at least 20 hours per week with repetitive movements
(19). Thus, combinations of exposures may interact with each other, and their
joint effect cannot be neatly assigned to a single category of risk factor.
It should be noted that although the model(s) proposed by Karasek et al. (96)
have been frequently quoted, few of the analysis reported in this review have
actually utilised the specific suggestions in that model, that it is the combination
of a high job demand and low decision latitude that is the cause of worker strain.
Most analyses have instead used job demand and decision latitude (control)
variables as separate (independent) factors. On the other hand, a critical test of the
Karasek model can be found in the study by Carayon (33), where such
combinations apparently failed to produce worker strain (including Óphysical
healthÓ) in excess of additivity, i.e. in excess of what should be expected if these
factors worked independently (see that study for further discussion). There is a
13
comment in the only study here that investigated such interactions (47), stating
that decision latitude did not significantly modify the effects of workload - thus
essentially agreeing with the comments of Carayon (33).
As pointed out by Frese (60), there are some conceptual issues that need to be
resolved when studying stress-mediated health effects and VDU work:
· First of all, the work performed is a vector for or primary cause of stress in
VDU situations; current understanding does not allow for a direct and general
effect of VDUs on stress or stress-mediated effects independent of the work
performed. In other words, the psychological demands and opportunities for
decision-making are determined by specific aspects of the job description,
hardware components and software packages used by the operator. Thus,
different studies that examine groups of workers doing different types of VDU
work should perhaps not be expected to result in similar findings regarding
psychological and social factors and stress-mediated musculoskeletal effects.
· Secondly, the dynamic relationships between the work content and the stress-
mediated reaction(s) could include an acute and reversible effect, an
accumulated effect, an adjustment, and a latent effect (appearing only after
cessation of the exposure). For example, an individual may experience
immediate stress when a new computer system is introduced. This may lessen
over time, as she/he both learns the necessary skills and adapts his/her work
pace and approach to accommodate the system«s demands. Nevertheless, long-
term frustration may build up, some of which may, in fact, not be apparent to
the user until after she/he is no longer economically dependent on that job. Few,
if any studies have specifically examined this, even if indirect suggestions
relevant to these models could be derived from studies that characterise VDU
work duration in hours per week vs. number of years employed.
· Finally, the studies on psychological, social and work organisational factors
reviewed here have basically been concerned with such factors that do also - in
principle - occur both in VDU and non-VDU work situations, even if their
prevalence may vary. Stressors unique to VDU work have been studied in a few
other investigations, one example being computer failure and adrenaline/
epinephrine responses (90), but not, to our awareness, in relation to musculo-
skeletal endpoints. Other factors such as abstractness of work, understanding of
work processes, Óvirtual reality,Ó etc., have been discussed in general, but - so
far - with little specific application to studies of musculoskeletal problems.
Interaction between VDU work and gender
Many studies on upper extremity and neck musculoskeletal disorders have found
higher prevalences among women than among men. (This observation is not
specific to VDU work.) However, the variable gender may be confounded by
occupational demands, such as less variety in work content or lower decision
latitude (96), or serve as a proxy for differences in exposure or other unmeasured
variables (128).
14
Possible causes of such observations can tentatively be summarised in the
following groups:
· Differences in task type allocations or work tasks between men and women.
· Higher physical or stress load of women from non-work activities such as
childcare and household work.
· Physiological differences, such as different body size or mass or endocrine
functions.
· Differences in the willingness to report or seek medical care for pain or
discomfort.
Utilising the variable ÓgenderÓ to evaluate the difference between men and
women in a study does not, by itself, help to clarify these possible explanations.
In a few studies only, added insights have been achieved due to the inclusion of
variables describing some of the specific factors above, such as child care or
presence of children at home, or details of work task design.
The possible differences between men and women in the risk of upper
extremity and neck musculoskeletal disorders related to VDU work can be
described in two different - but complementary - ways:
· Do women working with VDUs have a higher occurrence of such disorders
than men who work with VDU? In other words, is gender - or some factor(s)
related to gender - a risk factor for such disorders, conditional on exposure?
· Is the risk associated with VDU work different for men and women? That is, is
gender an effect modifier of work related risk factors?
Exposure categories and effect modification
In the reviewed studies, major attention is often given to some ÓglobalÓ
descriptors of VDU work; whether work is performed at a VDU or not, and/or the
number of hours per day (week) that is spent doing work on the VDU. Such data
may be used to evaluate the possibility that VDU work - irrespective of specific
details - may increase the occurrence of muscle problems. Nevertheless, the large
variations found between different VDU work situations make it imperative to
look in more details also at more specific work descriptors.
Therefore, it is often appropriate and necessary to compare multiple dimensions
of exposure in order to examine adequately the specific features of VDU use that
are associated with upper extremity morbidity. Combinations of exposures may
result in an effect that is the sum of the effect of each factor applied separately,
but it may also result in an interaction, where the effect of both factors together
are greater (or less) than the sum of each. There are also examples of a conditional
association, where one factor has an effect only if another factor is present. (See
also the discussion on the Karasek model above.)
15
As already indicated above in the case of work load, separating exposure
variables into categories such as ergonomic factors, task design/work organisation
and psychological and social factors etc. is not always straightforward -
substantial conceptual overlap may exist. Furthermore, it must be remembered
that VDU work is an integrated situation, where clear distinction in terms of these
categories do not exist. These categories are kept for the purpose of organising the
large material available, but should not be seen to imply that factors placed in
different categories are really independent, either in their origin or in their impact.
Many of these various measures of exposures (rapid work pace, static muscle
loading, limited job control etc.) are also not necessarily unique to VDU work; in
epidemiological studies, they should ideally be characterised for the non-VDU
users as well as for the operator group.
16
Methods and review of methodology
Aim and organisation of this review
In the past ten years, there have been few comprehensive reviews of the
epidemiologic evidence bearing on VDU work and its musculoskeletal health
effects. The reviews published to date have primarily described the ergonomics
issues and recommendations for work station layout and chair design (8, 34, 126).
Two articles partially summarised the epidemiology about 10 years ago (77, 127).
This review was therefore designed to examine critically the epidemiologic
literature on musculoskeletal problems of the upper extremity and the neck and
the occupational use of VDUs, in order to evaluate the strength of the evidence
with respect to causal inference and to identify the specific physical,
psychological and social demands or job features of VDU work that might be
associated with these disorders.
The review is based on analytic epidemiologic investigations that compared the
frequency of one or more upper extremity disorders among VDU operators or
between workers with occupational exposure to keyboard use and persons without
exposure either to keyboard use or to other significant ergonomic stressors acting
on the upper extremity. The following were excluded:
· analyses of other health endpoints (back pain, psychological mood states,
reproductive disorders, sick leave from all causes, etc.);
· descriptive or ecologic studies of exposure or of health status that did not
contain analyses of the relation between exposure and musculoskeletal health;
· experimental studies with only short-term outcomes, such as muscle activation
forces, productivity or error rates, operator satisfaction or discomfort during the
experiment; and
· studies of upper extremity disorders in which keyboard operators were
combined with other occupational groups and the data could not be separated.
Studies comparing upper extremity disorders among VDU users before and
after well-defined ergonomic or work organisational intervention were also
included. We have made an effort to include not only positive, but also non-
positive (ÒnegativeÓ) findings from the reviewed studies. However, we can not
assume that all non-positive findings were reported by all authors, and we can
therefore not fully exclude the possibility of a publication bias.
As described above, a large number of various predisposing factors have been
implicated as causes of these disorders. For the sake of discussion, they can be
seen as belonging to one of three groups; a/ individual or physiological factors, b/
ergonomic factors related to the work organisation or the physical work station,
and c/ psychological and social stressors in the work environment. As discussed
17
above, the distinctions between such groups are - in practice - less clear, partly
because they can be seen to exercise their impact in combinations, and partly
because many variables used in the research are descriptors that, in themselves,
cover several of these factors. Some examples are ÓVDU work,Ó Ódata entry
work,Ó Ógender,Ó Órepetitive and monotonous workÓ. Within the framework of
this review, we discuss the groups of possible risk factors with the following
definitions and caveats:
Work performed using a visual display unit (VDU) has been described in some
studies only as ÓVDU workÓ without further specification of task content or types
of exposures present. Since we are restricted by the information available in the
authors« presentations, these papers must be summarised in a general section on
VDU operation, including;
· VDU work vs. other (non-VDU) work situations;
· full-time versus part-time VDU work
· hours of VDU work per day or per week; and
· the duration (years) of employment in VDU work.
Papers that provide more specific and etiologically relevant descriptions of the
work environment are discussed under each of the headings to which they
contribute information.
Physical factors, as they occur and have been studied in VDU work situations,
are a varied group of characteristics. We include under this heading results of
studies on:
· Workstation dimensions and upper extremity postures that results from the need
to accommodate to those dimensions;
· The specific design or model of the keyboard.
· The use of a mouse or other input device than a keyboard; and
· Visual work demands, the use of corrective lenses, monitor placement, glare on
the screen, and the postural stresses that originate specifically in these aspects
of visual work demands;
As mentioned previously, several aspects of work organisation which are
conceptually similar are often found to be strongly correlated with each other in
actual workplaces, and therefore their effects may be very difficult to separate in
epidemiologic studies. For example, fast work pace, infrequent rest breaks, and
highly monotonous work are often found in data entry jobs. Nevertheless, to the
extent possible we have attempted here to address these separate dimensions in
separate sections. These include:
· Type of VDU task (e.g., data entry versus interactive communication).
18
· Work load and work demands, including factors such as time pressure, work
pace, typing speed, extensive overtime, presence of deadlines, insufficient
resources and requirements for Óclose attention;Ó
· Repetitiveness or monotony, which covers physical stereotypy of motion
patterns; low task variability (work flexibility, work variance, and task
rotation); and the degree of skill utilisation (the routine character of the work
and whether the work material Ómakes senseÓ or not);
· The temporal organisation of the VDU work, with specific factors such as work
task duration and rest break patterns; and
· External control mechanisms such as electronic monitoring, quality of
supervision, and production quota systems.
Psychological and social factors are less clearly VDU-specific factors;
nevertheless, a number of observations testify to the commonality of their
occurrence in many VDU work situations, probably as a consequence of the
changes in work organisation that often accompany computerisation of clerical
tasks. This motivates - in our view - a summary of research findings dealing with
such factors in VDU work situations, as defined above:
· Decision latitude and influence within the organisation;
· Social support from supervisors, peers or family and co-operation among
colleagues; and
· Job insecurity, fear of job loss and job dissatisfaction.
Note that job demands are often considered in the category of ÔpsychologicalÕ
stressors. However, because of the close connection between work load and task
design - and thus with physical ergonomic conditions such as work pace - this
aspect has been covered under the work organisation heading. This should be
understood as an attempt to avoid a false dichotomy between the physical and
psychological aspects of work load, as explained above.
Gender can be investigated either as a direct risk factor for musculo-skeletal
problems or as an Óeffect modifierÓ for other risk factors. This review is limited
to conditions and research findings as they have appeared in VDU work
situations.
A large number of other factors, such as age or non-occupational recreational
activities and smoking, are dealt with only as confounders, i.e., we are here in
principle not concerned with their impact on muscle problems per se, but only to
adjust for their possible impact on the research findings relative to VDU work-
specific factors.
Subsequent to the review of etiological findings and a reappraisal of the
methods used in these studies, a short description of a few intervention studies is
provided and summarised (see Table 19). It should be noted that this literature is
very limited to date.
19
Definitions of common epidemiological methods and measures
Some general comments are appropriate on epidemiological research methods, in
order to familiarise the reader with the concepts. The fundamental principles of
study design and measures of risk are widely accepted. However, it is
acknowledged that different researchers may take different approaches, especially
with regard to statistical analyses, and this summary is not intended to address the
merits or demerits of competing choices of statistical measures. The interested
reader is referred to the general epidemiologic literature for further discussion (for
example (103, 164) or other texts.)
The prevalence of a disorder describes how many - out of a certain group - that
have this at a given point in time, divided by the total number in the group. The
odds is a similar measure, describing how many have the disorder divided by the
number of individuals in the same group that do not. The incidence, on the other
hand, describes how many - out of a certain group that does not have the disorder
at the start - develop it during a specified time, again divided by the total number
in the group. The term ÒriskÓ is sometimes used in a loose manner to describe any
of these concept (Órisk of havingÓ, vs. Órisk of gettingÓ).
Just reporting the prevalence of a muscle disorder among VDU users would not
be very informative, without showing the background or expected prevalence.
Thus, epidemiological studies - as reviewed here - are basically comparisons of
these measures in one group with a defined exposure with the same measure in a
nonexposed group, to indicate the impact of that factor. The results of such
comparisons may be expressed as a prevalence ratio (PR), odds ratio (OR), risk
ratio (RR, sometimes referred to as relative risk) or incidence density ratio. These
ratios describe the disease frequency in the ÓexposedÓ group divided by the
frequency in the ÓnonexposedÓ group. A ratio of 1.0 indicates that the prevalence
(odds, incidence) are the same in both groups, suggesting that the factor(s) that
was investigated does not increase the risk of disease. The uncertainty in this ratio
is often given as a confidence interval - the ratio is ÓreasonablyÓ expected to fall
within this interval. As already suggested, many other measures of comparisons
exist, such as correlation coefficients, p-values from tests of significance etc.
To a large extent, the choice of a measure of effect depends on the study design
and the types of data that have been collected. Case-control studies can only
generate odds ratios, while the calculation of incidence (ratios) requires that the
study population has been followed for a defined period of time. In cross-
sectional studies, prevalence rather than incidence of disease is estimated, since
the participants are studied only at one moment in time. However, prevalence data
can be compared with either prevalence ratios or odds ratios, and there is
currently some discussion among epidemiologists about the situations when it is
and is not appropriate to report odds ratios. Without burdening the reader unduly,
we simply note here that when more than one of these measures can be calculated
from the same study, they will not give the same numerical results, even though
they are mathematically related to each other. Thus the exact value of the measure
20
of effect should not be given undue importance; in this review we are more
interested in the general consistency among the findings from multiple studies.
For epidemiological comparisons to be valid, the groups must be comparable in
the way that they have been studied and in their background risk of disease,
otherwise a bias may be said to exist. This could be due to the erroneous selection
of individuals in the groups, or to the presence of other factors that influence the
results (confounding). Say that one group is younger than the other, and the risk
of disease changes with age, then this would be called confounding by age -
causing uncertainty as to whether the difference between the groups is due to the
exposure or to age. Various methods exist to adjust for this. When the phrase
Óadjusted odds ratioÓ is used, it indicates that the odds ratio has been adjusted for
one or more specified factor - thereby reducing or removing the confounding
impact of that factor. Failure to adjust for age, gender, or other variables, when
necessary, is another reason that the estimated measure of effect may not have the
true (correct) value.
More detailed and specific comments as to epidemiological methods are given
in the text below at appropriate places and also summarised below (page 101-
107).
Criteria for causality
A statistical association found by epidemiologic methods between an exposure
and a disorder does not necessarily mean that this association is causal. The
establishment of causality requires the ruling out of alternative competing
explanations for the observed association, interpretation of the findings in light of
what is known about the biological processes involved (often from experimental
studies), as well as replication of the epidemiological findings.
There have been several proposed sets of criteria for determining causality, the
most well-known being that of Hill (82), which includes considerations of the
strength of associations, consistency between observational studies (replications),
specificity of effect from a factor, temporality - the exposure precedes the effect,
biological gradient or Òdose-responseÓ, biological plausibility and coherence,
experimental evidence and analogy with other known processes. As pointed out
by Hill, as well as others (e.g. Rothman, (164)), these criteria are not strict in the
sense that they Òbring indisputable evidence for or against the cause-and-effect
hypothesisÓ (apart from the criterion of temporality). Establishing causality is
fundamentally a process of judgement.
To formalise such a judgement process, the terms establishedÓ, ÒprobableÓ and
ÒpossibleÓ when describing the causality of an association, based on both epi-
demiologic and experimental findings, have been used by groups such as the
International Agency for Research on Cancer (IARC). The results of evaluations
of epidemiologic studies alone are, however, not described in such terms. In this
review of epidemiologic findings, we have therefore refrained from such
formalised statements, as they would necessitate a detailed review also of existing
experimental studies and other evidence in line with the above process.
21
Nevertheless, the outline of such processes has served as a general guidance in
our evaluations.
Finally, it should be emphasised that a study that fails to find an association
between an exposure and a disease does not necessarily prove that none exists.
There are two different situations where caution in such an interpretation is
warranted. First, with studies of limited power, although the estimator may be
close to the null value, the wide confidence interval would not give any
confidence in the statement. Secondly, various types of systematic errors may
cause an underestimate of the effect, i.e., the risk is estimated as closer to the null
(expected) value than it really is. (See further discussions by e.g. Ahlbom et al.
(3)). (See also the section below on Òexposure contrastÓ.)
Acquisition of data
Epidemiologic studies were sought through computerised databases (NIOSH-TIC,
MEDLINE, ARBLINE); in the citations of review articles and articles found from
the literature search; in the authorsÕ personal files; and by contacting researchers
active in this field.
At this stage, articles were included even if they had serious methodological
shortcomings or failed to define the population or methods sufficiently for
evaluation of potential misclassification or bias. In the Summary and conclusions
section and Appendix III, we indicate which studies were judged to be
methodologically valid and informative (III A), and which were too weak to be
relied upon (III B), together with comments regarding the reasons for this
selection. Further details are found in Appendix I, and these are also discussed in
the appropriate sections of the review.
A few investigators were contacted with queries regarding study methods and
findings. Where necessary, relative risks (RRs), odds ratios (ORs) and tests of
linear trend in effect with exposure level were calculated by the authors from raw
data in the articles. In some cases the raw data had first to be estimated from
graphs.
Several studies were compared to an external population that had estimated the
prevalence of hand/wrist disorders in low-exposure (industrial) jobs (176), since
all of the authors had used a comparable case definition. This permitted an
external comparison of the disorder frequencies in VDU or keyboard work with
the expected or background prevalence in the U.S. working population.
Descriptions of the design of those observational studies that were judged to be
relevant to the question of musculoskeletal disorders and VDU work are found in
Appendix I. The results of these studies are tabulated in Tables 1-18 and
commented on in the text. Intervention studies are tabulated in Table 19. A few
epidemiological studies that were excluded from the review - and the reasons for
the exclusions - are found in Appendix II. Finally, Appendix III separates the
methodologically stronger studies that - in our judgement - should form the basis
for conclusions from other studies.
22
Exposure contrasts
In the studies on upper extremity and neck disorders and occupational VDU use
that met the search criteria, a variety of exposure conditions were compared by
the original investigators (see Appendix I). In some studies, musculoskeletal
disorder rates were compared between VDU users and clerical employees who
did not use keyboards, permitting analysis of the main effect of keyboard use,
overall, or subdivided into data entry vs. conversation, hours per day, etc. In other
studies, internal comparisons of postural stressors and other job features were
made among VDU operators; this design does not allow analysis of the effect of
VDU use per se, but does permit identification of factors that increase the risk of
upper extremity disorders within the broad range of activities called ÓVDU
operationÓ. Some of these may also be thought of as effect modifying factors that
might also be independent risk factors in the absence of keyboard work.
VDU work hour contrasts
Figure 2 illustrates three possible comparisons of Òhours of VDU work performed
per dayÓ, based on differences in the underlying task distribution of the study
population(s).
Situation a/ shows a work-force in which individuals spend very different
numbers of hours per day at the VDU, ranging from none to 8 hours but covering
all the possibilities in between. Often, an investigator will decide to divide this
Hours of VDU
work per day
0 2 4 6 8
a/
b/
c/
Figure 2. Some examples of different contrasts in terms of hours of estimated VDU work
per day. In situation a/, individuals in the study cover the full range of VDU working
hours, and the delineation has been (arbitrarily) set at 4 hours, for a comparison between
Òworking more than 4 hours/dayÓ with Òworking less than 4 hours/dayÓ. In situation b/
the distribution of working hours is bimodal, with two distinct groups of VDU users (Òat
least 6 hours/dayÓ vs. Òless than 2 hours per dayÓ). Situation c/ essentially describes full-
time VDU users compared with office workers without VDU use.
23
range in the middle - in this example, to compare those doing VDU work for at
least 4 hours with those at the VDU for less than 4 hours per day. However, what
cannot be seen from this diagram is what number of people are at each level of
exposure, which can have a large effect on the study results. For example, one
possibility is that most of the study participants actually work from 3 to 5 hours
per day at the VDU. In this case, even though the contrast is labelled "4 to 8 hours
versus 0 to 4 hours," in reality it is closer to "5 hours per day versus 3 hours per
day." Because there is only a narrow range in the exposure variable, there is very
little power to study the effect of duration of VDU work, and the risk ratio will be
smaller than might be expected, even if there is a strong underlying exposure-
response relationship.
On the other hand, if the study subjects are spread uniformly over the range
from 0 to 8 hours per day of VDU work, then the same apparent comparison will
have more information because there will be a bigger difference on average
between the exposure levels of the two groups. In fact, if it is big enough, this
population would ideally be divided into more than two levels of exposure, in
order to study the shape of the exposure-response trend (i.e., to see if the disease
rate increases steadily as the level of exposure increases).
A third possibility exists if close to one-half of the subjects work less than one
hour per day at the VDU, and almost half work 7 to 8 hours per day, with very
few people in the middle. This situation will actually be more like figure c/ in
reality, even though it is still labelled as "4 to 8 hours versus 0 to 4 hours," and it
will likely give a much bigger relative risk estimate because the contrast in actual
hours worked is so much greater (again assuming that there is a true exposure-
response relationship).
Situation c/ would more or less describe many studies found in the section
ÒVDU operation in general, compared to non-VDU workÓ (if the VDU operators
are full-term users of VDUs). A complication is that in this case, different choices
of the comparison group (Ònon-VDU usersÓ) may also lead to different results, as
discussed further on page 71-72.
Epidemiologic investigators must generally study "natural experiments" - in
this case, the work situations of real people - rather than being able to decide the
number of people who will be exposed for each daily duration (or to any other
occupational factor). Thus, even though figures b/ and c/ illustrate cleaner
contrasts that are simpler to analyse, these situations are not always available for
study. The epidemiologist must decide how to analyse and present the available
data in order to learn as much as possible about the risk factors in a given work
situation.
In addition, it would not be desirable if only situations like b/ and c/ were
studied. Both of them show workplaces where at least one half of the population
(the highly exposed) probably performs very monotonous work, spending almost
the entire workday at the VDU. However, we are also interested in knowing
whether the risk changes when people's jobs are more varied. If the "non-VDU
work" performed by the study subjects does not involve some other type of
24
repetitive hand motion, then the group of people who spend neither all nor none
of the work day at the VDU provide us with an important opportunity to learn
about the variety in activity that might reduce people's risk of musculoskeletal
disorders.
Thus, a large number of contrasts are possible and have been studied in the
literature reviewed here. We have attempted to categorise these contrasts
appropriately in the various tables and to discuss in the text the actual contrast
provided by each study. Unfortunately, the actual population distribution of VDU
work hours has not been reported in all of the reviewed studies.
Contrasts in other exposure factors
A recurrent problem in several studies is that of inadequate contrast in terms of
various specific exposure factors. If all or most of the study population is
characterised by similar conditions in terms of one factor under study, then
obviously that study will not provide adequate information as to the impact of
variations in that factor. Figure 3 provides some graphical illustrations. This issue
of contrast is, of course, factor specific, since one study may provide adequate
contrast in one factor but not another (see figure 3a and b). Ideally, of course,
studies should have adequate contrast for all factors being investigated (figure 3c),
practical consequences e.g. in terms of populations available for study may often
make this difficult. It is, however, essential that such limitations in variations in
some factors are taken into considerations, especially when examining ÒnegativeÓ
results - i.e. data suggesting the absence of associations between some factors and
disorders. (In Bergqvist 1995b (20), the authors comment that Òsince the study
group may be favourably selected for, e.g., certain ergonomic conditions, it may
not be appropriate to generalise from the lack of associations.....Ó.) It should also
be recognised that studies of interactions between two factors would - generally -
require a broad range of contrasts in terms of both, and little correlation between
them in the population under study.
25
Mental
load
Physical load
b/ Narrow range of contrast in mental
load, broad in physical load
Mental
load
Physical load
a/ Broad range of contrast in mental
load, narrow in physical load
Mental
load
Physical load
c/ Broad range of contrast in both
mental and physical load
Figure 3. Contrasts in two factors, using Òphysical loadÓ and Òmental loadÓ as two
examples of exposure factors. In study a/, only mental load can be adequately
investigated, and its estimated effect may be conditional on the physical load experienced
by the study subjects. In study b/ only physical load may be studied, and there may be a
similar lack of generalisability to other workplaces because of the specific level and
narrow range of mental load. Study c/ would enable investigations of both physical and
mental loads as well as possible interactions between them, assuming that there is a
sufficient number of subjects in relevant combinations of these factors.
26
VDU work per se
The most common comparison in the studies identified was that of VDU use
versus no or little VDU work in offices or other similar occupations. VDU use
was defined variously, for example, at least 4 or 5 or 8 hours per day, and the
reference groups were also defined in various ways, such as no use at all or less
than 2 or 4 hours per day. Thus the magnitude of the estimated relative risks were
likely affected by the study design and the exposure differentials between study
groups, as already discussed above in the section of exposure contrasts (page 22-
25). In an attempt to clarify these distinctions, we list results for VDU work
compared to no VDU use (table 1) separately from the comparison of full vs. part-
time VDU work (table 2); the effects of hours of VDU use per day or week (table
3) here represents exposure-response relations estimated from at least 3 levels of
exposure.
VDU operation in general, compared to non-VDU work
In the 18 reports that compared VDU or keyboard operators to non-users, a
majority found clear and consistent increases in musculoskeletal disorder
prevalences associated with keyboard use. The results are shown in Table 1 and
commented on below.
Bernard and colleagues studied employees of a large newspaper and found that
keying for at least 60% of the work day conveyed a marked increase in risk of
hand/wrist disorders in both the cross-sectional and the prospective phases of their
investigation, compared to individuals keying less than 2 hrs/day (21, 22, Table
1).
Bozi Ferraz and colleagues (26) compared keyboard operators with "traditional
office workers" not using computer keyboards. The former were implied to work
full-time at the keyboard, although this was not stated explicitly. The operators
had much higher prevalences of upper extremity symptoms and diagnoses than
the non-keyboard workers (Table 1). However, a higher percentage of the
operators was female, so the results may have been confounded by gender.
Subjects with and without disorders were also compared on the average number
of keystrokes per hour during the previous month (obtained from company
records) and no significant difference was found. Unfortunately, the authors did
not state whether this comparison was made only within the keyboard operators,
or whether the non-keyboard workers were included as a referent group for this
analysis. Camerino and colleagues (32) found a significantly elevated prevalence
of cervical spine disorders among female VDU operators aged 26 to 35 years,
compared with female workers in the same age range without occupational
exposure to prolonged fixed postures (Table 1).
27
Table 1. Effect of visual display unit use (Yes/No) on frequency of upper extremity musculoskeletal disorders: Relative risk with 95% confidence
interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group, Health
Outcome
Findings
Bergqvist, 1992
(18)
Nk, Sh sx:
Cross-sectional (1981)
Cross-sectional (1987)
Follow-up (1981-87)
OR=1.0 (0.7-1.4)
OR=0.8 (0.5-1.4)
RR=1.0 (0.6-1.6)
Arm sx:
Follow-up
(1981-87)
RR=
1.3 (0.6-2.5)
Hand/wrist sx:
Cross-sectional (1981)
Cross-sectional (1987)
Follow-up (1981-87)
- All VDU use
- New VDU use during follow-up
OR= 1.0 (0.6-1.7)
OR= 1.3 (0.6-2.8)
RR= 2.8 (0.8-10.1)
RR= 4.0 (1.1-14.9)
Bergqvist, 1995a
(19)
Any symptoms
Intense symptoms
TNS diagnosis
Cervical diagnosis
Shoulder diagnosis
OR= 1.4 (0.8-2.4)
OR= 0.5 (0.2-1.8)
OR= 1.0 (0.5-1.9)
OR= 1.3 (0.6-2.6)
OR= 0.6 (0.3-1.5)
Any symptoms
Any diagnosis
OR=1.2 (0.6-2.2)
OR=0.7 (0.3-1.7)
Bernard, 1993,
1994 (21-22)
Phase I
Keying 6-8 vs 0-2 hr/day OR=2.5 (1.6-3.9)
Bozi Ferraz (26) Sx in any UE region,
past 7 days
Sx in any UE region,
past 12 months
Tension neck syndrome
Any UE MSD dx
PR=3.9 (2.5-6.2)
PR=3.7 (2.6-5.3)
PR=5.3 (1.2-23.8)
PR=4.4 (2.5-7.9)
Hand/wrist tenosynovitis dx PR=4.9 (1.9-12.5)
(for notes, see end of the table)
28
Table 1. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Camerino, 1995
(32)
Cervical disorders PR=1.9 (1.0-3.4)
Kemmlert, 1988
(102)
Neck: VDU primary
machine
Shoulder: VDU primary
machine
OR=2.0 (0.8-4.9)
OR=2.1 (0.9-5.2)
Knave, 1985 (104) VDU use (>5 hr/day) vs
none: Neck, upper arms,
and shoulder symptom
intensity
Higher for neck and
upper arm (left side)
and both shoulders
(p<0.05)
VDU use (>5
hr/day) vs none:
Forearm and elbow
symptom intensity
Higher for both
forearms (p<0.05)
but neither elbow
VDU use (>5 hr/day) vs
none: Hand symptom
intensity
Higher for both
hands (p<0.05)
Krapac, 1994 (105) Cervical syndrome
Cervicobrachial syndrome
PR=2.3 (0.8-6.4)
PR=1.0 (0.3-3.0)
Marcus, 1996 (125) Nk/sh sx:
Current use vs. never use
Former use vs. never use
PR=2.3 (1.4-3.7)
PR=2.5 (1.5-4.2)
Arm/hand
(see to the right)
Arm/hand:
Current use vs. never use
Former use vs. never use
PR=1.1 (0.7-2.0)
PR=1.7 (0.9-3.0)
Murata, 1996 (133) Subclinical CTS, by NCV:
wrist-to finger (wf)
wrist-to-palm
ratio of wf to palm-to-
finger value
number of symptoms
p=0.03
p=0.02
p=0.003
p<0.001
(for notes, see end of the table)
29
Table 1. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Nishiyama, 1984
(138)
Nk sx:
Sh sx:
p>0.05
p<0.05
Arm sx: p>0.05 Fingers sx: p>0.05 (right)
p<0.05 (left)
Ong, 1981 (142) Arm/shoulder aches (2
VDU groups combined) PR=1.9 (1.1-3.5)
Arm/shoulder
(see to the left)
Onishi, 1982 (144) VDU vs other office
machines
Highest median
trapezius EMG
levels
VDU vs other office
machines
Highest median
forearm extensor
EMG levels
Starr, 1982 (189) Nk: VDU vs paper task
Sh: VDU vs paper task
OR=2.0 (1.1-3.6)
OR=1.6 (0.9-2.8)
Upper arm: VDU vs paper
Elbow: VDU vs paper
OR=1.1 (0.5-2.5)
OR=0.8 (0.3-2.0)
VDU vs paper task OR=0.7 (0.3-1.6)
Starr, 1984 (187) Nk: VDU vs paper task
Sh: VDU vs paper task
OR=1.0 (0.5-1.7)
OR=1.3 (0.7-2.2)
Upper arm: VDU vs paper
Elbow: VDU vs paper
OR=1.0 (0.4-2.5)
OR=1.3 (0.5-3.4)
VDU vs paper task OR=1.2 (0.5-3.0)
Stellman, 1987
(190)
All-day VDU use OR=1.7 (1.1-2.6) All-day VDU use OR=2.0 (1.2-3.6) All-day VDU use OR=2.8 (1.6-4.9)
* All comparisons are for current VDU use (some) versus no VDU use, unless otherwise specified.
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
30
Hocking (83) presented selected data from passive surveillance of "repetition
strain injury" (RSI) at the Australian national telecommunications agency from
1981 through mid-1987. RSI reports were included from multiple data sources,
with no case criteria specified; in fact, the author stated that "the clinical details of
the cases had not been reviewed." Furthermore, individual subjects could have
contributed multiple reports, and exact denominators for each job group in each
year do not appear to have been available. The highest incidence rate was found
for telephonists and the lowest for telegraphists, but insufficient exposure
information is provided for the three job categories to interpret these differences
in terms of ergonomic factors. No unexposed group or reference rates are
provided. The author states that the incidence rates showed an inverse relationship
with keystrokes per hour and no association with length of employment, but no
data are shown and the sources of these data are not specified. The rates peaked in
1984-85 and then declined; this may have been in response to ergonomic
measures undertaken in the company before and during that period, although
inadequate information is available to conclude this with certainty. (These results
are not shown in Table 1).
Kemmlert and Kilbom (102) investigated those with primary use of VDUs vs.
other office workers and found elevated but non-significant excesses of neck and
shoulder discomforts (see Table 1). Similar increased odds ratios were also found
for non-VDU users who (primarily) used typewriters. This study did adjust for
other factors, though.
In a large cross-sectional study in Sweden, Knave et al (104) used symptom
severity as the endpoint, so prevalences and odds ratios could not be calculated
from published data; symptoms of the neck, shoulder, and upper arms were scored
as more intense among VDU users (more than five hours per day), although only
the differences at the shoulder reached the statistical significance of p=0.05 (Table
1). In other results, not shown in the paper, male subjects with musculoskeletal
disorders of any body region worked more hours per day at the VDU than their
counterparts without disorders (p<0.05). The authors did not use multivariate
analyses to examine multiple exposure dimensions simultaneously, and they
reported large differences in both symptom severity and total hours worked per
day among different industries, so these data may have been confounded by total
hours or by task characteristics.
The same population was followed up by Bergqvist and colleagues over a
seven-year period as a closed cohort with a series of cross-sectional surveys; of
the several articles published to date, two have examined the dichotomous
variable of VDU use (18, 19, Table 1). In both the cross-sectional and the follow-
up studies, VDU use (all types combined) was not associated with neck and
shoulder morbidity after controlling for covariates; for hand and wrist disorders
there were some elevated odds ratios but these were not statistically significant at
p=0.05 except among those subjects who began using VDUs during the follow-up
period. However, several categories of more intensive use, or use in adverse
conditions, did appear to increase the risk (19); these are described under the
31
various specific headings below. There was also a suggestion of a self-selection
process (Óhealthy worker effectÓ), in that among the subjects who dropped out of
the cohort, the VDU users were more than twice as likely to have had hand/wrist
disorders at baseline. Upon contact, however, few among this dropout confirmed
that muscle problems experienced during VDU work had contributed to their
leaving the work place.
Krapac and colleagues (105) also compared VDU users and non-users in a
small cross-sectional study of office workers. Pain and fatigue in the
musculoskeletal system (neck, upper extremity and back combined) during work
were about 50% more frequent in the operators, and cervical syndrome was
diagnosed in more than twice as many VDU users as non-users (Table 1). The
study suffered from potential confounding, as a higher proportion of the VDU
users were women; they were also slightly younger and had slightly less seniority
than the non-VDU workers.
Marcus and Gerr (125) also reported a cross-sectional study, comparing hours
per week and duration in years of VDU use within a large group of young women
office workers. Both neck/shoulder symptoms and hand/arm symptoms were
more prevalent among current users than workers who had never used a VDU.
Perhaps more importantly, the prevalence ratios were even higher for former
VDU users compared with never users. This provides further evidence that
workers who develop VDU-related symptoms are disproportionately likely to
seek jobs without VDU exposure, when able to do so, and thus that cross-
sectional studies likely systematically underestimate the risk associated with VDU
work. However, the subject participation level was only 70%, suggesting possible
selection bias, although participants and non-participants did not differ in their
distributions of age or job title.
Murata et al. (133) compared the sensory nerve conduction velocities (NCV) in
the right median nerve between a group of VDU operators and a referent group of
students with only occasional work processing activity. Despite the small study
size, the VDU operators were found to have markedly lower NCV values across
the carpal tunnel than the comparison group (Table 1). The operators also had
many more self-reported CTS and other upper extremity symptoms; the authors
stated that there was no correlation between symptoms and NCV, although that
analysis may have been diluted by inclusion of some symptoms not usually
considered typical of CTS, such as shoulder stiffness.
In a Japanese study in the newspaper industry (138), prevalences of dullness or
pain were significantly higher among VDU users than other keyboard operators
not using a VDU in the shoulders and (left) fingers. For the neck and left arm,
these differences were not significant, and for right arm, no differences were
found. The authors made the comment that the left hand was used for keying,
while the right hand was used for operating the cursor keys or the joy stick. It
should be noted that the operators used Kanji characters, not alphanumeric
characters, which involves more complex motion patterns in text entry (see e.g.
(131)). The two groups were of similar ages, but no adjustments for gender or
32
other covariates was performed, and the response rate was not stated, limiting the
conclusions that could be drawn from this study.
Two groups of VDU users investigated by Ong (142) had (combined) about
twice the prevalence of aches in the arm and shoulder than the non-VDU users. In
another study of female office workers operating a variety of office machines
(144), shoulder stiffness and fatigue of the hands, arms, and shoulders increased
across the shift, while staying higher for office machine operators; shoulder
compression tenderness thresholds decreased across the day and the work week.
Sustained contractions of the forearm and shoulder muscles were documented
among operators of all six machines studied; the median forces (as a percentage
of maximum strength capacity) were highest for VDU operators, about four times
the level recommended to prevent muscle fatigue and its sequelae (144, see Table
1).
In a pair of studies by Starr et al. with comparable methods, symptom
prevalences were compared between directory assistance operators who used
VDUs and those who used paper records to perform similar tasks (187, 189, Table
1). In one of these studies (189, but not in the other) neck and shoulder conditions
were about twice as frequent among VDU operators; no other differences were
observed between the two types of work. Several weaknesses limit the utility of
these results. The physical demands of the non-VDU jobs were not described, so
it could not be verified that the non-VDU work represented an appropriate
ÓcontrolÓ (i.e., low exposure) group. In the first of the two surveys there was
differential participation between the VDU users and non-users (76% and 95%,
respectively); it was also not clear whether subjects were assured of the
confidentiality of their questionnaire responses. Lastly, there was inadequate
adjustment for potential confounding, even though in the second paper the VDU
users were older and age was negatively associated with symptoms.
Smith and colleagues (180) studied 283 employees of a large newspaper. Pain
and stiffness in the arms, hands and legs combined was stated to be more frequent
in VDU users than non-users (no data presented) but was negatively associated
with length of employment. No results were shown for upper extremity
symptoms alone. This study was flawed by low participation (48%) of the target
population; data were available for about one-half of the non-participants, who
had less education and less VDU experience than the participants (results not
shown in Table 1).
Finally, a study by Stellman et al. examined the effects of three types of VDU
work (keypunch, data processing, and word processing) among women employed
in the public sector (190). The risk of neck and shoulder, arm, wrist and hand
disorders were about twice as high in full-time VDU users as in clerical workers
using only paper records, with neither VDUs nor typewriters (Table 1). The
subjects were all women of similar ages, although the authors did not adjust for
other covariates.
Although not strictly a study of VDU operators, an older study of accounting-
machine operators (87, 122, 123) deserves some interest. These operators were
33
compared to non-clerical workers (retail saleswomen) as the basis for estimating
the symptom frequencies that would be expected in workers with neither clerical
nor other significant ergonomic stresses to the upper extremity. All subjects were
women of comparable ages. The right hand and arm exclusively were used to
operate the keyboard, and the operators had higher prevalences of arm and hand
symptoms on the right than on the left (non-operating) side. In addition, pain in
the arms and hands was more common among the operators than the saleswomen,
especially on the right side (relative risks of 3.4 and 2.0, respectively, Table 1).
Thus, accounting machine operation was found to present a risk of arm and hand
pain which was significantly elevated relative to other, non-clerical work that
women of the same age might perform and which was localised to the limb
performing the keyboard work.
In summary, the overall comparisons between people doing VDU work and
people in other kinds of work produced numerous indications of the former group
being at higher risk of neck, shoulder, arm, wrist and hand musculoskeletal
problems. In general, the highest odds ratios and relative risks were found for
hand and wrist disorders. To the extent that these studies are not all consistent,
there are several possible reasons that might explain the discrepancies. One is
that, as discussed previously, "VDU work" in one study may represent full-time,
intensive keying with high psychological demands at a poorly designed
workstation and in another study may refer to jobs with frequent rest breaks, fully
adjustable workstations and a good psychosocial environment. Workers in the
first situation would likely have much higher risks than the workers in the second;
this may explain why, to some extent, high increases are found in the older studies
(e.g., Ong et al. (142), Onishi et al. (144), Stellman et al. (190)) and those
carried out in developing countries (e.g., Bosi Ferraz et al. (26) or Krapac and
Sakic (105)) but not in some of the recent studies in countries with high attention
to workplace ergonomics (e.g., Bergqvist et al. (18, 19)). A related consideration
is the choice of the reference group. Some authors were able to define and recruit
workers not exposed to upper extremity stressors (e.g., Camerino et al. (32)),
while others utilised workers in jobs that may have involved a great deal of
repetitive hand motion (e.g., Starr et al. (189) and Starr (187)).
Third is the issue of the "healthy worker effect," i.e., negative selection or self-
selection out of the job by people who develop work-related health problems. It
has been shown previously that workers who develop musculoskeletal disorders
are disproportionately likely to transfer out of jobs that are ergonomically
stressful (135, 145, 179). The effect of this transferring or leaving employment is
to produce an underestimate in the risk observed from the study data (152). This
type of selection effect could here explain the failure to find associations with
VDU use or duration of employment in some of the studies that examined those
variables. Indirect evidence for this effect has already been noted in the study by
Bergqvist et al. (18); furthermore, in that study, the odds ratio for hand/wrist
disorders in new VDU users during the follow-up period was somewhat higher
than that estimated for continuing users. Marcus and Gerr (125) showed that
34
former VDU users had a higher relative risk for upper extremity disorders than
current users (both groups compared with never users). Indirect evidence for this
effect was also found in Starr et al. (187), where age was associated with fewer
symptoms and VDU operators were somewhat older than the non-VDU clerks.
Full-time vs. part-time VDU operators
Studies of people performing VDU work all or most of the time, vs. those who
spent less time at the VDU, are reviewed here and summarised in Table 2.
The Swedish investigation described above included several analyses of the
effect of hours of VDU use. In the follow-up of the cohort (18), VDU users were
subdivided into those with up to and more than 30 hours per week of operation;
the longer duration demonstrated an increase in the risk of arm and hand/wrist
disorders (both 95% confidence intervals included 1.0), but not neck/shoulder
disorders (Table 2). The authors failed to find associations between overtime on
short notice and muscle problems, but frequent overtime was associated in the
final multivariate model with arm/hand discomforts (20).
Grieco et al. (71) studied a very large population of VDU operators at the
Italian Telecommunications Company. They presented prevalences of "frequent"
neck and upper limb disorders but did not specify the case definitions or any data
collection methods. Age-standardised morbidity ratios (SMR) were computed for
male and female VDU operators separately, using the entire work-force at the
same company as the reference group. Overall, both neck and upper extremity
disorders were strongly associated with VDU use for at least 4 hours per day
(Table 2).
In another large study of keyboard users in a mix of industries, Rossignol et al.
(163) reported the prevalences of several upper extremity endpoints by four strata
of hours of daily operation. Neck, shoulder, and arm disorders were all at least
twice as frequent among subjects with at least 4 hours exposure per day; the odds
ratios for hand pain and hand paresthesia were lower and not statistically
significant, except among banking, communications, and hospital employees
(Table 2).
In the only other cohort extensively studied to date, by Nathan et al. (135, 137),
nerve conduction velocity (NCV) measurements of the median nerve were
obtained as a clinical test for evidence of carpal tunnel syndrome (CTS). In the
baseline survey, an odds ratio point estimate of 2.75 for bilateral median nerve
slowing was calculated for more than 4 versus less than 4 hours of keying per day
among clerical workers (Óoccupational hand useÓ Classes I and II, (137), Table 2).
In addition, a test of linear trend in the odds of slowing across all classes was
highly statistically significant (p = 0.001). However, the data were not stratified
by gender, so potential confounding could not be ruled out. (This study was
reported as negative by the authors, who made some arguably inappropriate
choices in statistical analysis methods; several other reviewers (e.g., Stock (193)
and Hagberg et al. (75) have concurred that it contains positive findings.) NCV
findings were also reported after five years of follow-up (135). The study design
35
Table 2. Effect of full-time versus part-time visual display unit work on frequency of upper extremity musculoskeletal disorders. Relative risk with
95% confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study Group,
Health Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Bergqvist,
1992 (18)
Follow-up (1981-87)
VDU >30 vs < 30 hr/wk
(among users) RR=0.6 (0.4-1.0)
Follow-up (1981-87)
VDU >30 vs < 30
hr/wk (among users) RR=1.4 (0.8-2.7)
Follow-up (1981-87)
VDU >30 vs < 30
hr/wk (among users)
New VDU use >30 vs
< 30 hr/wk (among
users)
RR=
1.3 (0.6-2.9)
RR=
1.6 (0.7-3.6)
Bergqvist,
1995b (20)
Arm/hand
(see to the right)
Any arm/hand sx:
Extensive overtime
OR=
2.2 (1.2-4.4)
Fahrbach,
1990 (46)
Nk: VDU use >4 vs
4 hr/day)
Sh: VDU use >4 vs
4 hr/day)
OR= 1.3 (0.5-3.8)
OR=10.3 (2.4-43.3)
Grieco, 1989
(71)
Nk: 4-6 hr/day (men)
Nk: >6 hr/day (men)
Nk: 4-6 hr/day (women)
Nk: >6 hr/day (women)
SMR=152 (p<0.05) *
SMR=165 (p<0.05) *
SMR=108 (p>0.05) *
SMR=125 (p<0.05) *
UE: 4-6 hr/day (men)
UE: >6 hr/day (men)
UE: 4-6 hr/day (women)
UE: >6 hr/day (women)
SMR=144 (p<0.05) *
SMR=201 (p<0.05) *
SMR=102 (p<0.05) *
SMR=125 (p<0.05) *
UE (see to the left)
Hales, 1992,
1994 (78-79)
Overtime in past year E-R (neg.)
(for notes, see end of the table)
36
Table 2. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Heyer, 1990 (81) Nk: VDU use > 4 vs < 2
hr/day
Nk: VDU use >6 vs 4-6
hr/day
PR=1.4 (0.9-2.2)
PR=1.1 (0.9-1.4)
VDU use > 4 v
< 2 hr/day
VDU use >6 v
4-6 hr/day
PR=2.0 (0.8-5.5)
PR=0.9 (0.6-1.3)
VDU use > 4 vs < 2 hr/day
VDU use >6 vs 4-6 hr/day
PR=2.3 (1.1-4.4)
PR=1.0 (0.7-1.3)
Kamwendo, 1991a
(94)
Nk: Work with office
machines 5 vs <5 hr/day
Sh: Work with office
machines 5 vs <5 hr/day
OR=1.7 (1.0-2.7)
OR=1.9 (1.2-3.0)
Karlqvist, 1996
(101)
Mouse use 5.6 vs <5.6
hr/day:
- Neck
- Left scapular
- Right scapular
- Left shoulder
- Right shoulder
OR=1.1 (0.8- 1.7)
OR=1.2 (0.7- 2.0)
OR=1.1 (0.7- 1.7)
OR=4.0 (1.6-10.1)
OR=3.9 (1.8- 8.1)
Mouse use
5.6 vs <5.6
hr/day:
- Left elbow
- Right elbow
OR=4.3 (1.4-13)
OR=2.0 (1.0- 4.1)
Mouse use 5.6 vs <5.6
hr/day:
- Left wrist
- Right wrist
- Left hand/fingers
- Right hand/fingers
OR=3.4 (1.1-11)
OR=2.0 (1.0- 4.3)
OR=2.6 (1.0- 6.8)
OR=3.1 (1.5- 6.6)
Nathan, 1988 (137) Median nerve function:
Keyboard operator (>4
hr/day) vs. clerical (<4
hr/day) OR=2.8 (0.8-9.6)
(for notes, see end of the table)
37
Table 2. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Oxenburgh, 1987
(146)
VDU use >4 hr vs <3 hr/day OR=7.9 (2.6-32)
Rossignol, 1987
(163)
Nk: VDU use ( 4 vs 0-3
hr/day)
Sh: VDU use ( 4 vs 0-3
hr/day)
PR=2.8 (1.8-3.9)
PR=3.0 (1.9-4.4)
VDU use ( 4 vs
0-3 hr/day)
PR=
2.0 (1.2-3.7)
Numbness/tingling: VDU use
4 vs 0-3 hr/day
Hand/wrist pain: VDU use
4 vs 0-3 hr/day
Hand/wrist pain: VDU use
4 vs 0-3 hr/day in banking,
communications, and hospitals
PR=1.2 (0.3-5.7)
PR=1.5 (0.8-2.6)
PR=2.4 (1.0-6.0)
* Reference group for all SMRs shown: the entire telecommunications company workforce.
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
38
placed more emphasis on exposures in industrial jobs and the association between
Classes II and I alone could not be determined from the data provided; however,
occupational hand use at baseline, across all five classes, was a statistically
significant predictor of NCV slowing, along with age and hand dominance.
Oxenburgh (146) compared 46 cases of Órepetition strain injuryÓ with a sample
of referents without injury (all women). The two groups differed little with
respect to workstation and physical environment, but the cases worked many
more hours at a keyboard (87% using VDUs). No cases used keyboards fewer
than 3 hours per day, so the odds ratio could only be calculated for the effect of at
least 4 hours of operation (Table 2).
Karlqvist et al. (101) carried out the only epidemiologic study to date to focus
specifically on computer mouse users. They selected the CAD operators (n=542)
from a total study group of 652 telecommunications laboratory workers and
examined the associations of neck and upper extremity symptoms with hours per
day of mouse use and mouse location (see next section). After adjusting for age
and gender (see Tables 2), the odds ratios with prolonged mouse use were 2.2 or
higher for shoulder/upper arm, elbow, wrist, and hand/finger symptoms.
Somewhat surprisingly, even though most of the subjects operated the mouse with
the right arm, symptoms in the left shoulder and arm were as likely to be
associated with mouse use as the right side. Possible explanations for this may
include the effect of keyboard work with the non-mouse hand, operators having
changed sides of mouse use after developing symptoms, and sympathetic muscle
coactivation.
Odds ratios could also be computed from three further studies, and in general
the increase in risk of upper extremity morbidity for 4 or more versus less than 4
hours per day was approximately two (46, 81, 94, Table 2). However, two of
these studies suffered from potential selection bias (46, 81) and none of them
controlled for potential confounding by gender and/or the simultaneous effects of
multiple exposures (46, 81, 94).
Three studies were judged to have inadequate power to examine the
relationship because there was so little variability in exposure (i.e., all of the
subjects used VDUs 6 or more hours per day) (78, 84, 170). Nevertheless,
Hoekstra et al. found an association among telephone service workers with more
than 8 hours per day (the question was worded with respect to telephone use, but
subjects used both the telephone headset and the keyboard continuously during
work, so the effects of these two items could not be separated (84).
Hours of VDU operation per day
A number of studies compared the risk of MSDs at multiple levels of exposure,
characterised either as hours of keyboard operation per day or per week or as
percentage of the day spent keying (Table 3).
In the follow-up study of Bergqvist et al. (18), increases were found in
cumulative incidence of hand/wrist disorders (p<0.05) and arm/shoulder disorders
(p>0.05) per hour of weekly VDU use.
39
Table 3. Effect of number of hours of VDU work per day or week on frequency of upper extremity musculoskeletal disorders: Relative risk with 95%
confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health Outcome
Findings
Bergqvist, 1992
(18)
Follow-up (1981-87)
Change in cumulative
incidence, per 10 hr/week
VDU use
-4.6%
(-10.5 - 1.2%)
Follow-up (1981-87)
Change in cumulative
incidence, per 10
hr/week VDU use
+2.6%
(-2.0 - 6.1%)
Follow-up (1981-87)
Change in cumulative
incidence, per 10
hr/week VDU use
+3.2%
(0.2 - 6.1%)
Bernard, 1993,
1994 (21-22)
Keying 6-8 vs 0-2 h/day p>0.05 in multi-
variate analysis
Phase I
2-4 h/day
4-6 h/day
6-8 h/d
8 h/day
(cf with <2 h/day)
Phase II
Keying 60-79% of day
Keying 80-100% of day
One-year increase in
hr/day keying
ORs:
1.0 (0.6-1.8)
1.3 (0.8-2.2)
2.1 (1.3-3.6)
3.3 (1.2-8.9)
ORs:
7.6 (1.8-32)
2.3 (1.0-7.8)
9.1 (7.1-11.6)
Burt, 1990 (30) Nk: Keying 20-39% of day
Nk: Keying 40-59% of day
Nk: Keying 60-79% of day
Nk: Keying 80-100% of day
OR=2.0 (1.0-7.7)
OR=2.6 (1.4-5.0)
OR=2.2 (1.0-4.7)
OR=2.8 (1.4-5.4)
Keying 20-39% of day
Keying 40-59% of day
Keying 60-79% of day
Keying 80-100% of day
OR=1.2 (0.6-2.5)
OR=1.7 (0.8-3.5)
OR=1.9 (0.9-4.3)
OR=2.8 (1.4-5.7)
Percent of time keying p<0.01 in
intial, p>0.05
in multivariate
analysis
DeMatteo, 1993
(41)
MS sx, all UE regions:
VDU hr/day
E-R, p<0.03
(for notes, see end of the table)
40
Table 3. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Evans, 1987 (45) Nk/sh pain Òoften:Ó
VDU hr/day
E-R, p<0.0001
Faucett, 1994 (47) Nk/sh disorders: Per
hr/day VDU use OR=1.5 (1.1-2.1)
Disorders: Per hr/day
VDU use OR=1.5 (1.1-2.1)
Hales, 1992, 1994
(78-79)
dx: Per hr/day VDU use E-R (neg.)
Jeyaratnam, 1989
(89)
Nk sx: VDU hr/day
Sh sx: VDU hr/day
E-R, p<0.05
E-R, p<0.05
VDU hr/day no E-R VDU hr/day no E-R
Kamwendo, 1991b
(95)
Nk/sh fatigue and pain:
Typing hr/day
Positive corre-
lations within
shift (p<0.05) for
33% of subjects
Knave, 1985
(104)
MS sx, all body
regions
p<0.05 MS sx, all body
regions (see to the
left)
MS sx, all body regions
(see to the left)
Marcus, 1996 (125) Nk/sh: hr/week current
VDU use no E-R
Arm/hand
(see also to the right)
Hand/arm: hr/week
current VDU use no E-R
(for notes, see end of the table)
41
Table 3. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Oxenburgh, 1987
(146)
Per hr/day VDU use E-R, p<0.0001
Rossignol, 1987
(163)
Nk: VDU hr/day
Sh: VDU hr/day
E-R, p<0.001
E-R, p<0.001
Hours VDU use/day E-R, p=0.01 Pain
Numbness
E-R, p=0.14
no E-R. p=0.50
Sauter, 1983, 1984
(168-169)
Neck/shoulder/back:
VDU hr/day r=-0.15 (p<0.05)
Arm/wrist
(see also to the right)
Arm/wrist/hand:
VDU hr/day
p>0.05 in multi-
variate analysis
SHARP, 1993
(174)
Sh sx: per hr/day
keying OR=2.5 (p<0.10)
Sx: per hr/day of keying
Sx/PE: per hr/day of keying
OR=3.1 (1.2-8.4)
OR=1.8 (0.9-3.8)
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
42
DeMatteo et al. (41) compared full-time data entry clerks in highly repetitive
jobs with clerical workers who used the VDU fewer hours per day and had more
varied tasks. A correlation was found between the number of symptoms and hours
at the VDU, although this may have been confounded by the primary contrast
between job titles. No data were presented on gender, age, or other covariates, so
the potential for uncontrolled confounding could not be ruled out.
Evans (45) analysed 3,819 questionnaires received from a mail-in survey of
VDU users. The occurrence "often" of neck and shoulder pain increased three-
fold among respondents working more than 8 hours per day at the VDU compared
to those working 1-2 hours per day. However, this association is difficult to
interpret in the light of potential selection and information bias and lack of
statistical adjustment for covariates.
Jeyaratnam and colleagues (89) carried out a large cross-sectional study in
which they compared female VDU operators on several occupational and non-
occupational factors. An increasing prevalence of neck and of shoulder symptoms
was found with hours per day of operation, although this was not the case for
elbow or wrist/hand symptoms. Large differences were found among ethnic
groups and age strata, but no statistical adjustment was carried out for these
variables (Table 3).
In a nested case-referent study among 156 office workers, Oxenburgh (146)
found a distinct exposure-response relationship between VDU use and hand/wrist
problems. An exact test comparing at least 2 with less than 2 hours was also
highly significant (p<0.01). In contrast, Marcus and Gerr ((125)) did not find
significant exposure-responses with arm/hand problems, nor with neck and
shoulder problems in a larger, cross-sectional study (Table 3).
In the study by Rossignol et al. already described above (163), tests of linear
trend across all four strata of exposure duration, and pooling all industries, were
highly statistically significant for neck, shoulder, and arm pain (Table 3).
Examining the possible effect of modest amount of VDU work, the authors found
- in general - little difference in risk between 0 hours and 0.5 to 3 hours per day,
except for neck and shoulder symptoms in computer and data processing services,
public utilities, and state agencies.
Rubino et al. (165) carried out a very large study, the methods of which were
not described at all in the paper. Musculoskeletal symptoms of the neck, upper
limbs and lower limbs were combined into a single factor, which increased in
intensity with increasing number of hours per day at the VDU (no statistical test
reported). However, interpretation of these findings is difficult without more
knowledge of the study methods (data not shown in Table 3).
Sauter (168, 169) recruited a sample of 248 VDU users vs. 85 "traditional"
office workers and found a negative association between hours per day of VDU
use and upper torso musculoskeletal discomforts. A non-significant association
was found between extremity discomforts and VDU use (Table 1). Separate
analyses were not performed, however, for specific body locations such as the
neck or shoulder.
43
Three studies examined VDU use among newspaper employees. Two of these,
originally undertaken as health hazard evaluations by the US National Institute of
Occupational Safety and Health (NIOSH), were large investigations that achieved
sufficient variability in hours of exposure to obtain stable estimates; both also
utilised data on numerous covariates in multivariate analyses. Bernard and
colleagues found that keying for at least 60% of the work day conveyed a marked
increase in risk of hand/ wrist disorders in both the cross-sectional and the
prospective phases of their investigation, with odds ratios ranging from 2.3 to as
high as 9.1 for an increase in daily hours of keying in the previous year (21, 22,
Table 3). The results of Burt et al. (30) demonstrated clear exposure-response
relationships for both neck and elbow/forearm disorders with proportion of the
day keying (Table 3). Similar relationships for shoulder and wrist/hand disorders
were observed in univariate analyses but were no longer statistically significant
when typing speed was included in the multivariate models.
A smaller study of VDU operators in a newspaper editorial department
demonstrated strong dose-response trends with hours per day of VDU use and
disorders of both the hand/arm and the neck/shoulder/upper back, using
multivariate statistical analyses (47, Table 3). Each of the odds ratios exceeded
2.0 for as little as two hours per day of keyboard use. Gender and age were not
included in the regression models in the article, but unpublished data showed no
change in results with inclusion of either covariate (Julia Faucett, Univ. Calif. San
Francisco, personal communication, 1994-95). The participation was only 56%
among recruited individuals, but respondents and non-respondents were similar in
age, job title, and duration of employment at the newspaper. More women than
men responded, but gender was only weakly associated with the outcomes and did
not confound the crude analyses.
In an office of workersÕ compensation claim clerks, a follow-up intervention
study was carried out in which the third cross-sectional survey was also analysed
separately for associations with prevalent cases (174). Within this small group
(n=34), the number of hours keying per day was about one hour higher for each of
the shoulder, elbow, and wrist/hand, using both symptom-based case definitions
and case definitions based on physical examination (PE) (although only the
comparison for shoulder symptoms was significant at the level of p<0.05). From
multivariate analyses, the risk of shoulder symptoms, hand/wrist symptoms and
hand/wrist PE cases increased from two to three times per hour of daily keying
(Table 3).
In multiple measures of symptoms across the work shift, neck and shoulder
fatigue and pain increased as the number of hours worked (95). In contrast, Hales
et al. (78) found that hours per day were weakly but negatively associated with
severity of hand/wrist pain (78).
In summary, when comparisons were made between individuals performing
more vs. fewer hours per day, the great majority of studies showed a substantial
increase in neck, shoulder, arm and hand problems among those working for
44
longer hours on a VDU. The exceptions were primarily the three studies with
inadequate power, and those that failed to control for potential confounding.
Duration (years) of exposure to VDU or keyboard operation
Years of employment in a VDU-exposed job was used in several of the studies
reviewed as an estimator of cumulative exposure to VDU or keyboard work. In
one study, calculated person-years of VDU work was used. Results are shown in
Table 4, and discussed further below.
Duration of employment in VDU work type jobs was found to be associated
with an increased risk of neck and/or shoulder symptoms in several studies (22,
30, 32, 41, 71, 94, 125, 174) and of arm/elbow or hand/wrist symptoms in some
(71, 125, Table 4). Some additional results merit further discussion, though.
Camerino and colleagues (32) combined female VDU operators aged 26 to 35
years with a group of female office workers not using VDUs, whom they found to
have a similar index of postural fixity and a prevalence of cervical disorders
intermediate between the VDU operators and the reference group. In the same age
group, the prevalence of cervical disorders was significantly associated with years
of occupational exposure to fixed postures (Table 4). This indicates that some, but
perhaps (?) not all, of the elevated risk for the VDU operators can be attributed to
the static postural demands associated with prolonged sitting at a desk. In the
study by Jeyaratnam and colleagues (89) already described, no or even a negative
association was found with years of VDU work.
Knave et al. reported a small difference (not statistically significant) in length
of employment between subjects with any musculoskeletal disorder versus those
with none (104). However, in a later examination of the same cohort, with
adjustments for some confounders, Bergqvist et al. (19) failed to find increased
prevalences among individuals with more than 5 person-years of accumulated
VDU work compared to those with less. As already pointed out above, in this
study population, individuals who began VDU work during the follow-up period
(7 years) were at higher risks of hand and wrist problems. Furthermore, dropouts
during this period were also more likely to report hand/wrist problems prior to
dropout. Both of these factors may contribute to this lack of association with
person-years.
A very small study by Pickett et al. of intensive data entry work showed no
association with length of employment for any upper extremity symptoms, but
there was no adjustment for age or any other covariates, including poor
workstation ergonomics (150).
In summary, the comparisons of individuals with longer and shorter duration of
VDU work have not produced consistent results. However, all of the methodo-
logical concerns discussed above with respect to studying the effect of VDU work
per se are equally relevant here. It should be noted that the analyses that failed to
find such significant differences were either small (Picket et al), or belonged to
the Swedish study (Knave/Bergqvist) that overall gave the weakest support for an
association between VDU work per se and muscle (neck) prob lems. Furthermore,
45
Table 4. Effect of years of employment or person-years in VDU work on frequency of upper extremity musculoskeletal disorders. Relative risk with
95% confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Bergqvist, 1995a
(19)
Nk/sh sx: 5 p.y
>5 p.y
Intense nk/sh sx: 5 p.y
>5 p.y
TNS dx: 5 p.y
>5 p.y
Cervical disorders 5 p.y
>5 p.y
Any shoulder dx: 5 p.y
>5 p.y
OR=1.5 (0.8-3.1)
OR=1.3 (0.7-2.5)
OR=0.9 (0.3-3.1)
OR=0.8 (0.3-2.5)
OR=1.2 (0.5-2.8)
OR=1.0 (0.4-2.1)
OR=1.3 (0.6-3.0)
OR=1.1 (0.5-2.3)
OR=0.6 (0.2-1.6)
OR=0.7 (0.3-1.6)
Arm/hand
(see to the right)
Arm/hand sx:
5 p.y
>5 p.y
Arm/hand dx:
5 p.y
>5 p.y
OR=1.3 (0.6-2.9)
OR=1.4 (0.7-2.8)
OR=0.8 (0.3-2.3)
OR=0.6 (0.2-1.6)
Bernard, 1993,
1994 (21-22)
Sh: Per year of
employment
Nk: Length of employment
OR=1.4 (1.2-1.8)
p>0.05 in multi-variate analysis
Length of
employment
p>0.05 in
multivariate
analysis
Burt, 1990 (30) Neck: As reporter:
1 year
>1 - <5 years
5- <10 years
10 years
Neck: Other jobs:
1 year
>1 - <5 years
5- <10 years
10 years
OR=4.6 (1.3-15.7)
OR=6.5 (2.6-15.9)
OR=10.7 (3.6-32.5)
OR=7.6 (2.4-20.4)
OR=1.0 (reference)
OR=2.8 (1.3-6.4)
OR=4.4 (1.8-10.7)
OR=7.4 (3.1-17.7)
Length of
employment
p>0.05 in multi-
variate analysis
Length of
employment
p>0.05 in
multivariate
analysis
(for notes, see end of the table)
46
Table 4. (Continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Camerino, 1995
(32)
Nk: Years exposed to
fixed postures (ages
26-35 years) E-R, p=0.001
DeMatteo, 1993
(41)
MS sx, all UE regions:
Years of VDU experience E-R, p<0.03
UE sx
(see to the left)
UE sx
(see to the left)
Grieco, 1989 (71) Nk: > 4 years of VDU
work (men)
Nk: >4 years of VDU
work (women)
SMR=1.30 (p<0.05) *
SMR=1.12 (p<0.05) *
UE: > 4 years of
VDU work (men)
UE: >4 years of
VDU work
(women)
SMR=1.20 (p>0.05) *
SMR=1.14 (p<0.05) *
UE sx
(see to the left)
Kamwendo,
1991a (94)
Nk: Years employed
Sh: Years employed
E-R, p=0.01
E-R, p=0.002
Marcus, 1996
(125)
Nk/sh: VDU use:
<3 years
4-6 years
> 6 years
OR=4.1 (1.5-11.2)
OR=5.6 (2.0-15.7)
OR=4.3 (1.4-13.6)
Arm/hand
(see to the right)
Arm/hand: VDU
use: <3 years
4-6 years
> 6 years
OR=1.9 (0.7-5.2)
OR=1.9 (0.7-5.3)
OR=3.9 (1.2-12.0)
(for notes, see end of the table)
47
Table 4. (Continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Pickett, 1991 (150) Nk: Years employed
Sh: Years employed
p=0.53
p=0.76
Years employed p=0.83 Wrist: Years employed
Hand: Years employed
p=0.85
p=0.22
SHARP, 1993
(174)
Nk sx: Years in current job
Sh sx: Years in current job
p<0.05
p<0.08
* Reference group for all SMRs shown: the entire telecommunications company workforce
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
48
Marcus and Gerr (125) showed that, in the population they studied, former VDU
users had higher rates than never users, providing clear evidence of the difficulty
in studying this issue in cross-sectional studies.
VDU work in general, compared to low-exposed industrial jobs
In order to address, in part, the problem posed by the lack of appropriate
comparison groups, external reference rates were obtained from a study by
Silverstein and colleagues (176, 177) that quantified the prevalence of hand/wrist
disorders in US industrial employees according to their exposure to forceful
manual exertions and to repetitive manual work. The disorder rates in the jobs that
were scored both low in force (ÓLoFÓ in Table 5) and low in repetition (ÓLoRÓ)
were 9% based on symptoms alone and 2% for symptoms plus physical
examination findings (Table 5). Workers in highly repetitive manual tasks that
required low forces (ÓLoF.HiRÓ) had rates that were more than twice as high
(22% and 7%, respectively, for the two case definitions).
Five of the US studies reviewed here used a case definition based on symptoms
alone (22, 30, 84, 174), or on symptoms plus physical findings (78, 174) that was
identical to or slightly more restrictive than that of Silverstein et al. (176, 177).
Two of these described groups of newspaper employees, two were of telephone
service workers, and one a group of clerks in a state agency (Table 5). All of the
groups were of comparable ages except for the last, which was marginally
younger on average.
The relative risks for these five groups of office workers were estimated by
standardising the reference rates from the industrial workers in low exposure jobs
(ÓLoF.LoRÓ) to the gender distributions of each VDU population in turn. The
relative risks for hand/wrist disorders using the symptom-based case definition,
adjusted for gender, ranged from about 2 to 5; using the physical examination
case definition, the relative risks were about 5 to 9 (Table 5). In other words, the
magnitude of these risks was significantly higher than the background population
risk, and even exceeded that which would have been expected if they had been
employed in highly repetitive, low force jobs in an industrial setting.
In conclusion, there appears to be a substantial body of evidence showing that
VDU workers are at increased risk (probably more than doubled) of upper
extremity disorders, relative to the risk that they might expect if employed in low-
exposure office or industrial jobs. In light of the fact that almost all of these
studies are cross-sectional, which suffer from the particular limitation that they
likely underestimate the true increase in risk due to exposure, the number of
studies that show positive associations and have ruled out most or all competing
explanations for their findings is even more impressive.
49
Table 5. Frequency of hand/wrist disorders in full-time VDU operators compared to
industrial workers with low ergonomic exposures (studies with comparable case
definitions): Sex-standardised risk ratios (RR) with 95% confidence intervals (CIs).
Study, author Population Age: Mean Prevalence of symptoms Standardised
year (% female) ±SD
(or range)
Male Female Both relative risk
A. Symptoms on interview
Silverstein, 1985
(176),
Industrial
LoF.LoR (45%)
Industrial
LoF.HiR (70%)
39±10
41±11
3 %
12 %
16%
27%
9%
22 % 1.8 (1.3-2.6)
Bernard, 1993 (22) Newspaper
employees (56%)
39±11
(19-72)
22 % 2.2 (1.9-2.5)
Burt, 1990 (30) Newspaper
employees (55%)
40
(20-72)
23 % 2.2 (1.9-2.6)
Hoekstra, 1994 (84) Teleservice repre-
sentatives (64%)
42
(23-64)
30 % 2.6 (1.8-3.7)
SHARP, 1993 (174) Workers' compen-
sation clerks (67%)
35±9
(20-55)
50 % 4.2 (2.5-6.8)
B. Symptoms plus findings on physical examination
Silverstein, 1986
(177)
Industrial
LoF.LoR (45%)
Industrial
LoF.HiR (70%)
39±10
41±11
0%
2%
3%
9%
2%
7 % 3.1 (1.5-5.6)
Hales, 1992 (78) Telecommunica-
tions employees
(78%)
38±10
(19-68)
12 % 4.7 (3.6-6.0)
SHARP, 1993 (174) Workers' compen-
sation clerks (67%)
35±9
(20-55)
21 % 9.3 (3.8-19.2)
* Standardized by adjusting the overall prevalences in reference (176)
(Table 5.16, page 89) to the proportions of female and male workers in
each of the other populations in turn
"LoF.LoR" = Low force, low repetition; "LoF.HiR" = Low force, high
repetition
We have reached this conclusion even while acknowledging that the
comparison of VDU workers to non-VDU workers is often difficult to interpret
with regard to the range of, and degree of contrast in, ergonomic exposures. In
order to define better which aspects of VDU work are implicated as causal
factors, and in order to understand the degree of risk in VDU work compared with
other types of clerical work with other ergonomic stressors present, the sections
that follow examine specific dimensions of exposure that occur both in VDU and
in other office jobs.
50
Physical ergonomic factors in VDU work
situations
Workstation dimensions and postural stress
In 17 of the reviewed studies, direct observations or self-reports of body posture
or measurements of workstation dimensions were used to evaluate ergonomic
stress on various body parts during keyboard operation. One other author had
intended to examine such associations but found too little variance in workstation
configurations (78, 79).
The Swedish cohort study (19, 20) found that three of four neck/shoulder
endpoints were strongly associated with a high keyboard or VDU placement and
two of four with static work posture (Table 6). Wrist/hand disorders were
consistently associated with low keyboard height, lack of forearm support, and
non-neutral hand position. In a retrospective part of the study (15), new furniture
were - among men - associated with an increase in neck and shoulder discomforts
over the follow-up period. The authors offered no suggestive explanation for this
finding. Furthermore, it could not be determined whether the new furniture was
obtained before or after onset of the problems.
Bernard et al. (22) found, in a nested case-control study with exposure data on
the preceding three years, that at least 60% of the hand/wrist cases had received a
new chair or telephone equipment subsequent to their reporting problems to their
supervisors. About 40% of the cases who had received such new equipment
thought that it had helped to reduce their symptoms, although no comparison
group was available and thus no statistical associations could be estimated. Fewer
than 50% of the cases had received any other equipment, such as an adjustable
desk, footrest, or new keyboard.
DeMatteo et al. (41) found that the number of symptoms per worker was
statistically significantly correlated with self-reported workstation design
problems and indirectly with lack of postural mobility. However, a similar
difficulty in interpreting these findings arises as did with the association with
hours per day. Since the data clerks had inherently more repetitive jobs, with less
variation in arm and hand position and more wrist bending and twisting, it is not
possible to distinguish from the data provided whether some excess risk should be
specifically attributed to poor workstation features, or whether it is all contained
within the more macro-ergonomic aspect of poor work organisation that results in
little variability in activities and postures.
Faucett et al. (47) found that keyboard height and head rotation were correlated
with severity of neck/shoulder symptoms in a subset of 70 newspaper employees,
after adjusting for gender and age. In addition, keyboard height interacted with
work load, decision latitude and social support in explaining symptom severity for
51
both the hand/wrist and neck/shoulder (see also further discussion under Òwork
load and work demandÓ and Ópsychological and social factors and muscle
problemsÓ).
In a retrospective cohort study, Ferreira et al. (54) failed to find associations
between ergonomic improvements in workstation and VDUs and incidence rates
of MSD, but probably these factors could not be evaluated adequately because
they occurred close to the end of the follow-up period.
Green and Briggs (70) surveyed the keyboard users employed by a large
university, in order to determine the availability of adjustable workstations, the
information that workers had received regarding how to adjust their equipment,
and their satisfaction with the equipment. However, the response rate was rather
low (52%) and there was no analysis of potential confounding by age, gender, or
other covariates. Workers with and without symptoms were equally likely to have
adjustable chairs, desks, and monitors, although fewer than half of either group
had adjustable desks. In a small sample on whom anthropometric and workstation
measurements were collected, those with adjustable desks had lower desk heights,
on average, compatible with standard ergonomic recommendations. In the total
population, workers with symptoms were more likely to have received
information from someone without formal ergonomics training. They were also
more likely to have readjusted their workstations, where possible, but they were
less comfortable and less satisfied with the adjustability available. The authors
correctly point out the difficulty in determining whether the association between
symptoms and readjusting the equipment was due to coping with symptoms, less
adequate equipment, or less information about how to make the adjustments
(these results are not shown in Table 6).
Another study found positive univariate associations with postural stress but did
not examine multiple exposures simultaneously. In a group of VDU users from
various industries, non-neutral neck angle was associated with disorders of the
neck, arm, and hand (81). This neck strain was highly correlated with data entry
tasks; it may also have served as a proxy for more distal postural stresses on the
arm and hand, since these were not measured directly (Table 6).
ÓTeleservice representativesÓ of the US Social Security Administration, who all
worked full-time with VDUs and telephone headsets, were much more likely to
have symptoms and PE findings of all upper extremity regions if they were
employed in a facility with non-adjustable workstations and chairs and generally
very poor work layout (84). However, these univariate associations were replaced
for neck and arm/elbow disorders by more specific postural stresses such as non-
optimal chair, desk height, and screen location (Table 6). In the study by HŸnting
and co-workers (88), those few data entry workers (n=7) with more than 20
degrees of ulnar abduction of the right wrist had almost 3 times more pain in the
right arm than those with less abduction.
Among medical secretaries, sitting for at least five hours per day was associated
with both neck and shoulder symptoms (94), although this was likely confounded
by work with office machines while the operator was seated. Shifting from sitting
52
to standing more often per day was also found to be correlated (95), although it
could not be determined whether this was a cause or a consequence of the
neck/shoulder discomfort. Karlqvist and colleagues (101) compared workers on
the basis of whether or not their computer mouse placement was ÕoptimalÕ and
found symptoms in neck, shoulder, elbow, wrist and fingers to be more prevalent
with a non-optimal mouse location. Most of the differences between ÒoptimalÓ
and Ònon-optimalÓ location, especially for the proximal extremity, were
statistically significant.
Kemmlert and Kilbom (102) compared office workers with respect to the
ergonomic features of their workstations, and found no or even negative
associations with neck and shoulder symptoms. However, the physical
dimensions were assessed rather crudely, and exposure misclassification was
possible. More importantly, it appeared from workplace discussions (after the data
had been collected) that workers who had developed symptoms had subsequently
paid more attention to their workstations, so that the observations did not
necessarily reflect the conditions that had been in effect at the time of symptom
onset (personal communication, •. Kilbom, June 1997).
Lim et al. (116) developed two scales for describing physical activity and body
posture in VDU work. They found that working in uncomfortable positions and
using awkward motions was strongly correlated with the intensity and frequency
of upper extremity symptoms in multivariate analyses. In addition, dynamic
(whole body) work activity was strongly negatively correlated with UE
symptoms, meaning that being able to change work postures and walk around the
office was associated with fewer and less severe disorders. In a group of female
office workers operating a variety of office machines, the keyboard heights were
measured and found to be too high for all sub-groups (144).
In a study of public sector data entry operators, lack of training in chair
adjustment was the strongest single risk factor for neck/ shoulder/forearm
disorders (Table 6). Elbow flexion was also a risk factor, in the expected
direction, while forward arm flexion and distance to hard copy were negatively
associated; as the authors pointed out, in a cross-sectional study it could not be
determined whether these work station dimensions were causal or represented the
operatorsÕ attempts to accommodate their discomfort at work (166).
Sauter (168, 169) collected self-reported data on the physical environment from
333 office workers, mostly female, and made direct measurements of workstation
dimensions for a 25% sample of the 248 VDU users. In multivariate analyses,
"upper torso" symptoms were associated with self-reported physical workstation
problems, in general, in addition to using an uncomfortable chair. "Extremity"
symptoms were associated with self-reported workstation problems, and with
observed gaze angle and lack of a keyboard that could be detached and adjusted
away from the central processing unit. However, these latter two correlations
were not adjusted for age or other covariates.
53
Table 6. Effect of VDU workstation dimensions or work postures on frequency of upper extremity musculoskeletal disorders: Relative risk with 95%
confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome*
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group, Health
Outcome
Findings
Bergqvist, 1995a
(19)
Arm/hand
(see to the right)
Any arm/hand sx: Data entry
and low keyboard
Any arm/hand dx: > 20 h/wk,
limited rest breaks, and no arm
support
OR=2.8 (0.9-8.6)
OR=4.6 (1.2-17.9)
Bergqvist, 1995b
(20)
Any dx: Static work posture
Any sx: Static work posture
Any sx: High keyboard
Intense sx: VDU placed high
TNS dx: High keyboard
OR=5.1 (0.6-42)
OR=4.1 (0.9-1.8)
OR=3.1 (1.3-7.2)
OR=7.4 (0.9-60)
OR=4.4 (1.1-17.6)
Arm/hand
(see to the right)
Any sx: Non-neutral hand
position
Any sx: Keyboard too low
Any dx: No forearm support
Any dx: Limited rest breaks and
no forearm support
OR=3.8 (1.0-15.0)
OR=2.0 (0.9-4.5)
OR=2.7 (0.9-8.3)
OR=10.1 (2.4-43)
Bergqvist, 1995c
(15)
Any sx: New furniture (among
men) RR=4.2 (1.6-11.4)
DeMatteo, 1993
(41)
Any UE sx:
Workstation design problems
Lack of postural mobility
E-R, p<0.03
E-R, p<0.03
Faucett, 1994 (47) Nk/sh severity: Head rotation
Nk/sh severity: Keyboard
height
E-R (p<0.01)
E-R (p<0.05)
Numbness severity: Work
posture E-R (p<0.10)
(for notes, see end of the table)
54
Table 6. (continued )
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health
Outcome*
Findings Exposure, Study Group,
Health Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Heyer, 1990 (81) Nk: Non-neutral
neck angle PR=1.5 (1.2-1.9)
Non-neutral neck angle PR=2.5 (1.7-3.7) Non-neutral neck angle PR=1.9 (1.4-2.6)
Hoekstra, 1994 (84) Nk: Chair
discomfort
Sh: Poor overall
ergonomics
Sh: Non-optimal
desk height
Sh: Non-optimal
screen location
OR=3.5 (1.4-8.9)
OR=4.0 (1.2-13.1)
OR=5.1 (1.7-15.5)
OR=3.9 (1.4-11.5)
Non-optimal chair OR=4.0 (1.2-13.1)
HŸnting, 1981 (87) Nk/Sh PE: Neck
flexion >56° in data
entry work
p<0.05 PE: Ulnar abduction
>20° in data entry work
Ulnar abduction >20° in
conversational work
OR=15.2 (1.5-721)
OR=8.8 (0.5-137)
PE: Angle of wrist ulnar
abduction (all operators)
E-R (p-value not
given)
Karlqvist, 1996
(101)
Mouse location
non-optimal:
Neck sx:
Left scapular sx
Right scapular sx
Left shoulder sx
Right shoulder sx
OR=1.5 (1.0- 2.3)
OR=2.3 (1.3- 4.2)
OR=1.6 (1.0- 2.7)
OR=4.4 (1.3-15.0)
OR=2.6 (1.2- 5.9)
Mouse location non-
optimal:
Left elbow sx:
Right elbow sx
OR=4.2 (1.0-18.7)
OR=2.7 (1.1- 6.6)
Mouse location non-
optimal:
Left wrist sx
Right wrist sx
Left hand/fingers sx
Right hand/fingers sx
OR=2.3 (0.6- 8.3)
OR=2.4 (1.0- 5.9)
OR=3.0 (0.9-10.5)
OR=1.9 (0.8- 4.2)
(for notes, see end of the table)
55
Table 6. (continued )
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Kamwendo, 1991a
(94)
Nk: Sitting 5 hrs/day
Sh: Sitting 5 hrs/day
OR=1.5 (0.9-2.6)
OR=1.6 (0.9-2.7)
Kamwendo, 1991b
(95)
Nk/Sh fatigue and pain:
Number of shifts from sitting
to standing
Negative correlations for 72% of
subjects (fatigue) and 61% (pain), but
significant (p<0.05) for only 12%
(fatigue) and 3% (pain) of subjects.
Kemmlert, 1988
(102)
Nk/Sh discomforts: Various
postural elements No association with discomforts
Lim, 1994 (116) UE MSD sx:
Awkward postures
Dynamic work activity
UE CTD index:
Awkward postures
Dynamic work activity
r= 0.41 (p<0.001)r
r=-0.28 (p<0.001)
r= 0.47 (p<0.001)
r=-0.35 (p<0.001)
UE MSD
(see to the left)
UE MSD
(see to the left)
(for notes, see end of the table)
56
Table 6. (continued )
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Ryan, 1988 (166) Nk/sh/forearm:
No training in chair
adjustment
L elbow (more
acute angle)
L forward arm
flexion
Eye copy at closer
distance
OR=5.6 (1.7-18.5)
p=0.006
p=0.03
p=0.01
Forearm
(see to the left)
Sauter, 1983, 1984
(168-169)
Neck/shoulder/back
- Workstation
configuration
problems
- Chair comfort
r= 0.16 (p <0.05)
r=-0.21 (p <0.05)
Arm/wrist/hand
(see to the right)
Arm/wrist/hand discomforts:
- Workstation configuration
problems
- Detached keyboard
- Increased gaze angle
r=0.17 (p<0.05)
r=0.28 (p<0.05)
r=0.39 (p<0.05)
Sauter, 1991 (170) Trunk discomforts:
Head tilt and
viewing angle
Neither predictive of
discomforts.
Keyboard height
Document reach
distance
Shoulder flexion
Wrist ulnar deviation
E-R (p<0.10)
E-R (p<0.10)
E-R (p<0.10)
E-R (p<0.10)
(for notes, see end of the table)
57
Table 6. (continued )
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
SHARP, 1993
(174)
Sh sx/PE:
inadequate arm rests p<0.05
Elbow sx:
inadequate arm rests p<0.04
Starr, 1985 (188) Nk, Sh sx:
- Neck angle
- Upper arm angle
p>0.05
p>0.05
Upper arm, elbow sx:
- Upper arm angle
- Forearm angle
- Elbow angle
p>0.05
p>0.05
p>0.05
Wrist sx:
- Forearm angle
- Elbow angle
- Hand angle
p>0.05
p>0.05
p>0.05
Stellman, 1985
(191)
Nk, Sh sx:
Ergonomic stressor
score E-R
Arm/wrist sx:
Ergonomic stressor
score E-R
Hand/finger cramps:
Ergonomic stressor score E-R
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
58
In another study by the same principal author, in which postures and work
station dimensions were measured for data entry workers, arm discomfort
increased with keyboard height above seated elbow level (170). Right arm
discomfort increased with the degree of shoulder flexion, as the right arm reached
to the keyboard, and with ulnar deviation required to maintain the hand over the
keyboard (Table 6). With the left hand manipulating documents, left arm dis-
comfort was associated with the distance required to reach the documents. These
correlations were statistically significant at p=0.05 even though the measurements
were only obtained on 40 subjects and the analyses were adjusted for potential
confounding variables. Upper extremity symptoms were more frequent on the
right side than on the left for data entry, whereas no such difference was found
among VDU users with more varied tasks. In the small study of 34 state agency
billing clerks (174), both shoulder and elbow disorders were significantly
associated with no or inadequate arm rests, in multivariate analyses.
One study of directory assistance operators also examined work postures
(measured from photographs) in relation to symptoms, and found no statistically
significant correlations (188). However, misclassification of exposure to non-
neutral postures (based on one single observation per worker and derived from a
two-dimensional image) cannot be ruled out. Other flaws mentioned above for
this study include potential selection bias and failure to adjust for age, gender, or
other risk factors.
In a pair of studies on a large population of public sector employees, specific
job features were compared among office workers who used VDUs, typewriters,
or neither (190, 191). A composite scale of ergonomic stressors was constructed
from frequencies of awkward postures and of workstation features. The
musculoskeletal symptom score was highly correlated with the ergonomic stressor
score, both in the total population (191) and when the sample was restricted to
female employees (190). However, as noted previously, there was no statistical
adjustment for age or other covariates.
In summary, these studies are extremely consistent in showing large adverse
health effects of poorly physical workstation ergonomics. With the exception of a
few studies (102, 170, 188), virtually every posture and dimension studied was
shown to be associated with increased risk of upper extremity disorders.
Numerous observations attest, e.g., to associations between postural elements
such as ulnar deviation, shoulder flexion or static work position and
musculoskeletal problems. Likewise, work station elements related to postural
constraints such as too high or too low keyboard height were also found to be
associated with musculoskeletal problems. In some studies, an interaction
between these observations and the work performed were also suggested - by and
large, these relationships were more profound among data entry workers.
Keyboard model
Four studies attempted to examine the effect of keyboard model, in several
instances because the employees themselves complained about specific design
59
features to the investigators. However, only Bergqvist and colleagues were able
to assess such an effect, reporting that assignment of a new keyboard was
associated with recovery from neck/shoulder and upper arm disorders over the
seven-year follow-up period (15). The authors assumed (but did not specify) that
the new keyboards were ergonomic improvements compared to the old ones.
Burt et al. found that the keyboard model depended on department and job, so
in order to avoid confounding by job requirements, keyboard was only examined
within one department (of five), and no association was found, although statistical
power was limited by this constraint (30). The other two authors were unable to
study keyboard model because the equipment used by study subjects kept
changing during the study period (21, 22, 78, 79). Thus, little substantial
epidemiological evidence exists today concerning the different effects of using
various keyboard models. However, there are a number of experimental studies in
the literature (see the introduction) which provide the basis for strong a priori
hypotheses that could inform future epidemiologic research in this area.
Use of a mouse or other input devices
Several cases of various adverse hand/wrist problems among mouse users have
been reported, sometimes with only a few (e.g., 2-3) hours per day of usage.
Disorders included ulnar neuropathy, tenosynovitis affecting the wrist, finger
flexors, and lateral epicondyle attachment, and myofascial pain syndrome (39, 57,
130, 139, 148). In the United States, workers' compensation claims for
"cumulative trauma disorders of the upper extremity" increased markedly, both as
a percentage of all claims and as a percentage of claims for these disorders, from
1986 to 1993 (56). There is no evidence to date that there is a unique syndrome
associated specifically with mouse use; instead, the risk of upper extremity
disorders among mouse users appears to be similar to those observed in other
types of work with repetitive hand motion and static loading of the shoulder and
arm muscles (99). In a group of seven graphic artists with intensive mouse usage,
three prevalent, clinical cases of carpal tunnel syndrome were found (43%), in
contrast to no cases among 39 other office workers (59). Although these data are
very sparse and do not permit extensive statistical analysis, the difference between
the two groups is suggestive.
Among 652 civil engineers, longer work hours with the computer mouse were
associated with higher prevalences of symptoms of the upper arms, elbows, wrist,
and fingers, especially of the arm that used the mouse (99). In further multivariate
analyses restricted to 542 CAD operators, controlling for age and gender, both
mouse location and duration of mouse use were associated with the odds of
musculo-skeletal symptoms of the upper extremity on both the left and right sides
(see above, Tables 2 and 6). In a smaller study of 12 mouse and 12 non-mouse i.e.
keyboard users, it was observed that the two devices required different postural
deviations of the upper extremity. The keyboard users reported, on average, more
intense discomfort ratings, although the differences were not statistically
significant (98).
60
There appear to be no epidemiologic studies relating to the possibility of
musculoskeletal disorders from non-keyboard input devices other than the mouse.
One experimental study compared several short-term effects between mouse and
track-ball use in 15-minute work periods (100). Work with the mouse led to
higher shoulder elevation and higher shoulder muscle activity in the arm using the
device. More wrist extension was observed with the track-ball, but overall there
were no significant differences in discomfort between the two input devices. Any
further specific knowledge concerning effects of other input or interactive devices
such as haptic devices (118) is limited to predictions based on general knowledge
about postures, repetition rates, etc., which in turn are based more on studies of
non-VDU work, although still likely relevant.
In summary, there are substantial indications from case reports as well as from
experimental studies to suggest that certain aspects of using a mouse as an input
device could result in musculoskeletal problems. Current information from
epidemiological studies is - in this respect - rather limited. Nevertheless, the few
studies available do support the occurrence of adverse health conditions being
related to mouse use, especially for more than 5 hours per day or when placed far
from the body.
Visual demands, corrective lenses and monitor placement
In a few epidemiological studies, associations between use of various types of
spectacles (monofocal, bifocal and progressive) or contact lenses and muscle
problems have been investigated (see Table 7). The study by Bergqvist et al. (20)
reported that among VDU users, the use of spectacles (not specified as to type)
was associated with diagnosed cervical disorders. For other disorders or
discomforts, use of spectacles were not retained in the final models. Further
analysis also suggested that using bifocals or progressive glasses was an effect
modifier of VDU use duration - individuals who worked for more than 20
hrs/week at a VDU and used bifocals/progressive glasses had a high odds ratio for
tension neck syndrome (TNS) compared to non-VDU users, while individuals
working long hours with monofocal glasses or without spectacles or who worked
shorter hours were not at a higher risk for TNS (19).
In the study by Hales and co-workers (79), while no association between use (in
general) of spectacles during work was associated with neck disorders, using
bifocals was so related (Table 7). In some contrast, Sauter et al. (170) did not find
an impact of use of spectacles on discomforts in e.g. trunk discomforts. The fact
that neck discomfort was not specifically evaluated may have some bearing on
this, on the other hand, neck discomforts was rather highly correlated with upper
and lower back, motivating the authors to summarise discomforts in all these
regions into one index. In another investigation by the same principal author (168,
169), the use of corrective eyewear was correlated with both upper torso and
limb/extremities discomforts. The SHARP report (174) found that neck symptoms
were significantly associated in multivariate analysis with not wearing eyeglasses
61
or contact lenses. Presumably corrective eyewear and glare were not significantly
associated with shoulder disorders of either type.
The study by Bergqvist and co-workers (19) further suggested that the presence
of glare on the screen could be an effect modifier; among individuals who worked
more than 20 hours/week, the presence of glare was an apparent predictor of
cervical disorders. Associations between discomforts and glare were also reported
by Ryan and Bampton (166) for upper extremity symptoms (p=0.02, not specified
for shoulder/neck or lower arm) and by the SHARP report (174) in terms of the
neck (p<0.05, no point estimates given) and in terms of hand/wrist discomforts.
For elbow symptoms, glare was positively associated (p<0.01) in crude but not
multivariate analyses (174).
A few studies have also investigated the posture of the neck, specifically the
head-neck tilt angle and discomforts (see Table 6). HŸnting and colleagues (88)
found a positive association between this tilt angle and neck pain or stiffnes.
Sauter (168, 169) found a positive correlation between gaze angle and
limb/extremities complaints, but not with upper torso complaints. Neither study
did apparently adjust for other factors, though. Bergqvist et al. (20) noted that a
highly placed VDU were associated - in the univariate analysis - with increased
neck discomforts. Such a VDU placement is more consistent with a backwards
than a forward tilt of the head. However, this factor was not retained in the final
multivariate model - and it is conceivable that it may be influenced by the use or
non-use of various spectacle types. (When restricting the analysis to intensive
neck/shoulder discomforts, a highly placed VDU was retained in the final model -
but with a very uncertain estimate, see Table 6.) Finally, neither Sauter et al.
(170) nor Starr and co-workers (188) could find a relationship between head-neck
tilt angle or viewing angle with trunk or neck discomforts. In the study by
Hoekstra et al. (84), a Ônon-optimal adjusted VDU screen locationÕ was associated
with shoulder problems (see Table 6).
In a few studies, the use of spectacles, especially bifocals or progressive glasses
were found to be associated with increased neck problems, consistent with some
experimental data (see page 11). It should be observed, however, that such
associations are likely closely interrelated with the work performed, the position
of the screen and other visual elements etc. Accordingly, the complex
interrelationships between visual and postural demands and thus between visual
and neck discomforts should be further investigated.
62
Table 7. Use of spectacles during VDU work in relation to upper extremity musculoskeletal disorders: Relative risk with 95% confidence interval (CI)
and/or p-values based on hypotheses testing.
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health
Outcome*
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Bergqvist, 1995b
(20)
Nk/Sh sx: glasses in
VDU use (yes/no)
Cervical disorders:
VDU use with vs.
without spectacles
p>0.05 in multivariate
analysis
OR=4.0 (1.3-12.5)
Arm/hand
(see to the right)
Any arm/hand sx or
disorders: VDU use with
vs. without spectacles
p>0.05 in
multivariate
analysis
Bergqvist, 1995a
(19)
TNS: Use of bifocals
and work >20 h/w at a
VDU, vs no VDUs OR=6.9 (1.1-42.1)
Hales, 1992, 1994
(78-79)
Nk or Sh dx: VDU use
with vs. without
spectacles
Nk dx:
p>0.05 in multivariate
analysis
OR=3.8 (1.5-9.4)
Elbow dx: VDU
use with vs.
without spectacles
p <0.05 in initial
analysis, but p>0.05
in multivariate
analysis
Hand/wrist dx: VDU use
with vs. without
spectacles
p>0.05 in
multivariate
analysis
(for notes, see end of the table)
63
Table 7. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Sauter, 1983, 1984
(168-169)
ÒUpper torsoÓ discom-
fort: VDU use with vs.
without spectacles p<0.05
Limb/estremities
(see to the right)
ÒLimb/extremitiesÓ discom-
fort: VDU use with vs. with-
out spectacles
p<0.05
Sauter, 1991 (170) Trunk sx: glasses in
VDU use (yes/no)
p>0.05 in multivariate
analysis
Arm sx: VDU use
with vs. without
spectacles
p>0.05 in
multivariate
analysis
SHARP, 1993,
(174)
Neck sx: not using
glasses or contact
lenses
p<0.05
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
64
Types of VDU work and work organisa-
tional factors
VDU task types
In a study by Aronsson and co-workers, 2025 Swedish public employees per-
forming various types of VDU work were examined by means of a questionnaire
(10). Workers with at least 50% of their worktime allocated to data entry, data
acquisition, word processing, programming, system work, graphical production,
computer-assisted design (CAD) or mixes of these task types were compared with
workers with ÓotherÓ types of VDU work. The choice of the latter as a reference
group was dictated by a/ that being the most heterogeneous group in terms of
tasks and education, and b/ that having the lowest prevalence of muscle problems.
Based on discomfort data obtained by the ÓNordic questionnaireÓ (109), the
data entry group exhibited significantly higher odds ratios for neck and shoulder
discomforts compared to the reference group (Table 8). Similar results were
obtained for data acquisition workers. For word processors, only shoulder
problems exhibited a significant increase vs. the reference group. Utilising other
work organisational and psychological and social factors as confounders did not
result in any changes in the results for the data entry group. For the data
acquisition and the word processing groups, the shoulder ORs were somewhat
changed in these multivariate analyses (data not shown in the report). All other
task types did show increased (but non-significant) odds ratios compared to the
reference group.
It must be kept in mind, however, that the choice of the reference group was
based on that having the lowest prevalences, so at least part of these observations
could be seen as a result of the design of the study. What can be asserted is that -
after adjustments for psychological and social factors - the prevalence of neck
discomforts was highest in the data entry group, and the prevalence of shoulder
discomforts appeared to be highest in the word processing and the data entry
groups, when compared to other groups performing VDU work. Of more interest
is perhaps the use of psychological and social and other factors as possible effect
modifiers. The reported odds ratios were generally higher among men than among
women (see further below), higher for those reporting high time pressure and
demand for concentration, and reporting limited flexibility and control. For
individuals reporting frequently being tired, the contrast between task types
appeared to be less than among those not reporting tiredness (10), suggesting that
some, but not all, of the effect was mediated through fatigue.
In another Swedish study by Bergqvist et al. (19), using the same
questionnaires, Ódata entryÓ (data entry and/or word processing) workers and
ÓinteractiveÓ (other types) VDU workers were compared to non-VDU general
65
Table 8. Effect of different type of VDU tasks on frequency of upper extremity musculoskeletal disorders: Relative risk with 95% confidence interval
(CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Studies on questionnaire-derived discomforts - comparisons with non-VDU office workers
Bergqvist, 1995a
(19)
Data entry workers
Nk, Sh sx:
<20h/wk (at VDU)
20h/wk (at VDU)
OR=3.2 (1.1-9.8)
OR=1.7 (0.8-3.4)
Arm/hand
(see to the right)
Data entry workers
Arm/hand sx:
<20h/wk (at VDU)
20h/wk (at VDU)
OR=1.6 (0.6-4.5)
OR=1.8 (0.8-3.9)
Burt, 1990 (30) Arm/elbow sx: OR=2.5 (1.5-4.0) Hand/wrist sx: OR=2.4 (1.6-3.4)
Ong, 1981 (142) Arm/shoulder
(see to the right)
Data entry workers
Arm/shoulder aches: PR=1.9 (1.1-3.5)
Ong, 1984 (141) Data entry workers -
R side Nk sx:
- L side Nk sx:
Conversation
workers, Nk:
OR=7.3 (2.6-20.4)
OR=24 (7-80)
No differences with
non-VDU workers
Data entry workers -
R arm sx:
Conversation
workers, Nk:
OR=22 (6.3-76)
No differences with
non-VDU workers
Data entry workers
- R hand sx::
Conversation workers,
Nk:
OR=8.2 (2.9-23)
No differences with
non-VDU workers
Smith, 1981 (183) Clerical workers:
Professional
workers:
Significant (p<0.05)
increased occurrence of
13 UE discomfort
types or locations.
No significant
differences in these
discomforts.
UE discomfort
(see to the left)
UE discomfort
(see to the left)
(for notes, see end of the table)
66
Table 8. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health
Outcome*
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Studies on questionnaire-derived discomforts - comparisons between individuals with different VDU tasks
Aronsson, 1992
(10)
Data entry, Nk sx:
- Sh sx:
Data acquisition, Nk sx
- Sh sx:
Word processor, Nk sx
- Sh sx:
All compared to mixed
VDU tasks (group with
lowest sx prevalence)
OR=1.8 (1.2-2.7)
OR=2.1 (1.4-3.2)
OR=1.6 (1.1-2.4)
OR=2.2 (1.5-3.4)
OR=1.3 (0.8-2.0)
OR=1.7 (1.1-2.7)
DeMatteo, 1993
(41)
Data entry,
- at least six muscle
symptoms
- previously diagn MS
injury
All compared to general
office work (including
VDU)
OR=4.4 (1.7-11.5)
PR=3.0 , p<0.05
(for notes, see end of the table)
67
Table 8. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Studies on questionnaire-derived discomforts - comparisons between individuals with different VDU tasks (continued)
Grieco, 1989
(71)
For women:
Data entry, Nk sx,
Data acquisition, Nk sx
Dialogue, Nk sx
For men
Data entry, Nk sx,
Data acquisition, Nk sx
Dialogue, Nk sx
All compared to total
group of telecom.
workers (including
VDU)
SMR= 83 (p<0.05)*
SMR=114 (p<0.05)*
SMR=110 (p<0.05)*
SMR=100 (p>0.05)*
SMR= 77 (p<0.05)*
SMR= 98 (p>0.05)*
For women:
Data entry, UE sx,
Data acquisition, UE sx
Dialogue, UE sx
For men
Data entry, UE sx,
Data acquisition, UEk sx
Dialogue, UE sx
All compared to total
male or female group of
telecom. workers
(including VDU)
SMR= 77 (p<0.05)*
SMR=115 (p>0.05)*
SMR=116 (p<0.05)*
SMR= 98 (p<0.05)*
SMR= 68 (p<0.05)*
SMR=104 (p>0.05)*
UE sx:
(see to the left)
Heyer, 1990
(81)
Data entry, Nk sx
Compared to interactive
VDU work
PR=1.4 (1.1-1.8) Data entry, arm/elbow sx
Compared to interactive
VDU work
PR=2.2 (1.5-3.4) Data entry,
hand/wrist sx
Compared to
interactive VDU
work
PR=1.9 (1.4-2.7)
Studies examining disorders based on diagnoses - comparisons with non-VDU workers
Bergqvist,
1995a (19)
Cervical dx:
- data entry < 20 hr/wk
- data entry 20 hr/wk
- interactive
TNS dx:
OR=1.2 (0.4-4.3)
OR=1.7 (0.7-4.3)
No association
No association
Arm/hand
(see to the right)
Arm/hand dx:
- data entry
- interactive
No association
No association
(for notes, see end of the table)
68
Table 8. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
HŸnting, 1981 (87) Data entry, Nk/Sh PE:
(see further Table 9)
OR=4.8 (1.8-13.4) Data entry, forearm
PE:
(see further Table 9)
OR=4.8 (1.8-13.4)
Studies examining disorders based on diagnoses - comparisons between individuals with different VDU tasks
Hales, 1994 (79) Loop Provisioning
(telecom) workers,
hand/wrist dx:
Compared to Mail
Remittance (data
entry) workers
OR=1.9 (0.7-5.1)
HŸnting, 1981 (87) Data entry, Nk/Sh PE:
Compared to inter-
active workers
OR=1.5 (0.8-3.0) Data entry, forearm
PE:
Compared to inter-
active workers
OR=2.7 (1.2-6.0)
* Reference group for all SMRs shown: the entire telecommunications company workforce
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist.
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
69
office workers, after adjustment for a large number of individual, ergonomic and
work organisational/psychological and social factors. ÓData entryÓ workers
exhibited some excess odds ratios compared to non-VDU workers for
neck/shoulder and arm/hand discomforts as well as diagnosed cervical disorders.
These excesses were statistically significant only for neck/shoulder discomforts,
though, and then only for those with limited VDU work (see Table 8). One
tentative explanation for such a peak at an intermediate VDU work time (seen
also in a prospective part of the study (18)) is that the parameter Ôtask typeÕ may
be able to select adverse conditions better than VDU work time, since data entry
individuals may have equally (adverse) conditions also in the part of the work day
not spent working at a VDU (18). For ÓinteractiveÓ workers, no discomforts or
disorders were more prevalent among the VDU than the non-VDU workers (19).
Some other conditions modified the Ódata entryÓ effect in this study.
Neck/shoulder discomforts were found primarily among data entry workers with
limited (spontaneous) rest break ability (OR 4.8; 1.3-18.1) compared to inter-
active VDU or non-VDU workers (for the latter groups, rest break ability
appeared not to be relevant). Similarly, arm/hand discomforts was primarily found
among data entry workers who had their keyboard placed low (OR 2.8; 0.9-8.6).
There was a linear response - among data entry workers only - between (lower)
keyboard vertical position and (higher) odds ratio for arm/hand discomforts (19).
In a study among US newspaper workers, reporters were found to have more
elbow/forearm and hand/wrist problems in the past year than non-keyboard
workers (Table 8). A Canadian study (41) compared data entry (VDU) workers
with general office workers (with and without VDU work). 27% of the data entry
users reported 6 or more musculoskeletal symptoms, compared to 7.3% of the
general office workers. They also had more often previously diagnosed
musculoskeletal injuries (Table 8). No adjustments were made for other factors.
Furthermore, the data entry clerks took breaks more regularly but worked more
overtime and had less variability in physical activities and postures.
In a large Italian study on telecommunications workers (N=29 759), age-
adjusted standardised morbidity ratios (SMRs) were computed, comparing
questionnaire derived discomforts in neck and upper limbs among various type of
VDU workers with those of the general SIP (Italian Telecom) population. Women
with ÓenquiryÓ (data acquisition) or ÒdialogueÓ (interactive) work had slightly
increased SMR«s for neck discomforts and upper limb problems. In contrast,
female ÓloadingÓ (data entry) workers had SMR«s below 100 for both neck and
upper limb (Table 8). Male SMR«s were generally closer to 100, except for
ÒenquiryÓ, where substantially and significantly reduced SMR«s were noted. In
this study, non-VDU users had higher upper limb SMR«s than the VDU users
(71).
In another (US) telecommunication worker study (79), VDU/keyboard task
subjects were divided into various job titles. Mail Remittance workers, who
primarily performed data entry work, had average prevalences compared to the
70
other VDU workers in the study (20% vs. 22%) and were therefore chosen as the
reference group. The highest prevalences of diagnosed upper extremity
musculoskeletal disorders were found among the 69 Loop Provisioning
employees (Table 8). This work was characterised as the Ómost intellectually
demanding jobÓ. In the multivariate analysis, individuals with this job title
exhibited more hand-wrist disorders. It should be noted, however, that the relative
ranking of various job titles in terms of disorder prevalences varied substantially
across the three study sites.
Heyer et al. (81) reported that symptoms were more frequent in data entry than
in interactive tasks for all three regions of the upper extremity. However, as noted
previously, they did not adjust for the simultaneous effects of multiple
occupational exposures.
Hocking (83) reported that the highest annual incidence rates of "repetition
strain injury" were among telephonists, rather than clerical workers using VDUs
or telegraphists (a small group). Insufficient exposure information was provided
to interpret these differences, although the authors stated that they had an inverse
relationship with (average) keystrokes per hour. Other problems with this paper
include likely incomplete and non-uniform case reporting, possible double-
counting of cases through multiple data sources, and very approximate
denominators (numbers of workers at risk) for each job group. (Results not shown
in Table 8.)
A Swiss study (88) examined - by way of anamnesis, palpation and other tests -
various muscle problems in relation to task types, as seen in Tables 8 and 9. All
of the sub-groups were similar in age, but the data entry VDU users and the
typewriter users had much higher proportions of women than the other groups
(see Table 9). No comparisons were stratified by gender - and thus confounding
or effect modification may exist in some comparisons. As can be seen in table 9
(which is limited to comparisons of groups with similar gender composition), task
characteristics appear to be of primary importance for the risk of some muscle
disorders. Both data entry workers and non-VDU typists (presumably both
performing data entry work) experienced similar odds for disorders, at least in the
neck/shoulder region. In some contrast, substantial differences were found
between traditional (non-VDU) office workers and interactive VDU workers. In a
continuation of this study, van der Heiden and co-workers (197) compared daily
and occasional pain among 69 Computer Assisted Design (CAD) workers with
these other groups. Prevalences of neck, shoulder, arm and hand pain were lower
among the CAD workers than among all other groups (both VDU and non-VDU
groups). No statistical analysis was performed, however, and no information was
available on the gender of these CAD operators, precluding any definite
conclusion (data not presented in Table 8).
71
Table 9. Odds ratios for muscle problems in relation to various task types. Data from
HŸnting et al. (88).
Interactive or (presumably) varied
office tasks
Data entry tasks
Traditional office
work (n=54)
(60% women)
Interactive VDU
work (n=109)
(50% women)
Typewriters
(n=78)
(95% women)
VDU data entry
work (n=53)
(94% women)
Neck/shoulder
tendomyotic
pressure pains
1.0 3.2
(1.2-8.2)
1.0 1.1
(0.6-2.4)
Painful limited head
movability
1.0 3.4
(1.2-9.4)
1.0 0.7
(0.4-1.5)
Pain in isometric
forearm contrac-tion 1.0 1) 2.9
(0.8-10.5) 1) 1.0 1.6
(0.7-3.4)
Odds ratios and 95% confidence intervals calculated from published
prevalences. Note that the odds ratios are given for different
comparison,
i.e.
different choices of reference groups - the latter
indicated by an odds ratio of 1.0. Thus, for data entry workers, the odds
ratio is 1.1 compared to typists, while the odds ratio is 3.2 for interactive
VDU workers compared with traditional office workers (see ÒNeck/
shoulder tendomyotic pressure painsÓ). Due to varying gender
compositions, no comparisons are made between
e.g.
, data entry workers
and traditional office workers. No adjustments were made for other
factors.
1) Only three cases in the reference group, resulting in very wide
confdidence intervals for the odds ratios.
In a Singaporese study by Ong and colleagues (141, 142), VDU workers
performing data entry (n=36) or conversational VDU (n=26) tasks were compared
to 41 conventional office workers (without VDU work). Discomforts were more
prevalent among the data entry workers compared to non-VDU users for the neck,
right arm and right hand, whereas differences between conversational VDU and
non-VDU workers were not significant (Table 8). Non-VDU workers had slightly
higher prevalences of right side shoulder problems than the VDU workers, but the
data entry users had slightly higher prevalences in the left shoulder. The authors
attributed the preponderance of left side neck and shoulder problems among data
entry workers to the different work postures, since these workers had their
manuscripts placed at the left side (142).
Workers in newspapers and bank offices (n=412) were examined by Smith and
co-workers in the US (183). Professional VDU workers were reporters, editors
etc., with substantial control of their workpace. Clerical VDU workers, on the
other hand, performed data entry, data retrieval, telephone inquiry work etc., with
little control of their workpace. Among ÓclericalÓ workers using VDUs,
prevalences of a large number of body site complaints were significantly higher
compared to control subjects (not using VDUs). Such differences were not found
when comparing ÓprofessionalÓ VDU users with non-VDU workers, however. No
72
adjustments for other factors were performed and biased self-selection of
participants was possible.
Attempting to summarise this information, a strong case can - in our opinion -
be made for an increased occurrence of muscle problems among data entry
workers compared to non-VDU users, both as to arm and hand problems and as to
neck and shoulder problems. For data acquisition or interactive work, a similar
conclusion appears more uncertain. A definitive summary statement is slightly
compromised by the following considerations, though:
· Different choices of reference groups will have a major impact on the results
(see e.g. (120)). In these studies, the reference groups varied among other VDU
users, non-VDU users with varied work tasks, and all workers within the
organisation.
· The different task types do not appear to be well-defined in terms of various
psychological and social or work organisation parameters. Some indications of
this can be found by the large impact - in some studies - of various effect
modifiers. See for example the study by (10), where heterogeneity in these task
types is shown in terms of job tasks and educational level.
Nevertheless, these studies provide extensive evidence for an association
between certain type of VDU work types - especially data entry type jobs - and
musculoskeletal problems. It could be argued whether it is the type of job that
should be regarded as the causative agent, or the (likely) components of such jobs
such as repetitive, monotonous work situations with static postures, low decision
autonomy and skill utilisation etc. The advantages of clarifying the impact of such
factors within e.g., data entry work situations would be two; the ability to more
precisely design intervention, and the ability to extract and generalise findings
into other work types. On the other hand, failure to precisely describe the relative
impor-tance of or the interactions between such factors should not detract from
the fundamental finding of a strong association between data entry work and
musculoskeletal disorders.
A final point is that for many (perhaps most) of these factors, the occurrence
and impacts may well be strong also in certain office work situations without
VDUs. It could be, however, that such non-VDU work situations (e.g. extensive
typing) are disappearing with ongoing computerisation. (It should be noted that
the HŸnting study, which provides the strongest argument for similar risks in
ÓsimilarÓ non-VDU work situations - i.e., extensive typing - was performed in
1981.)
Work load and work demand
Aronsson and co-workers (10) reported Óelevated odds ratios and dose-response
relationshipsÓ between high work pace and discomforts in the neck and shoulders,
unfortunately without any details on the specific associations found. In contrast,
Bergqvist et al. (20), using the same questionnaire items, failed to find
73
associations between a high work pace or overtime on short notice and muscle
problems. Frequent overtime was associated in the final multivariate model with
arm/hand discomforts, see Table 2.
The two NIOSH newspaper studies cited above provided strong evidence of the
importance of work pace for upper extremity disorders. Bernard et al. (21, 22)
showed that working under deadline for at least 30 hours per week was associated
with neck and hand/wrist symptoms (Table 2) Restricting the analysis to routine
workers, the neck odds ratios were substantially increased (OR 2.8; 1.1-7.1). High
perceived job pressure was associated with shoulder symptoms The risk of
hand/wrist disorders consistent with carpal tunnel syndrome was also associated
with an increase in overall workload during the previous year (22). Burt and
colleagues found marked exposure-response relationships for both shoulder and
hand/wrist disorders with typing speed (30). Similar univariate trends were seen
for neck and elbow/forearm disorders but were no longer statistically significant
when hours of keying was also included in regression analysis. In addition,
reporters, who were observed to work at a faster pace and to get up less often
from their work stations than other employees, had more than twice the risk of
arm/elbow and hand/wrist disorders, even after controlling for other work
characteristics (Table 10).
In an Italian study by Camerino et al. (32), spinal/cervical complaints were
associated with insufficient personnel situations (p=0.05) and aggressive clients
(p=0.01), but not with time constraints. DeMatteo and co-workers (41) also failed
to find a relationship between UE symptoms and keying rate. Ferreira et al. (54)
found that the monthly incidence of cases of hand/wrist disorders increased when
there was a new management goal that increased time pressure at work.
In a study of telecommunications workers, Hales et. al (78, 79) found
associations in multivariate analyses between neck disorders and work pressure,
and with high information processing demands; between hand/wrist disorders and
high information processing demands; and between elbow disorders and surges in
workload. There was little power to examine the effect of keystrokes per day
among directory assistance operators, since data were only available for one job
title, in which the average number of keystrokes was low and there was little
variability. Surges in workload had an elevated risk of arm/elbow disorders; work
pressure, which has both physical and psychological attributes, was associated
with neck/shoulder disorders and the severity of arm/elbow and hand/wrist
disorders (Table 2). (See also Òpsychological and social factors,Ó below.)
Hoekstra et al., reporting on Óteleservice representatives,Ó also could not
examine the effect of hours per day because there was so little variability among
study subjects (84). However, they found that neck/shoulder problems were
associated with a continuously changing workload during the day. Since this job
feature was also present for all study subjects, it suggests that workers with
neck/shoulder symptoms were more bothered by the workload (i.e., potential
information bias).
74
Table 10. Effect of VDU work pace or intensity of work demands on frequency of upper extremity musculoskeletal disorders: Relative risk with 95%
confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group, Health
Outcome
Findings
Aronsson, 1992
(10)
Nk and Sh: High work
pace
Elevated odds
ratios and dose-
response (no data
given)
Bergqvist 1995b
(20)
Nk/sh sx: High work
pace
p>0.05 in multi-
variate analysis
Arm/hand
(see to the right)
Arm/hand sx: High work pace p>0.05 in multi-
variate analysis
Bernard, 1993,
1994 (21-22)
Nk: Hours on
deadline/wk: 30-39
vs 0-10
Nk: Highly variable
workload
Sh: Job pressure
OR=1.7 (1.4-3.0)
OR=1.5 (1.1-1.8)
OR=1.5 (1.0-2.2)
Phase I: Hours on deadline/wk:
30-39 vs 0-10
Phase II : One-year increase in
overall workload
OR=1.7 (1.2-2.3)
OR=3.2 (2.5-4.1)
Burt, 1990 (30) Sh: Typing speed
moderate
Sh: Typing speed fast
OR=2.6 (1.1-5.9)
OR=4.1 (1.8-9.4)
Typing speed p<0.02 in
initial,
p>0.05 in
multivariate
analysis
Typing speed moderate
Typing speed fast
OR=1.3 (0.6-3.1)
OR=2.5 (1.0-5.6)
Camerino 1995
(32)
Nk/sh sx: Time
constraints
No association
(p>0.05)
(for notes, see end of the table)
75
Table 10. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
DeMatteo, 1993
(41)
MS sx, all UE regions:
Keystroke rates p<0.41
Ferreira, 1997 (54) Increase in time
pressure
b=0.05, p=0.008
Hales, 1992, 1994
(78-79)
Nk: Work pressure
Nk: High information
processing demands
Sh: Work pressure
OR=2.4 (1.1-5.5)
OR=3.0 (1.4-6.2)
E-R (p<0.05)
Surges in workload
Work pressure
OR=2.4 (1.2-5.0)
E-R (p<0.05)
Work pressure
High information
processing demands
E-R (p<0.05)
OR=2.3 (1.3-4.3)
Hoekstra, 1994 (84) Nk: Highly variable
workload
OR=1.2 (1.0-1.4)
Kamwendo, 1991a
(94)
Nk: Too much to do
Sh: Too much to do
p=0.01
p=0.05
(for notes, see end of the table)
76
Table 10. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Lim, 1994 (116) UE MSD sx:
Work pressure
Production quota
UE CTD index:
Work pressure
Production quota
r=0.07 (p>0.05)
r=0.11 (p>0.05)
r=0.17 (p>0.05)
r=0.14 (p>0.05)
UE MSD sx
(see to the left)
UE MSD sx
(see to the left)
Ryan, 1988 (166) Neck/shoulder/forearm:
Work pressure >3
times/week OR=3.9 (1.3-12.2)
Nk/sh/forearm
(see to the left)
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
77
Lim and Carayon (116) compared a number of self-reported job characteristics
within a small group of office workers (n=129, 87% female). They posited an
etiological model with interactions among physical and organisational features of
work and analysed the data in a hierarchical manner consistent with their model.
They found that work pressure and production quotas were associated with
repetitive motions, awkward postures, dynamic work activity, and psychological
anxiety, all of which were in turn associated with upper extremity musculoskeletal
disorders. However, once the physical factors had been put into a multivariate
model, neither work pressure nor production quotas was associated directly with
upper extremity problems. (It should also be noted that there was no adjustment
for age or other non-occupational risk factors.)
Work demand was not associated with muscle problems according to Marcus
and Gerr (125), whereas Ryan and Bampton (166) did find relationships between
muscle disorders and Ópushing oneself more than 3 times/weekÓ.
In summary, a variety of indicators of job demands and workload has been
studied by different investigators. Of the two studies that directly examined the
effects of typing speed, the one that was larger and methodologically more sound
(30) found significant associations with shoulder and wrist/hand problems, while
the smaller, weaker one (41) did not. The majority of the studies that examined
some other indicator of workload or work pressure did find some evidence of its
role in upper extremity and neck problems. The failures of at least one study -
after extensive multivariate analysis - to verify an association between work load
or work demand and MSDs does, however, somewhat reduce the confidence in
this conclusion. It should also be noted that the effect of work load cannot easily
be partitioned into physical and psychological components or mechanisms, and
that Lim and Carayon (116) may be correct that work pressure may act through
(i.e., by causing and intensifying the effects of) physical factors such as repetitive
hand motion, rather than having an independent causal effects on MSD risk.
Repetitiveness of keyboard work
Work ÓrepetitivenessÓ may be characterised either in terms of the speed of body
motions or the extent to which those motions are repeated without variation
(ÓstereotypyÓ, 4). In the operation of VDUs, repetition has often been assessed
through keying speed (e.g., keystrokes per day) or as performance of data entry
versus interactive tasks; other variables have been available in a few studies.
Comparisons between data entry vs. interactive work are discussed above under
ÓVDU task types,Ó and typing speed is addressed under ÒWork load and work
demands.Ó Here, we direct our discussion to more explicit measures of
stereotypy, including monotonous work, lack of variation in tasks, and low skill
utilisation or opportunity to learn new things.
78
Table 11. Effect of repetitiveness or stereotypy of VDU work on frequency of upper extremity musculoskeletal disorders: Relative risk with 95%
confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure,
Study
Group,
Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Aronsson, 1992
(10)
Nk and Sh: Limited work
task flexibility
Elevated odds ratios
and dose-response (no
data given)
Bergqvist, 1995a
(19)
Intense nk/sh sx: Repetitive
movements 320 hr/wk, and
stomach reaction (stress) OR=3.9 (1.1-13.8)
Bergqvist, 1995b
(20)
Intense nk/sh sx: Repeated
work movements
Any sh dx: Low task
flexibility
OR=3.6 (0.4-29.6)
OR=3.2 (1.2-8.5)
Bergqvist, 1995c
(15)
Nk/sh sx: Decreased
manuscript use
RR=1.5 (0.8-2.7) Arm/hand
sx (see to
the right)
Arm/hand sx:
- Increased keyboard use at
newspaper:
- Increased keyboard use at
post office:
- Increased monotony of
work:
RR=5.3 (2.1-13.7)
RR=3.5 (0.4-30.3)
RR=3.1 (1.2-7.8)
(for notes, see end of the table)
79
Table 11. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Bernard, 1993 (22)
(21)
Nk: Low work variance OR=1.7 (1.2-2.5)
Hales, 1992, 1994
(78-79)
Nk: Variety of work tasks OR=2.9 (1.5-5.8)
Lim, 1994 (116) UE MSD sx:
Repetitiveness
UE CTD index:
Repetitiveness
r=0.18 (p<0.05)
r=0.24 (p<0.01)
UE MSD sx
(see to the left)
UE MSD sx
(see to the left)
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
80
In the study by Aronsson and colleagues (10), limited work task flexibility was
associated with neck and shoulders complaints (no data given). Bergqvist et al.
(20) found an association between the same task flexibility variable with shoulder
diagnoses after multivariate adjustment for other factors. In a longitudinal
analysis of the same study group, women who (retrospectively) reported
increasing monotony in their work tasks during a 7 year period also had a higher
incidence of hand/wrist problems during the same time period (15). In addition,
decreased use of hard copy and increased use of keyboards were also somewhat
associated with upper extremity problems. As noted above, elevated risks of
neck/shoulder symptoms were found for the combinations of data entry work with
limited opportunity to take rest breaks and VDU use for at least 20 hours per
week with repetitive movements (19). The authors also separated their study
participants according to whether the work could be classified as ÓroutineÓ or not,
but did not report any association with MSDs (20).
Among US telecommunication workers (21), limited variance in VDU work
was associated with neck/shoulder disorders (OR 1.7; 1.2-2.5) in the final
multivariate model. In an analysis on a sub-set of the population with
predominantly routine work, this odds ratio was higher (2.4; 0.9-6.9). In a
Finnish study (111), the index Ólimited satisfaction with the work contentÓ (which
included work variety) was associated with neck/shoulder symptoms (numerical
details unclear from the report).
In contrast, the study of Hales et al. (79) failed to indicate any positive
association between limited task rotation or limited task variation and muscle
problems. In fact, an opposite association was found for task variation. The
authors commented on this Ócounter intuitiveÓ finding by suggesting that Ótasks
within the present job offered little opportunity for relief of biomechanical or
psychic loads,Ó highlighting the limited value of job rotation or enlargement when
the tasks available to be combined all have similar demands.
Lim and Carayon (115, 116) found, when comparing office workers who used
VDUs with respect to a variety of ergonomic job features, that two indices of
upper extremity disorders were each strongly associated with repetitiveness
(Table 11). However, the reported analyses did not examine potential
confounding by age or other non-occupational risk factors.
In an investigation of 1032 female clerical workers (190), the prevalence of
various job characteristics was compared among part- or full-time typists, clerks
(these groups without VDU use) and part- or full-time VDU clerical work. For
Óunderstanding of work processÓ and Ólearn new things,Ó the part day VDU users
scored highest and the full day VDU users scored lowest among all the groups
(VDU and non-VDU). For Ówork Ômakes senseÕ,Ó both part and full time VDU
users scored low. Full-day VDU users (but not part-time VDU users) reported a
higher level of musculoskeletal problems than non-VDU workers, but direct
analysis of their association with the work content variables was not carried out.
Overall, there are thus several indications that repetitive work or limited task
variability have a role in the development of musculoskeletal disorders among
81
VDU workers. It is conceivable that this type of factor may be at least partly
responsible for the associations seen between upper extremity MSDs and VDU
work tasks (e.g., data entry) or other job features in other studies. High work load,
few rest breaks, and little task variability often occur together in actual
workplaces, which limits the ability to study their separate effects with
epidemiological methods. Experimental studies may be more informative, since
different combinations of exposure can be created artificially, if necessary. In
summary, we believe that the interrelationships between these dimensions of
work organisation still remain to be fully explored in relation to musculoskeletal
disorders.
Rest break patterns and duration of work tasks
In the study by Bergqvist and co-workers (20), limited spontaneous rest break
ability was - after adjustments for other factors - associated with neck/shoulder
discomforts as well as several diagnoses (see Table 12). The (non-significant)
association with arm-hand diagnoses was in a further analysis (19) shown to be
interrelated with some other factors; individuals with limited rest break
opportunity and non-use of lower arm support who worked more than 20
hrs/week with a VDU were found to have an increased odds ratio when compared
with those performing non-VDU office work. (See also above for interrelation-
ships between rest breaks and data entry work.)
In a Brazilian study (26) of both VDU and non-VDU users, self-reported
absence or insufficiency of rest breaks was associated with diagnosed upper
extremity musculoskeletal disorders. In an Australian study of data entry
operators (166), diagnosed cases of upper limb disorders reported to a lesser
extent than non-cases that there was Óenough time for rest breaksÓ (80% vs.
100%, p=0.035). Few further details were given in these two reports, and no
adjustments for other factors were made, however. Some other studies have exa-
mined the frequency of rest breaks without - in the final multi-variate analysis -
finding associations with muscle problems (21, 79).
The duration of VDU work periods between breaks or change of task was not
associated with self-reported or diagnosed disorders in two Swedish studies (10,
20). DeMatteo et al. (41) actually found a negative association (p<0.005). The
authors inferred that this could have been due to other factors. In their study, data
entry workers (with shorter work periods) were compared with general office
workers (with longer periods), but who had Ógreater opportunity to frequently
change postureÓ (41). Thus, they suggest a potential confounding effect of task
types with rest breaks, unfortunately without performing an analysis either to
verify or to discredit the suggestion. An alternative possibility is that workers with
fewer discomforts work longer before taking breaks or varying their activities.
Ferreira et al. (54) used multivariate time series analysis to examine factors
that predicted upper extremity musculoskeletal disorders in interactive VDU
work. In a retrospective, dynamic cohort of 106 operators, 24 cases occurred, all
or most in the hand/wrist region. The monthly incidence decreased after a regular
82
Table 12. Effect of self-evaluated sufficiency or frequency of rest breaks on frequency of upper extremity musculoskeletal disorders: Relative risk with
95% confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Bergqvist, 1995a
(19)
Arm/hand
(see to th eright)
Insufficiency of rest breaks.
non-use of lower arm
support and VDU work
³20 h/wk compared to non-
VDU work
- any arm/hand dx: OR=4.6 (1.2-17.9)
Bergqvist, 1995b
(20)
Insufficiency of rest breaks
- Nk/sh sx:
- TNS dx:
- any shoulder dx:
OR=2.7 (1.2-5.9)
OR=7.4 (3.1-17.4)
OR=3.3 (1.4-7.9)
Arm/hand
(see to th eright)
Insufficiency of rest breaks
- any arm/hand dx: OR=2.7 (0.8-9.1)
Bernard, 1994
(21)
Number of breaks p>0.05 in
multivariate
analysis
Number of breaks p>0.05 in
multivariate
analysis
Number of breaks p>0.05 in
multivariate
analysis
Bozi Ferraz, 1995
(26)
Any musculoskeletal dx:
Insufficiency of rest breaks
Associated,
p=0.012
(for notes, see end of the table)
83
Table 12. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Ferreira, 1997 (54) Rest/work
schedule
(10 min/hr)
linear regression:
b=-0.05, p=0.02
logistic regression:
OR=0.05, p=.02
Hales, 1994 (79) Number of breaks p>0.05 in
multivariate analysis
Number of
breaks
p>0.05 in
multivariate
analysis
Number of breaks p>0.05 in multivariate
analysis
Ryan, 1988 (166) Insufficiency of rest breaks:
- neck/shoulder/forearm sx:
Associated, p=0.035
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
84
rest/work schedule (10 minutes of rest per hour) was implemented. In logistic
regression modelling of high incidence rates, the rest/work schedule was the most
significant predictor of change (decrease) in monthly incidence.
The study by Ryan and Bampton (166) included insufficient rest breaks as one
of 7 examined variables. While they found an association with muscle disorders
(Table 12), unfortunately no multivariate analysis was performed, especially since
individuals in these workplaces were exposed to so many physical workstation
problems. It is plausible that these individuals might have a lower threshold for
ÒinsufficientÓ rest time, because of their poor ergonomic conditions.
Thus, the associations of rest break pattern, flexibility or duration of work tasks
with muscle problems appear rather inconsistent. While the strongest suggestion
of an effect was found in the study by Bergqvist and colleagues (20), that study
formulated the variable in such a way (Óopportunity to take unscheduled rest
breaksÓ) that a component of decision latitude (see below) may have been
included - as is often the case in actual workplaces. Another possible
interpretation is that workers with MSDs are more sensitive to or aware of the
inability to take breaks than are workers without MSDs. Given the wide variety of
other factors that could influence the ability to take rest breaks, further insight into
this factor could conceivably be obtained from experimental or intervention
studies, examining different fixed or flexible rest break patterns, while holding
other conditions constant (see the Introduction).
Monitoring and supervision
In the study by DeMatteo et al. (41), muscle problems and use of monitoring and
quota systems were analysed. Muscle problems were strongly related to stress
(see further below), and that outcome variable was in turn related to monitoring
(p<0.03). Unfortunately, it is not clear whether any ÓdirectÓ relationship between
muscle problems and monitoring was analysed - no such results were reported. In
another North American study which obtained data on monitoring, that of Hales
et al. (79), 93% of the study subjects were subject to monitoring, and therefore
there was little statistical power to detect any association. In this study, an
association was found between the absence of a productivity standard and neck
disorders (OR 3.5; 1.5-8.3). The authors noted this Ócounter-intuitiveÓ result, but
offered few suggestions as to its cause, confounding with other work factors is
one possibility.
In the study by Hoekstra et al. (84), 100% of the subjects were continuously
monitored electronically (for productivity), and at random by supervisor (for
quality and accuracy). Very high prevalences of job dissatisfaction, physical and
mental exhaustion were reported, along with very low decision autonomy.
However, since no comparison could be made with workers not being monitored,
the contribution (if any) of monitoring to MS disorders could not be determined.
Another North American study did provide such comparisons (182), and reported
significant increases in prevalences of a number of musculoskeletal complaints
(e.g., in the fingers, wrists, shoulders and neck). However, no multivariate
85
analysis was performed, so the relative role of monitoring, or (e.g.) workload was
not ascertained (the authors noted that workload was a predictor of strain also in
unmonitored work situations, though). The major obstacle when interpreting this
study is the very low participation rate (26%), and the authors rightly caution
against making firm conclusions based on these results.
In our minds, these few results can only be summarised with the statement that
an association between muscle problems and the type of monitoring or
supervision system among VDU users has (so far) not been established. It should
also be noted that both national regulatory environments and cultural differences
could influence these relationships as effect modifiers, so these factors should at
least be specified in future studies and, where possible, their effects should be
explored.
86
Psychological and social factors
WorkerÕs control and decision latitude
In the study by Aronsson and coworkers, limited organisational influence was
associated with discomforts in neck, and shoulder (no numerical data reported
(10). In another Swedish study (20), there was, however, no such reported
association in the final multivariate models. Bernard and co-workers (21) reported
an association between Óperceived lack of participation in job decision makingÓ
and shoulder symptoms, but not between job control and muscle problems (Table
13).
Decision latitude was associated with limb numbness severity in the substudy
(n=70, with ascertained ergonomic factors) of the study by Faucett and Rempel
(47), and they also found an interaction with keyboard vertical position;
individuals with low decision latitude and a highly placed keyboard (above elbow
height) reported more upper extremity numbness (p<0.05). A similar interaction
was also found for upper torso pain and stiffness (p<0.01). It should be noted that
in the full study (n=150, but without ergonomic factor ascertainment), no associa-
tion was noted with decision latitude, once workload were entered into the model.
Hales and co-workers (79) found associations between Óroutine work lacking
decision making opportunitiesÓ and neck and elbow disorders (Table 13). Marcus
and Gerr (125) did not find any associations between decision latitude and muscle
problems. The study by Hoekstra et al. (84) reported a small but significant
association between neck disorders and Òcontinually changing workload during
the dayÓ (see Table 10), which also suggests lack of control over tasks and pacing.
In addition, shoulder disorders were also associated with working at locations
where workers perceived themselves to have less control over work policy.
Linton and co-workers reported associations between ability to influence
working conditions and neck and shoulder pain (94). In the same study, using the
index Ówork contentÓ (which apparently included such questions), elevated odds
ratios and significant linear trends were found for neck and shoulder pain,
respectively (117).
In conclusion, while some studies appear to support a relationship between
muscle problems and decision latitude or job control, other studies - including
some with multivariate analysis - fail to do so. Thus, in our opinion, a definite
conclusion regarding this aspect of VDU work and its importance for MSDs is
currently not possible, although, again, the positive results are suggestive.
However, it should also be pointed out that, although numerous investigators
have used the Karasek Job Content Questionnaire (96) to characterise decision
latitude, as well as psychological job demands, remarkably, none to date have
used the corresponding Karasek-Theorell demand/control model to inform the
analysis of the data obtained. The model specifically posits the combination of
87
Table 13. Effect of perceived decision latitude and control on frequency of upper extremity musculoskeletal disorders: Relative risk with 95%
confidence interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Aronsson, 1992
(10)
Nk and sh sx: index on
low control and
influence
Associated (no data
given)
Bergqvist, 1995b
(20)
Nk and sh sx or dx:
index on low control
and influence
p>0.05 in
multivariate analysis
Arm/hand
(see to the right)
Arm/hand sx or dx: index
on low control and
influence
p>0.05 in
multivariate
analysis
Bernard, 1994 (21) Lack of participation in
job decision making
-Sh sx:
-Nk sx:
Job control, Nk and Sh
sx:
OR=1.6 (1.2-2.1)
Not associated
Not associated
Lack of participation in job
decision making,
hand/wrist sx:
Job control, hand/wrist sx:
Not associated
Not associated
Faucett, 1994 (47) Decision latitude, limb
numbness
Severity
associated
(p<0.05)
Hales, 1992, 1994
(78-79)
Routine work lacking
decision-making oppor-
tunities, Nk dx: OR=4.2 (2.1-8.6)
Routine work
lacking decision-
making oppor-
tunities; elbow dx:
OR=2.8 (1.4-5.7)
Routine work lacking
decision-making oppor-
tunities, hand/wrist dx:
Associated;
(linear regres-
sion of symp-
tom severity
disorders),
p>0.05
(for notes, see end of the table)
88
Table 13. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Hoekstra 1994 (84) Sh sx: Lack of control
on work policy
OR=4.0 (1.1-14.6)
Kamwendo, 1991a
(94), Linton, 1989
(117)
Inability to influence
working condition;
- Nk sx:
- Sh sx:
Work content
- Nk sx:
- Sh sx:
Poor psychosocial en-
vironment index
- Nk sx:
- Sh sx:
Associated (p<0.001)
Associated (p<0.003)
OR=2.2 (1.1-4.5)
OR=2.5 (1.3-4.9)
OR=2.3 (1.3-4.0)
OR=2.5 (1.5-4.3)
Marcus, 1996 (125) Decision latitude,
Nk/Sh sx:
p>0.05 in
multivariate analysis
Arm/hand
(see to the right)
Decision latitude,
arm/hand sx:
p>0.05 in multivariate
analysis
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
89
high demands and low control to be injurious, and the evidence with respect to
cardiovascular disease is strikingly consistent with this model. Yet all of the
studies that we cite here present analysis of associations with job demands and
decision latitude independent of each other and fail to examine whether there is a
higher risk of upper extremity MSDs specifically in VDU jobs that have both high
demands and low control.
Social support and co-operation
Among government employees, individuals reporting extreme peer contacts
(defined as both low and high compared to an intermediate level) were found to
have higher occurrences of neck and shoulder problems (no data given, 10).
Bergqvist et al. (20) found relationships between the same extreme peer contact
variable and arm/hand discomforts as well as with Óany arm/hand diagnosisÓ
(Table 14). In an a posteriori analysis, separating low and high peer contacts, it
was found a/ that both showed increased muscle problems compared to
intermediate levels, and b/ that this U-shaped pattern existed only for individuals
reporting stomach-related stress reactions. A similar interaction was also found
for tension neck syndrome. The authors ascribed the finding of limited peer
contact to that associated with limited social support, but could only speculate as
to the cause of the high peer contact finding; time-consuming peer contact may be
a stress factor, or extensive peer contact may be a reaction to problematic working
conditions (20).
Bernard and co-workers (21) found associations between Óperceived lack of
support from an immediate supervisorÓ and hand/wrist symptoms and between
Óperceived lack of importance for ergonomic issues by managementÓ and neck
symptoms (Table 14). Other support indicators (interaction with co-workers,
support from families) failed to exhibit associations with musculoskeletal
symptoms, though. In the study by Camerino et al. (32), indifferent colleagues, no
internal meetings, lack of communication and no support from colleagues were all
- in apparently separate analyses - associated with cervical complaints (Table 14).
Faucett and co-workers (47) reported co-worker support being associated with
upper extremity numbness (Table 14). Furthermore, low co-worker support and
increased relative keyboard height interacted as to upper torso stiffness (p<0.05).
Supervisor support and supervisor conflict interacted with both relative keyboard
height and relative seat back height in rather complex ways (see further (47)).
In the study by Hales et al. (79), no associations were reported between
interaction with co-workers or others and muscle disorders. A similar lack of
findings was reported by Marcus and Gerr (125). In contrast, Linton et al. (117)
found associations between the social support index and neck pain and (possibly)
also with shoulder pain. Finally, Levoska and KeinŠnen-Kiukaanniemi (111)
reported an uncertain tendency for work-related social support being associated
with disturbing neck/shoulder symptoms.
90
Table 14. Effect of social support and co-operation on frequency of upper extremity musculoskeletal disorders: Relative risk with 95% confidence
interval (CI) and/or p-value for test of linear trend in exposure-response relationship (E-R).
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study
Group, Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study Group,
Health Outcome
Findings
Aronsson, 1992
(10)
Extreme peer contact;
Nk and Sh sx:
Associated (no data
given)
Bergqvist, 1995b
(20)
Extreme peer contact;
nk/sh sx or dx:
p>0.05 in multivariate
analysis
Arm/hand
(see to the right)
Extreme peer contact;
-arm/hand sx:
-arm/hand dx:
2.1 (1.1-4.1)
4.5 (1.3-15.5)
Bernard, 1994 (21) Lack of support from
managers or super-
visors, Nk sx:
Peer contacs, family
support, Nk or sh sx:
OR=1.9 (1.4-2.4)
p>0.05 in multivariate
analysis
Lack of support from
managers or supervisors,
hand/wrist sx:
Peer contacs, family
support, hand/wrist sx:
OR=1.4 (1.2-2.5)
p>0.05 in multivariate
analysis
Camerino, 1995
(32)
Cervical symptoms:
- Indifferent colleagues,
- No internal meetings,
- Lack of communi-
cations
- No support from
colleagues
Associated (p=0.02)
Associated (p=0.03)
Associated (p=0.03)
Associated (p=0.009)
Faucett, 1994 (47) Less co-worker support;
upper extremity
numbness severity Associated (p=0.05)
(for notes, see end of the table)
91
Table 14. (continued)
Study Neck/shoulder Arm/elbow Hand/wrist
Exposure, Study Group,
Health Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings Exposure, Study
Group, Health
Outcome
Findings
Hales, 1992, 1994
(78-79)
Lack of co-worker
support, elbow dx:
p<0.05 in initial,
but p>0.05 in
multivariate
analysis
Lack of supervisor
support, hand/
wrist dx:
p<0.05 in initial, but
p>0.05 in multi-
variate analysis
Levoska, 1994
(111)
Low social support index,
Nk/sh sx:
Uncertain
association
(p=0.07)
Linton, 1989 (117) Low social support index,
- Nk sx:
- Sh sx:
1.8 (1.0-3.2)
1.6 (0.95-2.8)
Marcus, 1996 (125) Co-worker or supervisor
support, Nk/sh sx:
p>0.05 in
multivariate
analysis
Arm/hand sx
(see to the right)
Co-worker or
supervisor
support, arm/hand
sx:
p>0.05 in
multivariate analysis
Abbreviations: CI = confidence interval; CTD = cumulative trauma disorders, CTS = carpal tunnel syndrome; EMG = electro-
myography; E-R = exposure-response relationship; dx = diagnosis; Ha = hand; hr = hours; L = left; NCV = nerve conduction
velocity; Nk = neck; PE = physical examination; PR = prevalence ratio; OR = odds ratio; r = correlation coefficient, R = right;
RR = risk ratio; Sh = shoulder; SMR = Standardized Morbidity Ratio (by age); sx = symptoms; TNS = tension neck syndrome;
UE = upper extremity; VDU = visual display unit; Wr=wrist
Relative risks (expressed as RR, OR or PR) from multivariate analyses if available, otherwise from univariate analyses.
E-R p-values from Mantel test for linear trend in proportions, statistical testing on regression coefficient, t-test or analysis
of variance. Some raw data estimated from graphics for calculation of odds ratios and confidence intervals.
92
The majority of the studies reviewed above reported associations between
limited social support or some other indication of interaction with others and
muscle problems. Unfortunately, the precise ways in which these types of
variables were ascertained varied considerably among studies, making it difficult
to evaluate the consistency of the findings in more than a general way. In light of
this limitation, the following speculative comment is offered: The U-shaped
relationship between peer contacts and various muscle problems, found by both
Aronsson et al. (10) and Bergqvist et al. (20), raises the question whether
extensive contacts should be seen as only positive. Some other studies asked
specifically for both support and conflict (e.g. (21, 47)). However, closer
examination of the questions used in the two Swedish studies to derive the peer
contact index reveal that they were asked in a positive manner (Ómuch con-
tact...help each otherÓ, Ópossibility to talk with others during workÓ and
Ó...discussing solutions to.....Ó, 10). Another possibility is that extensive peer
contact may reflect the need to discuss adverse work conditions. If so, then peer
contacts may here be linked to other, more adverse conditions. The complex
interactions demonstrated by Faucett and Rempel (47) could conceivably be
explainable in this sense, in that the results may perhaps not reflect a direct
interaction between a low supervisor support and a high seatback height, but that -
in this specific locale - one of these two variables is actually closely related to
another (nonmeasured) variable that could better explain the interaction. On the
other hand, the authors do offer some possible rational explanations of their
findings, such that perhaps lack of results in attempting to improve seating may
deteriorate supervisor support.
Fear and insecurity or job dissatisfaction
No association between job security and muscle problems was found by Bernard
et al. (21). Likewise, job insecurity was not a main factor according to Faucett
and Rempel (47) - although an interaction between job insecurity and low relative
keyboard height was noted. In a study on telecommunication workers, fear of
being replaced by computers was associated with neck (OR 3.0; 1.5-6.1), shoulder
(OR 2.7; 1.3-5.8) and elbow disorders (OR 2.9; 1.4-6.1), while neck, elbow and
hand/wrist disorders were associated with uncertain job future (p<0.05) (78, 79).
Marcus and Gerr (125) found that neck/shoulder symptoms were associated with
lower perceived job security, although the low participation rate (70%) detracts
somewhat from these results. The limited number of studies, and the fact that only
two showed a positive association, although suggestive, precludes any attempt at
generalisation.
For job dissatisfaction, Burt (30) found associations between this variable and
certain MSDs; the adjusted odds ratio was 1.9 (1.1-3.4) for neck, and 2.3 (1.2-4.3)
for shoulder symptoms. Elbow, forearm and handds or wrist were not associated
with job dissatisfaction (p>0.05). The SHARP report (174), in contrast, found no
association between, neck, shoulder or hand/wrist disorders, but an uncertain
93
association (p=0.07) with elbow and forearm symptoms. These findings are both
inconsistent and difficult to interpret, since it is unclear whether job
dissatisfaction should be seen as a cause of, or perhaps consequence of MSDs.
Stress reactions
Aronsson et al. (10) reported dose-response associations between stress reactions
(both Óstomach reactionsÓ and Ópsychological tirednessÓ) and various muscle
problems - although no numerical details were reported. Bergqvist and co-
workers (20) used the same two stress variables, and reported numerous
associations between Óstomach reactionsÓ with different muscle problems;
neck/shoulder discomforts (OR 3.5; 1.5-8.2), intensive neck/shoulder discomforts
(OR 5.4; 1.6-17.6), cervical diagnoses (OR 3.9; 2.0-7.7), any shoulder diagnosis
(OR 4.8; 1.2-10.7), arm/hand discomforts (OR 3.8; 2.0-7.3) and arm/hand
diagnoses (OR 3.4; 1.3-8.4). Furthermore, there was an interaction of stomach
reactions with extreme peer contacts for TNS (see above), and with performing
repetitive movements for at least 20 hrs/week for intense neck/shoulder
symptoms. For Ópsychological tirednessÓ, only cervical diagnoses were associated
(OR 1.9; 1.0-3.5). As already commented on above, stress reactions were
generally not put in the same multivariate model as work organisational/
psychological and social factors, since they could form a link between exposure
(e.g. adverse psychological and social conditions) and the outcome. It may be
more useful to identify this as a psychosomatic indicator of persons who are
developing problems, and would benefit from secondary prevention measures.
A strong correlation was also found between stress and musculoskeletal
symptoms (p<0.03) by DeMatteo et al. (41). In multivariate analyses, Marcus and
Gerr (125) found that neck/shoulder and hand/arm symptoms were associated
with psychosocial job stress in the previous 2 weeks. In contrast, Camerino and
colleagues (32) failed to report any association between cervical complaint and
stress.
The (mostly) strong associations found between stress and muscle problems can
tentatively be seen as an overall verification of the importance of work
organisational, psychological and social factors - especially since such
associations were just as strong with objective verifications of muscle problems as
with questionnaire based endpoints.
94
Gender
Gender as a risk factor for musculoskeletal disorders
Odds ratios for the effect of gender on upper body disorders, as ascertained by
questionnaires are shown, based on data given in or computed from a number of
studies (10, 20, 21, 30, 45, 71, 78, 98, 101, 174), are shown in Tables 15 and 16.
Only three studies failed to indicate a substantial excess of neck and shoulder
discomforts among women compared to men (20, 78, 174). However, it should be
noted that the SHARP study (174) is small, with only 11 male subjects, so that the
higher crude prevalence among women may not have been statistically significant
simply because of sparse data (the odds ratios and CIs were not shown). The
Bergqvist study (20) was performed at workplaces with a high degree of routine
work for both men and women. This implies that observed gender differences
actually may represent differences in type of work. Some support for this
explanatory suggestion is found in both the study of Aronsson et al. (10) and that
of Bernard et al. (21), where smaller gender odds ratios for neck and for shoulder
problems were noted when the study population was limited to men and women
with more routine work, or with more comparable jobs (see also references 1 and
12 in the report by Bernard et al. (21)). However, the opposite trend was found by
Aronsson et al. (10) when examining professional (CAD and/or programming)
work (see Table 15). Furthermore, in an older investigation of the same study
population that was investigated by Bergqvist et al. (20), Knave and co-workers
(104), found higher discomfort scores (frequency and intensity) for women
compared to men in neck, shoulder, upper arm, elbow, forearm and hand.
Although there was no detectable difference between men and women in the
prevalence of disorders in the stidy by Hales et al. (78), female gender had a weak
positive association with increasing score of combined severity, frequency, and
duration for both neck and shoulder symptoms.
Few studies have specifically reported the influence of other factors on these
gender-specific odds ratios, when based on discomforts reported in question-
naires. In the study by Bernard et al. (21), the odds ratio for neck, shoulder and
hand/wrist symptoms were adjusted for a number of job task and psychological
and social factors. (Unfortunately, crude gender odds ratios were not reported, so
it cannot be determined whether these covariates had confounded the association
with gender.) Also of interest are the findings of substantial differences in the
work performed by men and women at the VDU found by Evans (45): men had
greater variety of VDU tasks, used the VDU for fewer hours per day in total, and
worked fewer hours continuously without a break. Among men only, task type
was found to modify the relationships between symptoms and VDU work, with
higher prevalences in data processing than in programming and a more rapid rate
of increase with increasing hours per day at the VDU.
95
Table 15. Odds ratios for upper torso musculoskeletal discomfort in women compared to
men.
Study, Study Odds ratio in upper torso location Comments
author, date population Neck Shoulder
Aronsson, 1992
(10)
General VDU 2.2 (1.8-2.7) 2.8 (2.3-3.6) In the last 12
months, cOR
Routine VDU 1.5 (1.0-2.2) 2.2 (1.4-3.4) In the last 12
months, cOR
CAD or prog.
work
2.9 (1.6-5.5) 3.1 (1.6-5.8)
Bergqvist,
1995b (20)
Routine VDU, 1.3 1, 0.7-2.3) 1.3 1, 0.7-2.3) In the last 12
months, cOR
Bernard, 1994
(21)
General VDU 2.1 (1.4-2.4) 2.2 (1.5-3.3) In the last 12
months, aOR
VDU jobs with
uniform gender
distribution
1.9 (0.8-4.5) 1.5 (0.5-4.8) In the last 12
months, aOR
Burt, 1990 (30) Newspaper workers 2.3 (1.5-3.5) 1.3, p=0.15 In the last 12
months, aOR
(neck only),
cOR for
shoulder)
Evans, 1987
(45)
General VDU 2.6 (2.0-3.3) ND Self-selected,
cOR
Grieco, 1989
(71)
General VDU 3.6 2, 3.1-4.1) 3.6 2, 3.1-4.1) cOR
Hales, 1992,
(78)
Various telecommu-
nication workers
No association
(p>0.05)
No association
(p>0.05)
Karlqvist, 1995
(98)
CAD work 2.0 (1.2-3.4) 3.0 (1.7-5.2) Current, cOR
Karlqvist, 1995
(101)
General VDU 2.6 (1.7-3.8) 4.2 (2.7-6.7) In the last 12
months, aOR
SHARP, 1993
(174)
Billing clerks
(routine work)
OR not given
p>0.05
OR not given
p>0.05
In the last 12
months, aOR
Notes and explanations : 1) Neck and shoulder discomforts reported
together, 2) Upper limb problems. (For other explanations, see bottom of
Table 16.)
96
Table 16. Odds ratios for extremity discomfort in women compared to men.
Study, Study population Odds ratio in extremity location Comments
author, date Elbow Hand/wrist
Aronsson, 1992
(10)
General VDU, 2.3 (1.5-3.7) 2.5 (1.8-3.6) In the last 12
months, cOR
Bergqvist,
1995b (20)
Routine VDU, ND 1.1 1, 0.6-2.2) In the last 12
months, cOR
Bernard, 1994
(21)
General VDU, ND 1.7 (1.2-2.4) In the last 12
months, aOR
VDU work with
similar gender
occupancy
ND 1.7 (0.8-3.6) In the last 12
months, aOR
Burt, 1990 (30) Newspaper workers 1.3, p=0.26 1.2 (0.9-1.7) In the last 12
months, cOR
Grieco, 1989
(71)
General VDU ND 3.6 2, 3.1-4.1) cOR
Hales, 1992,
(78)
Various telecommu-
nication workers
No association
(p>0.05)
No association
(p>0.05)
Karlqvist, 1995
(98)
CAD work 1.8 (0.8-4.3) 1.7 3, 0.7-4.3)
2.1 4, 0.9-4.3)
Current, cOR
SHARP, 1993
(174)
Billing clerks
(routine work)
OR not given
p>0.05
OR not given
p>0.05
In the last 12
months, aOR
Notes and explanations : OR=odds ratios comparing women with men;
cOR=crude OR (no adjustments made for other variables), aOR=adjusted
OR for other variables. ND=not determined or reported, CAD=Computer
Assisted Design. Prog.=programming. Selfselected= study population
self-selected (respondents to questionnaire in newspaper). Current and
Óin the last 12 monthsÓ refer to period of recall of discomforts.
1) Arm and/or hand discomforts. 2) Upper limb problems, 3) Wrist
discomforts, right side only. 4) Hand problems, right side only.
Three studies where data on men and women can be compared have utilised
physical examinations resulting in diagnoses in order to obtain diagnoses (20, 26,
79). Table 17 summarises this information.
Some points are noteworthy:
· The one study by Bergqvist et al. (20) which failed to exhibit gender
differences based on questionnaires now does so (c.f. Tables 15-17; see further
discussion below).
· In that study, both TNS (tension neck syndrome) and (any) arm/hand diagnoses
appear to be influenced by the womens child-care duties (see Table 17).
The largest of these fewer studies, by Hales et al. (79), did not show that gender
was a risk factor.
97
Table 17. Odds ratios for musculoskeletal diagnoses in women compared to men in the
reviewed studies.
Study, Study Odds ratios for Comments
author, date popula-
tion
Tension
neck synd-
rome
Cervical
disorders
Any
shoulder
diagnosis
Any arm/
hand diag-
nosis
Bergqvist,
1995b (20)
Routine
VDU
2.2
(1.0-4.9)
1.3
(0.6-2.7)
5.1
(ND)
7.4
(ND)
cOR
Routine
VDU
6.4 1)
(1.9-22)
ND 7.1
(1.6-32)
5.2 1)
(1.2-2.3)
aOR
Bozi Ferraz,
1995 (26)
Routine
VDU
2.6 2)
(1.2-5.9)
cOR
Hales, 1992,
1994 (78-79)
General
VDU
Neck (p<0.05) and shoulder (p<0.05) associated with
gender in crude analysis, however, gender was not
associated with any of the four upper extremity
disorders in the final model
aOR
Notes and explanations : OR=odds ratios comparing women with men;
cOR=crude OR (no adjustments made for other variables), aOR=adjusted
OR for other variables. ND=not determined or reported.
1) For women with children at home. For women without children at
home, the aOR was not significantly different than for men, for
e.g.
tension neck syndrome, the aOR was 2.0 (0.7-5.6) when comparing
women without children at home with all men. (There was no
discernible difference between men with and without children at
home.)
2) Reported only for all UE MSD (upper extremity musculoskeletal disorders)
combined.
In the study by Bergqvist et al. (20), both symptoms and diagnoses were
ascertained, enabling a comparison between these endpoints to be made. As
already noted (Tables 15-17), differences between men and women in that study
were fairly small in the questionnaire data. In contrast, diagnosed disorders were
from 1.5 to 10 times more frequent among women than men (see Table 18).
As can be seen in Table 18, women (as a group) were considerably more likely
than men to have their upper extremity and neck discomforts ÓconfirmedÓ by a
diagnosis in the relevant region. These data are not consistent with the suggestion
that the difference between men and women should result from a higher reporting
tendency among women, as such an explanation would be consistent with lower
ÒconfirmationÓ rates among women. Some cautionary notes are appropriate here,
though. First, VDU-specific data are available only for this one study; secondly,
this study is the only one not suggesting overall differences in questionnaire data;
and finally, the published data allow only comparisons of group prevalences, not
associations between individuals discomfort and diagnoses ratings.
In most but not all of these reviewed papers, the odds for neck, shoulder and
wrist/hand discomforts appear to be higher for women than men, at the order of
about 2 (possibly slightly higher for shoulder). For elbow discomfort, two studies
98
suggest similar elevated odds ratios, whereas two do not. The exceptions to this
general increase in some discomfort locations are the results of the studies by
Bergqvist et al. (20) and the SHARP report (174) who generally failed to find
indications of such differences in discomforts between men and women, and the
study by Burt (30) and colleagues, who failed to do so for some endpoints.
Among several possible explanations for this, it can be noted that at least two
these three non-positive studies were performed at workplaces with a high degree
of routine work.
Table 18. Comparison of prevalences of discomforts ascertained by questionnaires and
diagnoses obtained through medical examinations in men and women (data from (20)).
Men (n=58) Women (n=189)
Location Discomforts Diagnoses Ratio Discomforts Diagnoses Ratio
Neck 45.7% 1) 32.8% 2) 0.72 54.4% 1) 49.2% 2) 0.90
Shoulder 44.3% 1) 3.3% 0.07 49.6% 1) 14.7% 0.30
Arm/hand 28.1% 1.6% 0.06 30.4% 11.0% 0.36
1) Separation of neck and/or shoulder discomfort prevalences, unpub-
lished data. 2) Sum of prevalences of tension neck syndrome and
cervical diagnoses.
Data on specific causes of this gender difference are very sparse among VDU
use-based studies, but a few indications have appeared that could be seen to
support the suggestions that both work-specific and non-work specific factors
may be involved, such as child care (20). See also the discussion above on routine
vs. non-routine work, and by Aronsson et al. (10). In the one study that enabled
such an analysis, no suggestion for a higher reporting tendency among women
appeared - since the women in that study had higher ÒconfirmationÓ rates than
men.
There have been few opportunities to compare the effects of similar job features
on men and women, in any industry or sector, because there is so much sex
segregation in workplaces world-wide. However, Silverstein et al. (177) found
that women had an increased risk of hand/wrist disorders compared to men, after
controlling for job demands; most of the gender difference appeared to occur in
jobs with high manual forces but low repetition rates, jobs where many more male
than female subjects were employed. However, when the outcome was restricted
to carpal tunnel syndrome by symptoms and physical examination, there was no
longer an effect of gender after including job demands in the multivariate model
(178).
In a more recent study of automobile manufacturing workers, women were
found to have about 1.5 times higher prevalences of upper extremity disorders, by
either symptoms or physical examination findings (153), although there was
substantial effect modification of physical exposure by gender, in which women
had a higher Ôbaseline' risk than men in the lower exposure groups but about the
same risk in the highest exposure group. It also appeared that there was a healthy
99
worker selection effect which was especially predominant among women
employed in the most physically demanding jobs. However, in a one-year follow-
up survey of the same population, the risk of new upper extremity disorders
associated with physical exposures did not vary between men and women.
Gender as an effect modifier
The study by Aronsson et al. (10) also investigated whether the effect of VDU
work on neck and shoulder discomforts differed for men and women (i.e. gender
as an effect modifier of VDU work). For the neck, the male-specific odds ratios
were high for data entry (OR=5.5; 2.1-14.7), data acquisition (OR=3.8; 1.6-9.1),
word processing (OR=3.5; 1.3-2.1) and programming work (OR=3.0; 1.2-7.5),
compared with the reference group (mixed and varied VDU work). For shoulder
problems, male data entry workers had higher odds than male reference workers
(OR=3.4; 1.2-9.0) - as also illustrated in figure 4. For female workers,
corresponding neck and shoulder odds ratios did not differ significantly from 1.0
(they varied between 0.8 and 1.4). Thus, gender was a strong effect modifier, in
that men had higher risks for neck and shoulder discomforts in relation to
variations in work than women. For CAD workers, however, no such trend was
seen. Furthermore, no such tendency was seen for the arm and hand region.
To further illustrate this, it can be seen in figure 4 (which is based on data from
Aronsson et al. (10)), that a/ the difference in shoulder discomfort prevalences
between men and women is larger in the reference group (presumably with more
varied task types) than in the data entry group (with presumed more homogeneous
and monotonous work), b/ women show higher prevalences than men in both
groups, and c/ the difference between the data entry and reference groups is
higher for men than women. The authors speculated that: Óin situations with low
self control of work, high time pressure etc., the differences between men and
women are fairly small - these ÔexternalÕ workplace related conditions dominates.
The differences between men and women tends to increase in groups where
individual (work-related) conditions are allowed to vary. In such situations,
conditions outside of the workplace (Ôdouble workÕ) may tend to ÔkeepÕ the
women at a higher discomfort prevalence.Ó
Higher VDU work related odds ratios for men than women were also found by
Evans (45) for neck/shoulder discomforts (men 3.4; 1.8-6.3, women 1.9; 1.4-2.7);
tendencies in the same direction were also noted by Grieco et al. (71) for both
neck and upper extremity problems in relation to hours per day at the VDU (Table
3). In contrast to this, Bergqvist et al. (15) found an increased risk of hand/wrist
discomforts among women with increased monotony of work during a six year
period (risk ratio of 3.1; 1.2-7.8), whereas no such risk was found among men.
100
0
20%
40%
60%
Reference group
(mixed and varied
VDU work)
Data entry work
Men
Women
Prevalence of shoulder
discomfort
Figure 4. Prevalence of shoulder discomforts in different groups. Data from Aronsson et
al. (10).
Again, the number of studies are too few to enable any definite summary.
However, if it is true that men experience greater neck/shoulder effects of VDU
work than women, even though women have higher background risk, then this has
both substantive and epidemiological implications. The selection of women only
(based on having the highest prevalence of problems) in order to study VDU-
related neck/shoulder problems may actually be misleading, since a high
prevalence might not necessarily indicate a high contrast in the study.
101
Methodologic considerations and inter-
pretation of findings in the reviewed
epidemiological studies
The studies reviewed above are diverse in their study designs, populations,
exposures, and health outcomes. Nevertheless, as a group they contribute
important information bearing on the relationship of upper extremity musculo-
skeletal disorders and computer keyboard operation. The discussion below
focuses primarily on the methodologically strongest studies, excluding those that
had evidence of substantial potential selection bias or confounding by
occupational or non-occupational factors.
Characterisation of ergonomic exposures
In the large majority of the studies reviewed, information on job title or task type
appears to have been obtained from employer records or direct observation. In
over one-third of the papers (19, 20, 32, 81, 84, 88, 93, 95, 102, 137, 142, 144,
150, 163, 166, 168-170, 187, 188), workplace visits were conducted by the
investigators to obtain data on working conditions and exposure to ergonomic
stressors. In several of these, the information available suggested that postural
stresses were observed on only a single occasion or for a brief period (e.g., ten
minutes) per subject. Such limited observations might lead to misclassification if
the postures utilised by operators were not constant over time; however, no data
are available to indicate how likely this is to occur. Two other studies obtained
keystroke or other output data from employer records (26, 78, 79) and six
conducted exposure validation studies on a subsample of workers (21, 22, 30, 47,
78, 79, 95, 101).
The remaining investigations utilised questionnaires to obtain self-reports on
ergonomic exposures; reliance on self-reported data does imply the possibility of
either non-differential misclassification or information bias. There is a body of
literature addressing the validity and reproducibility of self-reported ergonomic
exposure measures and demonstrating that higher levels of exposure intensity
have generally been obtained in occupations and tasks where the exposures were
independently documented to be more severe (e.g., (6, 47, 61, 154, 205)). For
example, the study by Bernard et al. showed that the duration of keyboard use
was described consistently by people in the same job titles. Although all subjects
tended to over-report their keying time, they did so nondifferentially with regard
to their case status (22). Thus, information bias in this measure is unlikely on the
basis of the data presented. Similarly, Faucett and Rempel (48) analysed the
agreement between individualsÕ self-reported time at the VDU and average time
at the VDU observed for each of two job groups. They found that subjects
reported longer durations than observed, although copy editors spent more time at
102
the VDU than reporters by either measure. Self-reported duration was higher than
the observed group mean for younger subjects and those with higher reported
workload, but did not differ by symptom status. Among the 13 individuals
observed, there was a moderate correlation between the two measures (r=0.50,
p=0.08), between the values reported by Bernard et al. (22) and that found by
Burt et al. (30). The papers by Nathan et al. (136) include one showing very good
agreement between observed and self-reported exposures for the entire range of
ergonomic exposures studied, including keyboard work.
Characterisation of work organisational and psychological and social
exposures
In some studies, standardised questionnaires for work organisational and
psychological or social factors were used, enabling a comparison of results
between these studies. For example, the Job Content Instrument was used in the
studies by Faucett and Rempel (47) and by Marcus and Gerr (125) to define
variables such as decision latitude, job demands and social support scales. The
studies by Aronsson et al. (10) and Bergqvist et al. (19, 20) also used identical
questionnaires in order to derive indices for limited work task flexibility, high
work pace, limited organisational influence and limited or extensive peer contacts
(see (10) for a closer description). Similarly, the studies of Bernard et al. (21, 22)
and Hales et al. (78, 79) both used the NIOSH job-stress instrument. With these
and possibly some other exceptions, methods used to ascertain work
organisational and psychological and social appear to be unique for each study,
and - in several studies - not well defined or described.
The interpretation of specific items in terms of work organisational,
psychological and social factors is far from self-evident, however. For example,
limited rest break opportunity (as used by Bergqvist et al. (20)) is (here) placed
among the factors describing the temporal organisation of work. Lack of such
opportunity may, however, also be a result of a high work load and/or insufficient
decision latitude. This example serves not only to illustrate nomenclature
difficulties, but also to illustrate the problem of (indiscriminately) using a number
of interdependent work organisational, psychological and social and stress
variables in the same model, as it may result in insufficient ability to detect
associations (see further e.g. (164)). In some studies, such problems were
addressed by performing a factor analysis on the variables to be used, e.g. by
Bernard et al. (21), by otherwise checking for collinearity (125), or by a priori
restricting variables into appropriate models (20). However, in cross-sectional
studies this concern cannot be eliminated, because the temporal relationships
among these variables cannot usually be determined.
Characterisation of upper extremity disorders
The definition of Ómusculoskeletal disorderÓ varied widely among these studies.
One of the investigations under review reported only nerve conduction velocity
103
measures but not symptoms (137). All of the remaining studies utilised question-
naires to obtain information on musculoskeletal pain and other symptoms; some
of these also involved physical examinations of some or all study subjects, to
validate or elucidate the symptom reports or to separately assess these alternative
endpoints; (19, 20, 22, 26, 78, 79, 88, 105, 133, 144, 146, 166, 174). For example,
one study found that neck and shoulder pain and sensory discomfort was
correlated with lowered hand skin temperature, suggesting a mechanism of
compression of the neurovascular plexus (88). Camerino et al. (32) stated that
their questionnaire had been previously validated with a standardised clinical
examination protocol.
Even among the studies that relied on self-reported symptoms alone, the case
definition ranged from Ófatigue feelings,Ó with no minimum duration, frequency
or severity (144), to symptoms that lasted for more than one week or occurred at
least once a month within the past year and were reported as ÓmoderateÓ or worse
on a five-point intensity scale (22). The former endpoint represents very short-
term effects and provides potentially weaker evidence of musculoskeletal
disorders. However, that particular study was one in which physical examinations
of the shoulder were conducted on a subset of workers, which served both to
validate the self-reported symptom findings and to indicate the presence of
chronic health effects associated with those symptoms.
Again, misclassification of outcome might be a concern with respect to self-
reported symptoms. Since individuals differ in their experience of and threshold
for pain, questionnaire responses (especially to a single question) might differ
among study subjects with similar symptom experiences. Subjective symptoms
are not as standardised as physical examination findings or objective test results
might be.
On the other hand, there is evidence that the available, more objective tests for
work-related soft tissue disorders are not sufficiently sensitive (i.e., that they give
negative results for too large a proportion of affected individuals). Self-reported
symptoms appear to be the most sensitive indicator, especially of disorders still in
their early stages. Furthermore, the use of multiple questions to construct a score
or a composite outcome variable, as was done in many of these papers, e.g. (22,
30, 78, 84, 116, 168-170, 190, 191), produces an outcome with improved
precision and validity compared with the use of a single question to determine
symptoms.
Questionnaire data have been shown in numerous studies to have great utility in
identifying high-risk jobs or job features. It has been shown, for example, that
self-reported symptoms are predictive of seeking medical services for
musculoskeletal disorders (203). In addition, cases defined by symptoms alone
and those defined by findings on physical examination showed extremely similar
associations with the force and repetition characteristics of subjectsÕ jobs;
symptom-based case definitions generally appear to be both unbiased and more
sensitive (e.g., (176, 177, 203). Lastly, there is a strong correlation between the
frequency of symptoms among a group of occupations and the frequency of
104
workersÕ compensation claims and recorded work-related repetitive trauma
disorders in those same occupations (55). In studies where similar methods were
used (22, 78), similar proportions (about 50%) of symptomatic individuals had
physical examination findings, comparable to what has been observed in studies
of other populations (e.g., (27, 177)).
With particular reference to carpal tunnel syndrome (CTS), self-reported
symptoms have been found to have reasonably high validity when compared with
the diagnostic procedure of nerve conduction velocity testing, as well as with
other objective tests (121, 136). In the work of Nathan and colleagues there was a
strong association between the prevalence of Óprobable CTSÓ (based entirely on
self-reported symptoms) and the severity of slowed median nerve NCV for both
1984 and 1989 data, 135). Similarly, the study by Bernard et al. (21, 22) utilised
nerve conduction velocity (NCV) testing, primarily to validate the symptom-
based case definition. The results showed that a subject meeting the case
definition was 43 times more likely to have abnormal median nerve latency than a
subject who was not considered a case. The odds ratios for both symptomatic
cases and abnormal NCVs were strongly associated with the amount of time spent
typing at the keyboard.
In some studies the case definitions may include symptoms that are difficult to
interpret as a single disease entity from the point of view of clinical medicine.
However, the opinion has been advanced that clinical medicine does not yet
provide the necessary diagnostic techniques for these conditions, especially in
their early stages, or offer the appropriate taxonomy (e.g., (121)). Epidemiology
can make a contribution here to the development of an appropriate case definition.
In addition, the symptom complexes shown to be statistically associated with
physical examination findings may prove to be markers for early stages in
pathogenesis. These have value precisely because they offer a greater possibility
of identifying affected individuals before clinical disease has developed and when
secondary prevention will likely be more effective.
Various questionnaire designs have been evaluated with respect to their
validity, reproducibility, and sensitivity to change; standardised questionnaire
items have been proposed (e.g. (109, 110) and the international comparability of
case definitions is increasing. For example, the validated Nordic Musculoskeletal
Questionnaire (NMQ) was used is several studies reviewed here (e.g. (10, 16, 19,
20, 89, 94, 102, 117). Most epidemiologic studies published now provide a clear
and explicit definition of the criteria by which ÓcasesÓ were identified and a
clarification of whether or not this definition should be considered a clinical
Ódiagnosis.Ó Furthermore, the use of questionnaire items to determine case status
is not unique to the study of musculoskeletal disorders; a standardised
questionnaire for the assessment of chronic pulmonary obstructive disease,
developed initially by the British Medical Council, has been widely used by
epidemiologists for a number of years.
105
Temporal sequence of cause and effect
A few studies were designed around changes of exposure and/or disease status;
two of those examined cross-shift and cross-week changes (94, 144), and one
included a nested case-control study with five-month follow-up of both cases and
referents (22). Two fixed cohort studies obtained seven years of follow-up on a
population of office workers (18, 15) and five years of follow-up on a mixed
industrial population (135), respectively, while a third (16) included a one year
follow-up in order to differentiate between immediate effects of computerisation
of work tasks and effects remaining after (later) adjustments. (Unfortunately, this
last study suffered from a large non-response, effectively precluding conclusions.)
With these few exceptions, the studies reviewed here are cross-sectional, thus
evaluating health problems and work situation at the same point in time. As
pointed out by several of the authors (e.g. (47, 79, 125)), this limits the causal
interpretation that can be given the results. Work conditions may have been
changed by the worker attempting to reduce his/her exposures subsequent to the
development of a disorder (see e.g. by Bergqvist et al. (20)). It is arguable that
such remedies could tend to ÒhideÓ an association existing at the point in time
when a disorder appeared. However, it is unlikely that workers who had already
developed musculoskeletal pain would preferentially transfer into jobs with
higher physical exposures such as repetitive motions or awkward postures. On the
contrary, as discussed earlier, it has been demonstrated in several populations that
workers who develop musculoskeletal disorders seek to transfer into jobs that
have fewer physical ergonomic stressors (125, 145, 152, 179), which would again
lead to an underestimation of the difference in disease rates. These problems can
be addressed, at least in part, if cases are restricted to those with onset reported
after first employment in the occupation under study. Of the studies described,
four (22, 30, 78, 84) utilised a case definition restricted to onset of symptoms after
first employment in the study job.
A second caution is warranted due to the possibility that the existence of a
disorder could influence the perception or reporting of work environment
features, possibly leading to inflated associations. Most work organisational,
psychological and social factors are - in principle - subject to such problems, since
they are essentially always derived from the individual«s perception in e.g. a
questionnaire. However, it should be noted that in some studies, also ergonomic
factors were ascertained in this manner. Both problems are in principle avoided in
a prospective study of incident cases, where assessment of adverse conditions is
done prior to (some) individuals developing muscle problem.
Potential confounding
The statistical data analysis was inadequate in several of the studies reviewed
here. About one-half utilised modern, multivariate techniques (stratified or
regression analyses) to control for potential confounding variables and to explore
the interaction or effect modification of several ergonomic factors simultaneously.
106
Several investigators also collected data on and examined (in the statistical
analyses) various non-occupational factors, such as obesity, educational level,
socio-economic status, history of acute injury, outside hobbies and activities
Nevertheless, not all of the remaining studies were likely affected by
uncontrolled confounding. The majority of the investigators controlled potential
confounding by age and gender, at least, either by restriction in study design or in
the analysis. The studies that did not at least restrict or examine gender (16, 26,
41, 46, 70, 102, 105, 137, 138, 165, 189, 188, 191, 197) are weaker and more
difficult to interpret with respect to aetiology, as a result. Some studies, where
subjects were mostly females, or consisted of females to 85% or more, or where
certain comparisons were shown to be unaffected by gender confounding were -
in our judgement - acceptable from this point of view (47, 83, 88, 116, 163, 168-
170, 182). (See also the discussion on page 99-100.)
Potential selection bias
In a number of studies (71, 83, 105, 116, 137, 138, 142, 141, 144, 165, 197), the
authors failed to give any information about how subjects were recruited and what
the participation rate was among the workers eligible for inclusion in the study.
Evans (45), as noted above, distributed a survey through an occupational health
and safety magazine; the study base was undefined and participation level was
impossible to determine. Five studies (46, 70, 81, 183, 180, 182) had participation
rates below 65%; in the study by Karlqvist (101) it was borderline at 67%, and the
authors failed to demonstrate that the participants were not a biased sample of the
entire population (i.e., that they were similar to non-participants in their exposure
and disease status). Faucett and Rempel (47) reported only 56% participation, but
compared participants and non-participants and found that they were similar in
age and job title. In contrast, Starr (189) obtained only 76% participation from
VDU users versus 95% from hard copy workers, suggesting a response bias in the
study population.
Two studies first recruited employers from designated industries to participate
and then sought volunteers from the participating firms (81, 163). The number of
employers that participated was limited, especially in the first investigation.
Possible consequences of this limited employer participation are that subjects may
not be representative of the entire industry in their working conditions or there
may not be sufficient variability in exposures for a stable measure of effect.
However, potential selection bias was not a concern for the second of these two
studies, in which the participation of individual employees was 94% to 99% in
each site.
A major concern regarding cross-sectional investigations is related to the fact
that only active workers are identified for study. This is likely to result in an
underestimate of the effect of exposure, due to self-selection of affected workers
out of employment (the Óhealthy worker effectÓ). With exception of the cross-
sectional part of the Swedish cohort study (19, 20), none of the investigators
attempted to contact workers who had left employment prior to the study,
107
although one (22) attempted to reach employees who were absent for medical
reasons. Because of this potential selection effect, there is a possibility that the
true effects of exposure are greater than that shown by the data in all of these
studies. Indirect evidence of this effect is found in the studies by Bergqvist et al.
(18), as noted above; by Nathan et al. (135), where fewer ÓnoviceÓ workers at
baseline returned at follow-up, and those who did had markedly reduced
frequencies of symptoms and slowed NCVs; and by Starr (187), where age was
slightly higher among VDU operators and was also weakly associated with fewer
symptoms. This selection effect could also contribute to an explanation of the
failure to find associations with duration of employment in several of the studies
that examined that variable (see page 44-48).
108
Intervention studies
Nine intervention studies were identified that evaluated the effectiveness of one or
more ergonomic or work organisational control measures in keyboard operation
(Table 19), two of these also provided analyses of single surveys that were
summarised above as cross-sectional findings. One additional study described a
clinical (treatment) intervention. As in the etiologic studies, various dimensions of
exposure were identified for the interventions that were undertaken. All of these
involved before/after comparisons, but only four studies also included a