Psychosocial factors at work and musculoskeletal disease

Psychosocial factors at work and musculoskeletal disease. Scand J Work Environ Health 1993;19:297-312. The objective of this review is to establish whether the epidemiologic literature presents evidence of an association between psychosocial work factors and musculoskeletal disease. In a hypothetical model it is sug gested that individual characteristics and stress symptoms can modify this relationship. The reviewed studies do not present conclusive evidence due to high correlations between psychosocial factors and physical load and to difficulties in measuring dependent and independent variables. Nevertheless, it is concluded that monotonous work, high perceived work load, and time pressure are related to mus culoskeletal symptoms. The data also suggest that low control on the job and lack of social support by colleagues are positively associated with musculoskeletal disease. Perceived stress may be an in termediary in this process. In addition, stress symptomsare often associated with musculoskeletal dis ease, and some studies indicate that stress symptoms contribute to the development of this disease.

Work-rel ated mu sculoskeletal di sease (in particular back pa in and back disorders) accounts for a large numb er of wor kers' compensation days and disability in numerous countries. In the last decade disability due to mu sculoskeletal dise ase was still increasing, makin g work-related mu sculoskeletal problem s one of the most expensive health problems in mod ern indu strial socie ty. Data from the American Natio nal Center fo r Health Statistics show that the number of people disabl ed fro m back pain increased by 168% between 1971 and 1986 ; thi s increase is 14 times faster than the population growth ( I) .
It is generall y agreed that back pain and other musculos ke le tal disea ses are of multifactorial origin . Many epide mio logic studies have been co ncerned with the relation sh ip between mu scul oskeletal disease and physical load during work . (See, for example, references 2---4.) Some of the se studie s report an association, while others do not. Several authors have presented possible explanations for the sometim es weak or contradictory evidence for the relation ship between physical load and mu sculoskeletal disease. Poor measurement of current and histori cal exposure, lack of a specific diagn osis, and healthba sed se lec tio n in industry are th e most impor tant explanation s. Nevertheless, heavy physical work, prol onged static load , and motor vehicle dri ving are ge nerally acce pted as risk factors fo r back trouble, and repetitive and static work load are co nsidered ris k fac tors for neck and shoulde r trouble.
In addition to phy sical load , several ind ividu al factors (eg, age , previou s symptoms) seem to pose risks for musculoskeletal complaints. It is generally agreed thou gh that work -related physical load in relati on to indiv idua l functi on al capacity can only partially explain the high prevalence of mu sculoskeletal disease . Wal sh et al (5) estimated that, for sy mptoms of the back, the etiolog ic fraction explained by ph ysic al load is only 20 %. Rec entl y the potential etiol ogic significance of psycho social factors to mu scul oskeletal troubl e has recei ved increasing attention. Anal ysis of the rol e of the se va ria bles in the development of musculoskel etal trouble may contribute to the understanding and redu ction of work-related mu scul oskeletal disease and disabil ity. To our knowledge , no review of the epidemio logic literature on this relationship has been publ ished earlier. Since, in our opinion, the studies concerning this topic are too het erogeneous to combine in a meta-analysis, we condu cted a qualitative literature review.
The epidemiologic studies on the relationship between psychosocial variables and musculoskeletal disease are heterogeneou s in study de sign, measur ement of out come, and the psychosocial variables studied. Th ese studies originate from three d ifferent traditions of research : (i) stud ies on the rel ation ship bet ween (wor k-related) stress ors , stress , and sy mptom s of ill health ; (ii) studies on personality and psychologic al disord ers of chronic (back-) pain patients; and (iii) epi demio log ic studies on the determinant s of mu scul oskeletal d isease. The main con cepts used in these research areas are briefly illu strated in the follo win g text.
I. Stress ca n arise whe n people feel unable to cope with the dem and s pla ced upon them. Person al char- Scand J Work Envi ron Health 1993, vol 19, no 5 acteristics such as age, gender, exp erience, ambition, need s, and per sonality influence coping capacity. According to Karasek et al (6) high work demands, lack of cla rity, and confli ct at work are among the rele vant work-related charac teristics which can predict j ob dissatisfaction , work stress, and ill health. The effect of these variables is mod erated by low control over work, poor career development opportuniti es, and poor social support at work. In particular, the combination of high demand s and low control at work is rep orted to be stress ful and is relat ed to adverse health effects. Most rep orts in the liter ature on the relationshi p between stress and adverse health effects have analyzed the rel evance of stress to the development of heart disease, gastrointes tinal problems, or poor subjec tive health in genera l. (See, for example, refe ren ces 7 and 8.) Only very few of thes e studies have investigated the relationship between work stress and musculoskeletal symptoms.
2. The correlatio n between person ality, psychological problems, and chro nicity of musculoskeletal pain has been exte nsive ly studied in the clinical setting (9,10). Although the foc us in this area of research has been on the chronic pain pati ent, the result s of these studies support the opinion that indivi dual psychological capac ity is important when muscu loskeletal symptoms are being dealt with. These individu al factors may not only be important for the exacerb ation or continuation of symptoms, but also for the development of symptoms . 3. In epidemiologic researc h on musculoskeletal disease, individual capac ity has mainl y been interpreted as physical capacity (eg , muscle stre ngth, range of movement), whereas little atte ntion has bee n give n to psychological copi ng capacity or the interac tion between physical load and coping capacity . In this area of research the foc us has mainl y been on mechani cal factors. The mechanical load is partly determ ined by worker organization variables such as speed of work and variation in tasks (eg, pos ture).
When the concepts of these three research traditions are combin ed and applied to the subject of this literatur e review, the following associa tions betwee n psychosocial factors, stress, indivi dual characteristics, and musc uloskeletal disease are suggested (illustrated in figure I): I . Psychosocial factors at work directly influence the mechanical load throu gh changes in posture, movement, and exerted fo rces. (For exam ple, time pressure may increase hurr ied move ments with high accelerations or poor posture.) 2. Psychosocial facto rs at work (demands and factors such as jo b contro l and socia l support) , together with the personal capacity to cope with such facto rs, may increase work-related stress (stre ss symp toms). This increase in stress may (a) incr ease muscl e tone, which may in the long term lead to the development of muscul oskeletal symptoms or increase musculoskeletal symptoms due to some other, yet unknown but specific physiological mecha nism (eg, hormonal path), (b) moderate the relationship between mechanical load and musculoskeletal sympto ms due to enhancemen t of the perceptio n of symptoms or the reduction of the capacity to cope with them. Th erefore the symptoms due to mechanical loading may be prolonged or intensified , or sympto ms of poor health in general, including musculoskeletal symptoms, may increase.
Thus stress may either increase muscul oskeletal signs in itself or increase the perception of symptoms due to other risk fac tors. We do not pretend to present an explanatory model. Figure I serves mainly an illustrative purpose and provides a structure for the discussion of possible assoc iations between psychosoc ial factors and musculoskeletal disease presented in the literature. In tabl e I all of the factors indica ted in figure I are cl ustered into five categories . This division seemed the most appropriate from the practical and theoretical point of view, although it can be argued that some of these variables could as well belong to other categories. For example, de- pressive symptoms can also be regarded as stress symptoms, and job dissatisfaction does not properly fit in any of the categories. In the following description and discussion of the literature, attention is given to each of these clusters.
In summary, the main questions of this literature review are (i) to what extent are the psychosocial factors at work (demand, control, and support variables) related to musculoskeletal symptoms, in particular back, neck, or shoulder symptoms; (ii) do individual (psychological) characteristics influence this relationship; and (iii) what is the role of stress symptoms in this process, and is the relationship between psychosocial factors at work and musculoskeletal symptoms a specific relationship or is it the result of a general detrimental effect of these variables on health or health behavior?

Selection of the literature
The available literature was selected in an automatic search of several data bases, aSH-ROM, CIS-DOC, Psychinfo and Medline, a manual searching through Current Contents (Life Sciences and Social Sciences), Psychological Abstracts, Safety and Health at Work, ILO/CIS Bulletin, and CIS-Abstracts, a screening of recent issues of scientific journals on occupational musculoskeletal disorders and work-related stress, and a check of the relevant references in recent publications on the subject. Initially, all studies analyzing the relationship between musculoskeletal disease and any of the variables in table I were selected. However, studies focusing on determinants of the prognosis of (chronic) back pain in patients (12)(13)(14) were not included, since the main objective of this review is to investigate the influence of psychosocial variables on the development of musculoskeletal problems, not its influence on prognosis. In this presentation the emphasis is on work-related psychosocial factors, and therefore the influence of psychosocial factors in the nonwork situation, such as emotional life events, problems at home, or living alone, are not discussed. These variables are included in a more extensive report on the subject (15). Descriptive studies that did not compare the results of exposed workers to a reference population of non-or less-exposed workers were also excluded. No other exclusion criteria were applied.

Presentation of the literature
Because there are no a priori hypotheses on the similarity of the relationships between psychosocial factors and symptoms of each musculoskeletal site, the studies on symptoms of the back, the neck and shoulder region, and the musculoskeletal system (several or all sites combined) are discussed separately. The results of cross-sectional and longitudinal studies are also presented separately, since the results of the cross-sectional studies can suffer from bias due to differential health-based selection and do not provide information on the temporal relationship between events. In addition, comments are made on the study population and whether the study dealt with potential confounders (in particular physical load). The presentation of the longitudinal studies includes mention of adjustment for a history of musculoskeletal pain at the start of follow-up. A more extensive description of the literature is presented in another report (15).

Back pain and back disorders
All of the studies relating the variables of table I to back trouble are presented in table 2 (chronological order). The cross-sectional epidemiologic studies on back pain and back disorders were heterogeneous in design, outcome variable, independent variables, and the way potential confounders were treated. Most of the studies on the working population dealt with work-related variables, either self-reported psychosocial factors during work or perceived stress during work (16)(17)(18)(19)(20)(21). Few of the population-based Table 2. Summary of the associations between back pain (BP) and back disorders (BO) and psychosocial factors at workdemands and control (A), psychosocial factors at work -social support (B), individual characteristics (C), stress symp toms (D), phys ical and behavioral health indicators (E). (+ = posi tive association , + + = positive assoc iat ion after adjustment for confounder, + /-= conflicting data , -= no assoc iation; 0 = not stud ied , LBP = low-back pain, HNP = hernia nuclei pulposi) cro ss-sectional studies dealt with work-re lated variabl es (22)(23)(24)(25)(26). The majority wa s co nce rn ed with the relationship bet ween back trouble and variables such as social cla ss and educational le vel (27)(28)(29), psychological and emotional problems, personality traits (27)(28)(29)(30), or parameters of po or health (24)(25)(26)(27). In most of the studie s only questionn aires were administered to assess back trouble (16- ( 19,26,29). Se veral cross-sectional stud ies d id not adjust for co nfo unding fac tors ( 16, 17, 20, 27, 30,31 ,34) or only for age (2 1,29). Onl y eig ht stud ies took age and other co nfo unders, including physical load, into consi dera tion ( 19, 23--26, 28, 32, 35). Information on the strength of the association was present ed in about half of the crosssectional studies (16,19,21,26,28,29,32,33). It can be co ncl uded, the refore, that the most informative cross-sectional studi es are the mult ivariate studies in vest igating the relation sh ip bet ween wo rk-related psychosocial variables and back trouble (other than " Have you e ver had back pain ?" ) after adju stment fo r physical load ( 19,26,32). 300 Although the longitudin al studies are definitel y more homogeneou s in their measurement of dependent and independent variables than the cross-sectional studies, still man y differences exi st. The duration of foll ow-up varied from one year (29,36,37) to almost lifelong (29,38,39). The measurement of back pain included self-reported pain in the lower back (29,(36)(37)(38) or self-reported sciatica (40), episodes registered at the medi cal service (4 1), hospital ad mitta nce du e to hern iated disc or sciatica (42,43), or disability due to back pain (44). Except for so me of the ea rly studies (29,37,38), potenti al confounders were dealt with in the design or ana lys is in all of the longitudinal studies. Th e emphasis was on personal psychological and emotiona l problems (29,36,37,(40)(41)(42)44), stress sy mptoms (34,40), experienced health (36,37,42,44), or utilization of medical servi ces and demograph ic variables (36,37,41,42). In only a few studi es were self-reported work-related psychosocial variables, such as satisfaction with the job (36, 38, 4 1), soci al relat ions at work (4 1) and job dem and s (44), studied . Almos t all of the studies also present informati on on the inde pendent fac tors mea sured before the development of sy mp-toms. Strength of the association was presented in most ofthe longitudinal studies (29,36,39,(40)(41)(42)44). Table 3 indicates, for each of the five clusters of table 1, which of the cross-sectional and longitudinal studies reported a positive association between one of these variables and back trouble and which studied that relationship but did not report a positive association. These associations are discussed next.
Psychosocial factors at work -demands and control. The evidence for a relationship between back trouble and work demands is contradictory. In the cross-sectional studies a positive univariate association was observed between back trouble and selfreported high demands on concentration and possibilities to take a break in some occupations but not in all (16). Other studies also observed such an association, but it did not remain statistically significant after adjustment for other variables (24). A univariate relationship between back pain and high responsibility has been reported for bank employees but not for other occupations (16), and not in another study after adjustment for confounders (19). After adjustment for potential confounders, including physical load and previous back pain, one of the largescale cross-sectional studies reported that a sum score for monotony, time pressure, and worry about mistakes was related to back symptoms established in a physical examination (odds ratio 2.0) (26). A sum score for high time pressure, mental strain, and monotony was also associated with back trouble (questionnaire) in another large-scale cross-sectional study (odds ratio 1.7) (33). However, this association was not adjusted for physical load. The importance of monotonous work is supported by other researchers, who reported an age-adjusted association between this variable and self-reported back pain (24).
As far as time pressure is concerned, perceived time pressure during work is related to perceived back load during the same period (45). The positive relationship between high time pressure and back symptoms (26,33,45) was not supported by two cross-sectional studies, which analyzed symptoms (32) and clinical signs (19) and also adjusted for physical load and other confounders. Neither did the work-related stressors measured in the study of Astrand (19) (time pressure and high responsibility) predict future disability due to back disorders in the same population (44). None of the other longitudinal studies investigated the relationship between job demands and back trouble. Only one of these studies investigated the influence of both job demands and control variables (24). These variables were not significantly related to back trouble after adjustment for confounders.
In summary, although contradictory results were reported, the cross-sectional studies yielded some evidence for a relationship between back trouble and aspects of self-reported work demands, in particular monotonous work. The evidence for the effect of working under time pressure is contradictory. The relationship between back trouble and other aspects of job demands (demands on concentration, responsibility, poor career prospects and possibility to take extra breaks) is not clear due to the few data of high quality. In the only longitudinal study that analyzed job demand variables, the variables did not predict future disability due to back disorders. Table 3 Summary of the associations between neck pain (NP), neck disorders (ND), and shoulder pain (SP) and psychosocial factors at work -demands an? control (A).~sychosocial factors at work -social support (B), individual characteristics (C), stress symptoms (D), and physical and behavioral health indicators (E). (+ = positive association, + + = positive association after adjustment for confounder, +1-= conflicting data, -= no association, 0 = not studied) Psychosocial fa ctors at work -social support. Low social support by colleagues and superiors without adjustment for confounders was associated with back pain in one cross-sectional study (17). A sum score for the psychosocial work environment (several questions on work content and social relations at work) was positively associated with back trouble (no adjustment for physical load) in two studies (20,21). However, in a similar study no signific ant association between self-reported poor psycho social environment and back trouble was observed after adjustment for other variabl es, including physical load (32). One study even observed a positive association between a good relationship with colleagues and troublesome back pain (18). In a prospective longitudinal study workers with poor social relation s at work more frequently reported an episode of back pain to the medical department or filed a claim in the next three years (41). This situation also applied to workers without back pain episodes at the start of follow-up. In the analy sis adjustment was made for physical load and other potential confounders. A similar association between poor social support by superiors and episod es of back pain reported to the medical office was previou sly found in a retro spective study by the same authors (46). They stated that these variables may have been related to the back pain episodes, as well as to the readin ess to report them to the medical department or file a claim .
In summary, poor social support at work was associated with the incidence of reported back trouble in the only relevant longitudinal study. The results of the few cross-sectional studies that analyzed the relation ship betwe en social support at work and back pain are contradictory. None of the studies on back troubl e investigated the effect of demands, control , and support variables simultaneously.
Indi vidual characteristi cs. Almo st all of the crosssectional studies reported a positive association between psychological or emotional probl ems and selfreported back pain (19, 22, 24, 25, 27, 29, 30, 32, 42, 47 , 48 ) and "back disorders" diagnosed in a physical examination (19,26). One cross-sectional study reported no association between back episodes registered at the medic al service and psychological problems or nervousness (35). Two prospecti ve longitudinal studies found a positive association between certain personality traits, measured with the Minnesota Multiphasic Personality Inventory (MMPI), and episod es of back pain reported to the medical service or insurance company (odds ratio 1.4) (41) and self-reported back pain (odds ratio 2.8) (29). Thi s situation also applied to those respondent s not reporting previous back pain at the start of follow-up. Psychosocial distres s was also related to hospital admittance due to a herni ated disc 10 years later (relative risk 1.7) (42). Neuroticism, howe ver, did not predict disability due to back pain (44), although it showed a strong cross-sectional association (odd s ratio 2.8) in the same population (19) . In addition neurot icism was not associated with sick leave due to back trouble (22). All of the cros s-sect ional studies that investigated the influence of type A behavior (competitiveness) on back trouble reported a positive association (31,49,50). However, adjustment for physical load was lacking in these studies. Variable s of a different category of individual characteristics such as social class and education are highl y intercorrelated and are also correlated with physical demands at work. (See reference 28.) Therefore, a correlation between these variables and musculoskeletal trouble is very likely biased by differences in work-related physical load. Low educational level was associated with back-pain prevalen ce in several cross-sectional studie s (19,(27)(28)(29)51), but not in all (25). It is remarkable that in one study variables such as low income and little education were related not only to back pain, but also to back abnormalities, established in a physical examinat ion, the latter association bein g even stronger (19). These variables did not, howe ver, predict later disability due to back problem s (44). Two other longitudinal studies (39,42) tentati vely supported a relation ship between low educational level or low social class (relative risk 2.5) and back trouble at a later age after some adjustment for physical load during work.
In summary, personality trait s and emotional problems are associated with back trouble in both crosssectional studies and longitudinal studies. The results suggest that these variables also predi ct back trouble for respondents who have not previously experienced back pain . The result s, howe ver, were not consistent, and the se studies investigated a variety of different dependent variables. Few studies investigated the association with low education or low social class while adjusting for physical load at work. A tentativ e association between the development of back pain and these variables independent of physical load at work was reported by two longitudinal studies.
Stress symptoms. An association betwe en back trouble and stress symptoms has been reported in several cross-sectional studies (16. 24-27, 29, 30, 32, 33), but not in all (35). Feelings of stress (nervousness, tension , sleeping difficulties) were univariately associated with back trouble in several studies (27,29,30). This association was also significant after adjustment for other risk factors (24,25,32,33) with a reported odds ratio of up to 3.5 (33). Worry about making mistakes (combined in one score with time pressure and monotonous work ) was positivel y associated with "obj ectively" assessed back pain (26). However, perceived demands and tension or physiological indicators of stress symptoms did not relate to simultaneously experienced back load (45). Although severa l studies reported an association between exhaustion and fatigue and back trouble, this association disappeared after adju stment for other risk factor s, including physical load (24,25,35,36).
In the longitudinal studies, stress symptoms also increased the risk of symptoms of back pain (29) or sciatica (43,40), although the odds ratios were close to one and not significant for all groups. Thi s situation also applied to those respondents without a history of sciatica or low-back pain before the start of follow-up. In three cross-sectional studies, dissatisfa ction with work conditions was significantl y associated with self-reported symptoms of the back and sick leave due to back trouble without additional adju stment for confounders (14,16,17,22,24,29,36,38,41). However, after adjustment for confounder s, a positive association between job dissatisfaction and back troubl e was reported by one cross-sectional study (25) and one longitudinal study (4 1) (odds ratio 1.7), but not by three cross-sectional studies (19,24,32) or one longitudinal study (36).
In summ ary, the association betwe en back trouble and stress symptoms reported in several cross-sectional studies seems to be tentatively supported by the result s of longitud inal studies. Th is situation applies to self-reported back pain , as well as to signs and symptoms of the back in a physical examination.
The results on the relation ship betw een job dissatisfaction and (self-reported) back troubl e are not consistent.
Physical and beha vioral health indicators. Several cross-sec tional studies have reported a positive association between perceived poor health and self-reported (19,27,29) or clini cally assess ed symptoms of the back (19). This variable also predicted future disability due to back trouble after adjustment for other risk factors (relative risk 3.4) (44). In one of the longitud inal studies, a similar relation ship was observed, but it was no longer significant when all of the other variables were taken into account (36). This latter study reported that "rumbling of the stomach" in women showed a signifi cant correlation in a multivariate analysis with back troubl e. Epigastric pain proved to be an indica tor of first-time occurrence of low-back trouble. The authors stated that these (psyc hosomatic) symptom s also showed a correlation with back trouble in other studies (12,13,22). Diagnosed respirato ry disease was, after adjustment for confounders, associated with self-reported symptoms of the back (5 1), with clinically assesse d unspecified low-back pain (26), and with back trouble after one year (52), but not with back trouble in severa l other studies (19,24,36) . Respiratory disease was also a predictor of hospitalization among men due to a herniated disc and sciatica after 10 years (42). In other studies clinicall y diagnosed cardiova scular disease was associated with clini cally assessed unspecified low-back pain after adjustment for confounders (26), but not with self-report ed back trouble ( 19,24,36) .
In summary, in several cross -sectional and longitudinal studies, an assoc iation between back trouble (both self-reported and clinically assessed symptoms) and other symptoms of poor health was observed. However, in several cases, the association was not significant after adj ustment for other fact ors. Some studies report ed no association at all between back troubl e and other indicators of poor health .

Symptoms and disorde rs of the neck or shoulders
We found 13 cross-sectional studies and five longitudinal studie s on the relationship between psychosocial factor s at work or stress and symptom s of the neck or shoulder region (  (58), or symptoms and signs reported in a physical examination (49,55,59). The analysis of each of the cross-sectional studies concentrated on self-reported work characteristics. In several studies the association between psychosocial factors at work and symptoms of the neck was determined after adjustment for some aspects of postural load (39 , 53, 54, 59). The longitudinal studies are very different in desi gn. The association between psychosocial factors and neck and shoulder symptoms are discussed for the same five clusters of table I.
Psychoso cial fac tors at work -demands and control. In most of the cross-sectional studies a relationship betwe en sever al job demand variables and symptom s of the neck and shoulder region are reported (table 3). Only some of the result s are conflicting. High responsibility did not have a direct significant influence on the prevalence of shoulder or neck complaints in one study (58) . The other studies provided no data on this variable. In only one study was no relation observ ed between symptoms of the neck or shoulder and the measured j ob demand variables ( 17). All of the other cross-sectional studies that investigated this relationship reported a positive association between neck or shoulder pain and jo b demand variables such as monotonous work, time pressure, poor work content, and high perceived work load (20, 2 1, [58][59][60][61]. Several studies also reported a positive association between the symptoms of the neck or shoulder region and a combination of job demand variables (33,62,63) or of job demand and control variables (20, 59-61 ), However, onl y one cross-sectional study adjusted for phy sical load (62). Some other studies indicated that the combined effect of variables for psychosocial factors and physical load was larger than the individual effects (21,61). The reported odd s ratio for the relationship between demand vari ables and neck and shoulder symptoms was 1,2 after adjustment for phy sical load (62) and 2.7 after adju stment for occupational clas s only (33). The reported odds ratios for the effect of control at work were 3,3 (20) and 3.9 (59) . Howe ver the se point estimates of the odds ratios are hard to compare due to the large var iatio n in mea surement and treatment in the analy sis of these variables. Onl y two of the longitudinal studies investigated the relationship between job demands (time pre ssure) and symptoms of the neck and shoulder (56 , 64). Both reported a positive association, after adjustment for postural load.
In summary, as the majority of the cross -sectional studies reported a relationship between psychosocial variables and symptoms of the neck or shoulders, an asso ciation seems likel y. Variables such as monotonous work, time pre ssure, poor work content, and high work load seem to be important. The longitudinal studies that con sidered this rel ationship reported a positive relationship between time pressure at work and neck pain .
Psychosocial factors at work -social support. Data on the influence of social support on neck or shoulder symptoms arc not con sistent. One study observed no effect of social suppo rt by colleagues or superiors (17 ), while others reported a po siti ve association for support by coll eague s (20) , superiors (2 \) , or a combination of the two (63). In addition, some reported an effect of supp ort by colleagues but not of support by superiors (59 , 64) or vice versa (58). All of the studies relating psychosocial work environment (co mbining both demand and supp ort variables) to the prev alence of neck or sho ulder pain reported a positi ve assoc iation (20,2 1, 56 ).
In summary, the sources on support by colleag ues or superiors pre sent contradictory dat a. Yet there is evidence that high demands in combination with these variables, in addition to the influence of ph ysical load at work, increase the prevalence of neck or shoulder symptoms.
Individual characteristics. In one cross -sectional study a relationship was found between emotional pro blems and symptom s of the neck or shoulders (6 1). However, depressive episodes in the year be-304 fore employment were not associ ated with the development of symptoms of the neck or shoulders in a new job (64) , Some results suggest a relationship between the mental resources of people to promote their own health (poor sense of coherence) and neck and shoulder symptoms one year later (39). The two studies on type A behavior in relation to neck or shoulder symptoms reported a positive associ ation (49,50). Social cla ss was not sign ificantly related to neck and shoulder sy mptoms after adj ustment for phy sica l demands (53) . After adju stm ent for several confounders, including bent and twi sted posture, low education was po siti vely related to the pre valen ce of neck and shoulder symptoms (54) . Ho we ver in the longitudinal studies no clear relation ship between low education (at childhood and later on) and symptoms in middle age was establi shed (38,39).
In summary, few studies analyzed the relation between psychological problems and neck or shoulder symptoms, and the variables investigated are diverse (emotional problems, depressive symptoms, and attitude towards own health). Th e scarce data on neck or shoulder sympto ms in relation to social cla ss and ed ucation do not sugg est a strong relationship, Stress symptoms, In several cro ss-sectional studies a positive asso ciation between neck trouble and perceived stres s or stress symptoms was observed (33, 57 , 58 ). High percei ved stress also predicted development of sy mptoms in two of the longitudinal studies (56,64). One of the studies reported a stro nger relationship of stress sympto ms with neck pain than with back pain (relative risk of 2,7 and 1.7, respectivel y) (33) . lob dissati sfaction was cross-sectionally rel ated with physical findings of neck and shoulder trouble after adjustment for confounders (54). Thi s relationship was also true for shoulder pain in women, but not in men (55), This vari able did not however predict neck or shoulder sy mptoms one yea r later (39).
In summary, according to the literature, a relation between stress symptoms or percei ved stress at work and symptoms of the neck or shoulders seems likely. No con sistent relation bet ween job dissat isfaction and neck or sho ulder trouble wa s observed.

Physical and behavioral health indicators. A rela -
tionship with severa l other non spe cific symptom s of poor health and ind icators of health behavior was obs erved in the only study that concerned th is relation ship (62) .

Musculoskeletal trouble, locations other than the back or neck and shoulders only
Tabl e 4 summarizes the results of studies with a rather di verse set of end point s. Included are studies on the relat ionship between psycho social factors and a combination of back, neck, and sometimes upperlimb disorders (50,65,66), all types of musculoskeletal signs or sympt oms (67,68,69), and more general muscl e pain (6,66,(70)(71)(72). In the crosssectional study of Ursin et al (70), feelings of anxiety, coping styles, and j ob stress were analyzed in relation to muscle pain. The other cross -sectional stud ies have analyzed the relat ionship bet ween psychosocial factors at work (dem ands, control, and support variables) and musculoskeletal symptoms (6, 65-69, 71, 73). Most of these latter studies adj usted for physical load or conducted the analysis within one, homogeneously expose d group (65)(66)(67)(68).
The one longitudinal study (68,74) concerned musculoskeletal symptoms determined with a questionnaire and with a physical examination in relation to a sum variable for stress symptoms, perceived psychoso cial stress (at work and at home ), and soc ial support (both at work and outside work). Table 4 show s the results of the study for each group of variables shown in table I.
Psychosocial fa ctors at work -demands and control. In almost all of the studies describ ed in table 4 a positive relation betwe en one or more psycho social factors at work and musculo skeletal symptoms was reported. The followi ng job demand variables were related to the prevalence of musculo skeletal trouble : monotonous work (7 1), time pressure (67,69,73), high demands on conc ent ration (65), and high perceived (mental) work load (6,65,66,69). Thi s last relationship was confirmed in the longitu- Health 1993, vol 19, no 5 din al study of Leino (74). However no significa nt influenc e on musculoskeletal trouble was observed for possibilities to take a break (65 ) and poor work content (69) . In contrast to the other studi es, time pressure was not significantly related to musculo skeletal trouble in one study (66). In this study, adju stment for physical load co nsiderably decreased the association between time pressure and symptoms of the back, neck, or shoulders. The variabl es job demands and lack of po ssibiliti es to talk were not only related to back, neck, and shoulder disorders, but also to self-reported muscle tension (66) . Low job control was positively related to musculo skelet al symptom s in all of the studies (6, 65, 66, 71) except one (69) . However the exception report ed that only little variation in this variable was present : autonomy was low for most of the workers. Although seve ral of the studies on psycho social factors and musculoskeletal trouble analyzed the influence of both jo b demand and support or co ntrol variables, each of these variables was treat ed as an independent variable, and no inform ation on the interaction between them was presented.
In summary, monotonou s work, time pressure , and high perc eived work load each seem to be related to musculoskeletal problem s. Almo st all of the studies stressed the importance of little control ove r one' s j ob.
Psychosocial fa ctors at work -social support. Althou gh most studies have observed a relation ship between musculoskeletal trouble and poor social support at work by colleagues (65,69,71), this relationship was not consistent in all studies . Two reported no associa tion between social support at work and back, neck, and shoulder symptoms and symptoms of other joi nts, regardle ss of adju stment for physi- Table 4. Summary of the assoc iations bet ween mus culoskeletal disease not specified and psychosocial factors at wor kdemands and control (A), psychosocial factors at work -social support (B), individual characteristics (C), stre ss symptoms (D), physical and behavioral health indicators (E). (+ =positive association, + + =positive association after adjustment for confounders, + 1-=conflicting data, -=no association; 0 =not studied , NP =neck pain, BP =back pain, LBP =low-back pain, SP = shoulder pain)  cal load (6,66). Poor social support by superiors was also positively related to musculoskeletal problems in most studies (66,67,69,71), but not in all (73), or only among women and not among men (6). For blue-collar women , a significant cross-sectional association between poor social support and base-line musculoskeletal findings was reported in a longitudinal study (68) . Poor social support was also associated with musculoskeletal symptoms and signs during the follow-up , but this association was not significant.
In summary, most of the cross-sectional studies reported that high job demands, low control , and poor social support were associated with self-reported musculo skeletal trouble. This was also the situation for those studies that adjusted for physical load. However the association did not consistently apply to all of the variables or all of the respondents.
Individual characteristics. Anxiety and depressive symptoms were positively related to musculoskeletal symptoms (70). An association between type A behavior and musculoskeletal trouble was also reported (50,66). Only one study analyzed the association between low income (low social class) and musculo skeletal trouble, and it reported a positive association (6).
In summary, a relationship between psychological and emotional problems and musculoskeletal symptoms was reported in the few studies that investigated this relationship.
Stress symptoms. In one of the cross-sectional studies a positive association between various stress symptoms (anger, worry, fatigue, sleep disturbance s, loss of appetite) and self-reported muscle tension was reported (66). Muscle tension was in turn associated with back , neck , and shoulder symptoms, as were other types of tension (breath tension, chewing tension, and type A tension), but not several physiological parameters (systolic blood pre ssure and plasma cortisol). No direct associations between stress symptoms and back, neck, and shoulder symptoms were reported in this study. Perception of the work environment as stressful was related to symptoms in one cross-sectional study (71) but not in another (70). Leino (74) reported that a sum score for self-reported stress symptoms was related to self-reported base-line musculo skeletal symptoms and clini cal findings . In addition , the base-line score for stress symptoms was related to the clinical findings after 10 years of follow-up. On the other hand , the baseline musculoskeletal symptoms were also related to stress symptoms after 10 years of follow-up . This re-306 lationship was not, however, equally consistent for all of the groups as the reverse relationship was. Thus this study presented some evidence that stress symptoms were not only related to self-reported musculoskeletal complaints, but also to adverse changes in clinical symptoms and signs after 10 years of followup, also for those respondents without musculoskeletal symptoms at the start of the study. It also showed that, for perceived psycho social stress, a sum score that contains items on work-and nonwork-rel ated psychosocial factors (overstrained by work, pressed work pace, mentally strenuous work, financial problems, trouble with kids or relatives, fear of making mistake s) was cro ss-sectionally related to musculoskeletal symptoms and signs among men only in the base-line measurements (68). However the initial perceived stress did not predict symptoms after 10 years of follow-up but did predict musculoskeletal findings among blue-collar workers.
Only one study analyzed the association between job dissatisfaction and musculoskeletal complaints and reported a positive association (71). In the study of Karasek et al (6), job (dis)satisfaction was treated as a dependent variable, and no information on its associat ion with musculoskeletal trouble was presented.
In summary, a relationship between stress symptoms and the development or exacerbation of musculoskeletal symptoms established by interv iew and physical examination seems likely . Muscle tension may be an intermediate of this process. Only few data are available on the relation between job satisfaction and musculoskeletal trouble.

Physical and behavioral health indicators.
Only two studies reported on the association between other health effects and musculoskeletal disorders. Both poor health in general (67) and psychosomatic symptoms such as gastrointestinal trouble were related to musculoskeletal trouble (66) .

Discussion
Quality of the studies We found 44 cross-sectional and 15 longitudinal studie s reporting empirical results on the subject of this review. Only 29 cross-sectional and three longitudinal studies specifically analyzed the influence of psycho social factors at work. The other studies dealt with the relationship between factors such as stress symptoms or psychological problems and musculoskeletal symptoms. Not all of the reviewed studies were of high quality. Only 22 adjusted in some way for physical load. In nine the outcome was established after a physical examination.

Summary of the resu lts for all sites combined
As the presented results do not appear to be essentially different for each musculoskeletal site, table 5 presents a summary of the relat ionship s presented in the Result s section, when the results for back , neck and shoulder, and general musculo skeletal problems are combined. Due to several shortcomings , to be discussed later on, the reviewed studies do not pres ent conclu sive evidence. Nevertheless, a qualitati ve summary evalu ation of the evidence for the various relationships is presented in the following.
Psychological fa ctors at work -demands and control. Althou gh the results on time pressure are conflicting, the epidem iologic studies support a relationship between monotonous work, percei ved work load, and work under time pressure on one hand and musculoskeletal trouble on the other. The studies Scand J Work Environ Health 1993, vol 19, no 5 dealing with neck or shoulder sympt oms or muscle pain empha size the importance of control at work.
Psychosocial f actors at work -socia l support. Several studies, on both back disorders and musculoskeletal troubl e, lend support to the rele vance of social support by colleagues at work, although the result s are not consistent. In addition, a combination of job demands and support was consistently related to musculoskeletal trouble.
Ind ividual characteristics. Several emot ional and psycholog ical probl ems arc related to musculoskeletal trouble. In the cross-sectional studies these problems were either a cause of musculo skeletal symptoms or a result of them. Two longitudin al studies showed that some personal ity traits predict musculoskeletal trouble , whereas two others did not support this finding. Therefore the role of these varia- Table 5. Summary of the epidemiologic evidence for t he rela ti onship between psychosocial factors, person al characterist ics, st ress, and musculoskeletal disease. (+ =positive evidence for an association, -=association absent , + / -=conflict ing data , ? =too little information, BP =back pain , NP/SP =neck or shoulder pain , MP =symptoms of the musculos keletal system, no location spec if ied (neck or shoulder and back or all musculoskeletal sites)) Stress symptoms. Some of the studies that measured stress before the onset of symptoms and adj usted for other risk factors point to a role of stress in the development of musculoskeletal trouble.
Physical and beha vioral indicators of health. The data suggest that back and neck trouble are often accompanied by other symptoms of poor health. Several of the reported result s were conflicting. The differences may be due to the heterogeneous character of the studies with respect to the measured independent and dependent variables.

Independent variables
Many different methods were emplo yed to measure the independent variables. Thi s statem ent applie s to the measurement of psychosocial factor s at work, stress symptoms, and personal characteristics. In addition the variables included in the category personal characteristics, for example, emotional problems and personality traits, are very diverse. It is therefore difficult to draw any overa ll conclu sion s.

Confounding variabl es
At the workplace a high correlation often exists between psychosocial factors and physical load. In assessing the importance of psycho social variables for musculoskelet al symptoms, it is important to adjust for physical load. Many of the reviewed studies failed to do so. All of the studies that appl ied some adj ustment for physical load, except one (45), relied on self-reports for the assessm ent of physical load. Several publications have shown that the validity of self-reported physical exposure is questionable (75,76). In several studies within a specific occupation (eg, operators of visual displ ay units or health care workers) high mutual correlations between self-reported psycho social factor s and physical load ca n make it difficult to disentan gle the effec t of each of these variables. Due to the limited assessment of physical load, none of the reviewed studies present s conclusive evidence of the role of work-related psychosocia l variables or stress symptoms in musculoskeletal disease.

Outcome
Few studies conducted a physical examination for the assessment of the outcome variable. When both independent and depend ent variables are self-reports, a relationship between these variables can arise from 308 general dissati sfaction or readiness to repo rt complaints. The same is true for the relation ship between self-reported work-related psychosocial variables and reported claims for disability (41). Therefore the relationship between self-reported psychosocial factors and self-reported musculo skeletal symptoms or claim s is expected to be stronger than the relationship between these variables and symptoms and signs established in a physical examination. Several studies not only showed an association between these variabl es and self-reported musculo skeletal complaint s, but also between these variabl es and physical finding s (19,63). Toomingas et al (63) reported a similar association between j ob demands and selfreported symptoms of the neck or shoulder region (odds ratio 1.5) and tenderness by palpation (odds ratio 1.9). This association applied even more strongly to the relationships between these outcome variables and social support at work (odds ratios 1.7 and 5.0, respectivel y). Some longitudinal studies have also reported that psycho social factors and low jo b satisfaction predi ct future musculo skeletal problems after adjustment for back trouble at the start of fol-lOW-Up. These data showed that not all of the reported results can be fully attributed to spurious associati ons due to the fact that most of the studie s relied on self-reports for both the dependent and independent variables.
It can be hypothesized that psychosocial factors, stress symptoms, and individual characteristics are more strongly related to nonspecified back pain than to back disorders, such as a herni ated disc. In the stud y of Helio vaara et al (26) the odds ratio for the relation ship betw een psychosocial factors and unspecifi ed low-back pain (odds ratio 1.4) was une xpectedly lower than that for the relationship between psychosocial factors and self-reported sciatica (odds ratio 2.0).

Specificity of the f indings f or each musculoskeletal site
Some of the studies reported odds ratios of similar magnitud e for the relati onships between musculoskeletal symptoms of the different body regions and psychosocial factors. The cross-sec tional study of Katilainen (33) show ed an odds ratio of 2.7 for the relationship between a sum score for mental load (distra ctions, tight time schedule, mental strain, monotony, and workpace) and neck and shoulder symptoms and 1.7 for back symptoms.
Recently, strong associations have been reported between back symptoms or neck symptoms and osteoarthrit is of the knee, hip, or hand (odds ratio of 5.3 and 1.6, respectively) (26,62). In the study of Makela et al (62) a high association between symptoms of the neck and low back was also reported (odds ratio 4.3). Th is high assoc iation between several musculo skelet al symptoms may have been due to the percepti on or reporting of symptoms, but it may also have been an indication of exposure to shared risk factors or of common etiologic factors.

Physical load versus psy choso cial factors
Only few studies present quantitative data on the relative importance of physical load and psychosocial load to musculoskeletal trouble (4,26,33,64,62). Heliovaara et al (26) reported an odds ratio of 2.5 for the relationship between unspecified low-ba ck pain and a sum score for physical load (highest category versus lowest category) and an odds ratio of 1.4 for a sum score for psychosocial factors. For selfreported sciatica the odds ratio s were almost similar ( 1.9 and 2.0, respectively). For the same population Makela et al (62) reported similar odds ratios of 1.3 and 1.2 for the relat ionship between neck or shoulder symptoms and a sum score for physical load and psychosocial factors (odds ratios computed for each addition al score level). For a comparable population odds ratios were reported of similar magnitude (odds ratios around 2) for physical load and psychosocial factor s in relation to neck symptoms (33). With respect to back pain the odds ratios for physical load and psychosocial factors were 2.3 and 1.7, respectively. Veiersted & Westgaard (64) observed that both strenuou s posture and psychosocial factors were risk factor s for trapeziu s myalgia, with point estimates of the relative risks of around 10 for both factors. (Due to small numbers these estimates are imprecise.) The se data tentatively suggest that for neck and shoulder symptoms similar risk estimates were reported for both physical load and psycho social factors, whereas, for back symptom s, the risk estimate for the physical load tended to be slightly higher than for psychosocial factor s.

Nonwork psychosocial factors
The relation ships between psycho social factors outside work and musculoskeletal symptoms are not inluded in this review. If these factors are of major Importance, the conclusions based on work-related factors alone may not be correct. In a more extensive report on the subject (15) we concluded that studies investigating the relationships between musculoskeletal symptoms and psychosocial factor s outside the work environment, such as life events (29,30) or social support in the family (19,35,38,41,59,37), did not report a strong relationship between these factors and musculo skeletal symptom s.

Chance findings
Conflicting data can also be due to chance. In several studies many associations were tested, and, therefore, there was the possibility that several would tum out to be significant, if only due to chance. How- Scand J Work Environ Health 1993, vol 19. no 5 ever it should be noted that only few studies showed associations between musculoskeletal trouble and psycho social variables in the direction opposite to that expected.

Discussion of the results in view of the hypothesized mechanisms
In the following discussion the relationships presented in the Results section are presented according to the hypothesized associations in figure I. These association s are (i) stressors at work directly influenc e mechanical load and (ii) psycho social factors at work (demands, control , and social support), modified by individual characteristics, increase workrelated stress, which may increa se muscle tone and lead to the development of musculoskeletal symptoms or increase the musculoskeletal symptoms due to some unknown physiological mechanism and which may increase the perception of musculoskeletal symptoms and thu s prolong or intensify these symptom s or increase symptom s of poor health in general, including musculoskeletal symptoms.
When all of the reported data are combined, it is concluded that monotonou s work, time pressure, and perceived high work load each show a positive relation with musculo skeletal trouble , although such a relationship was not observed consistently in all of the studies. Part or all of this association can be attributed to the high mechanical load associated with these variables. This conclusion has been illustrated by Theorell et al (66), who showed that the strength of the association between perceived time pressure and symptoms of the back, neck, or shoulder decreased considerably after adju stment for physic al load. In several studies, however, a statistically significant odds ratio remained even after adjustment for physical load. Th is difference can be due to residu al confounding, since physical load is hard to measure accurately, or it may indicate an effect of these demand variables that is additional to the associated mechanical load. Some of the other job demand variables that are probably less strongly related to mechanical load , such as lack of clarity or high demands on concentration, showed no clear associat ion with musculo skeletal symptom s, but few of the studies analyzed the role of these variables. However, the positive association between low control , poor social support by collea gues, and combination s of these vari ables on one hand and musculoskeletal trouble on the other seems to indicate that an association between (some) psycho social factor s independent of increased mechan ical load seem s likely.
It was hypothesized that perceived stress or stress symptoms are an intermediate in the relationship between psychosocial factors at work and the development of musculoskeletal symptoms. Before it can be determined whether stress symptom s are a result or a cause of musculo skeletal pain, stress symptoms should be measured before the onset of musculoskeletal symptoms. A few of the studies that both did so and adjusted for other factors provide some evidence for the role of stress in the development of musculoskeletal trouble. The evidence supporting the assumption that stress predicts musculoskeletal symptoms is slightly more convincing than that for the opposite relationship. Although both relationships are probably true. One study (33) showed that both psychosocial factors and stress symptoms were related to back (odds ratio 1.7 and 3.5, respectively) and neck or shoulder troubl e (odds ratio 2.7 and 4.2, respectively). If the relationship between psychosocial factor s and back or neck or shoulder trouble was adju sted for the stress symptom score, the odds ratios were 1.7 and 2.0 and remained statistically significant. This result show s that stress symptoms are not necessarily an intermediate in the relationship between psychosocial factors at work and back trouble, although the stress symptom score in itself was strongly associated with back trouble.
Only one of the epidemiologic studies (66) provided data on the possible intermediate role of (chronic) muscle tension . In this study several selfreported stress symptom s were related to self-reported muscle tension. Muscle tension in turn was related to back, neck, or shoulder symptoms. No information has been presented on the direct association between stre ss symptoms and symptoms of the back , neck, or shoulders. Some of the psycho social factor s (high demands and lack of opportunities to talk) were directly associated with muscle tension. The stres s variables were also associated with other types of self-reported tension (eg, breath tension). These types of tension were also related to musculoskelet al symptoms. Se veral experimental studies have previousl y shown that muscle tension increa ses with increasing perceived stress. The only other study that tested the relationship between physiological parameters and musculoskeletal symptoms obser ved no positive association (45).
It cannot be inferred from the presented data whether the relationship between stress symptoms and musculoskeletal trouble is the result of an increased general perception of symptom s or a specific (physiological) mechani sm. Most of the studie s that analyzed the relationship between other physical and behavioral health indicators and back troubl e reported a positive association. It is remarkable however that several of these studies report a significant univariate association that was no longer significant after other variables were included, for example , stress symptoms and indi vidual characteristics, in a multivariate model. Some studies (sec, cg, reference 44) however lend support to the hypothesis that poor health is associated with and may predict back and neck trouble. This finding seems to suggest that stress influences the perception of both musculoskeletal symptoms and symptoms coming from other parts of the body. It does not necessarily mean how-310 ever that increased musculoskeletal symptoms are due to increased perception. This correlation between stress and indicators of poor health may also be partly due to shared risk factors for both stress and musculoskeletal trouble (eg, age). In addition, in some cases, medical explanation s may account for the relationship . For example, increa sed spinal pressure due to chronic cough may be responsible for the association between frequent coughing and back pain .
Additional data are needed for more informati on on the plausibility of the various explanations for the empiric al associat ion s reported in the literature.

Recom mendations
In future research on the relat ionship between psychosocial variable s and musculoskeletal disea se, it seems important to assess (i) the psychosocial factors at work through observ ation or with neutral questions, (ii) the perception of worker s concerning these variables, (iii) the self-reported stress and stress symptoms, special attention being given to musclular tension , (iv) mechanical load by measurement, and (v) musculoskeletal symptoms (duration and type of symptoms) from self-reports and musculoskeletal symptoms and signs from physical examinations.
With respect to the psychosocial factor s the variables from the job demands and job decision latitude model for work stress (demands, control, and social support) should receive special attention. Epidemiologic research should attempt to assess the relevance of all these variables in relation to each other. Longitudinal studies may provide inform ation on the temporal relation and are therefore of primary importance. Moreover, in study design and analys is, a clear distinction between risk factors for the development of musculoskeletal trouble , the persistenc e of symptom s, and the prediction of sick leave and disability appears to be important.
Thus future studies on psycho social variable s related to musculo skeletal trouble should ideally be longitudinal, directed towards the analy sis of the development or persistence of symptoms or pathology (disability), and pay attent ion to the independent effect and interaction between mechani cal load , psychosocial factors at work, and stress symptoms . Such studies would provide society with better tools to set prior ities in the prevention of work-related musculoskeletal disease and job (re)design.