Gender differences in the relations between work-related physical and psychosocial risk factors and musculoskeletal complaints

Gender differences in the relations between work-related physical and psychosocial risk factors and musculoskeletal complaints. Scand J Work Environ Health 2004;30(4):261–278. Gender differences in the prevalence of musculoskeletal complaints might be explained by differences in the effect of exposure to work-related physical and psychosocial risk factors. A systematic review was conducted to examine gender differences in the relations between these risk factors and musculoskeletal complaints. Several electronic databases were searched. The strength of the evidence was determined on the basis of the methodological quality and consistency of the study results. For lifting, strong evidence was found that men have a higher risk of back complaints than women. The same was found for the relation between hand–arm vibration and neck– shoulder complaints. For arm posture, strong evidence was found that women have a higher risk of neck– shoulder complaints than men. For social support, no evidence of a gender difference was found for either neck– shoulder or back complaints. For hand–wrist and lower-extremity complaints, inconclusive evidence was found due to a lack of high-quality studies.

Many studies have reported gender differences in the prevalence of musculoskeletal complaints. For example, in a large population-based study in The Netherlands, 79.3% of the women and 71.5% of the men reported one or more musculoskeletal complaints in the past year (1). The one-year prevalence of self-reported spinal pain (including lower back, upper back, and neck) in a sample of 35-to 45-year-old Swedish residents was 69.5% for women and 63.2% for men (2). In the United States the prevalence of chronic joint symptoms in 2001 was 37.3% for women and 28.4% for men (3).
This gender difference seems to be more distinct for neck and upper-extremity complaints than for back complaints. The prevalences of neck and upper-extremity complaints has been found to be consistently higher for women than for men (1,4,5), while the prevalence of back complaints has been shown to be markedly higher for women (6) or slightly higher for women (1,4) and also slightly higher for men (7).
Several explanations have been proposed for the gender difference in prevalence (8)(9)(10)(11). First, it has been proposed that men and women have different exposure to risk factors, either because of differences in exposures outside work or because of differences in work exposure due to the sex segregation of the labor market. This last factor has been suggested to be the most important explanation for the sex difference in the prevalence of musculoskeletal complaints. However, the difference in prevalence remains when men and women from the same occupational class (12), or with the same work tasks (13) are compared. Second, it is claimed that women are more prone to express pain and symptoms, either because they have a lower threshold for detecting pain and symptoms or because they are more willing to express their feelings than men, who are taught not to complain (14). If this were true, one would expect that the gender difference in the prevalence of self-reported pain or symptoms would be larger than for objectively Scand J Work Environ Health 2004, vol 30, no 4 measured problems. Yet, Punnett & Herbert (8), who reported that some of the largest gender differences were found in studies in which objective measures were used, did not show this result. The third explanation to be suggested is that the same risk factors may have a different effect on men and women. In this respect, it has been pointed out that joint laxity seems to be influenced by sex hormones (15,16), women therefore being more vulnerable for musculoskeletal pain. In addition, women, on the average, have smaller body dimensions, lower muscle force, and a lower aerobic capacity. Therefore tasks performed with the same (absolute) exposure will, in most cases, result in a higher relative workload for women (8,17,18), which could lead to more complaints. Finally, men and women have been found to use different coping strategies for dealing with occupational stressors (19), and this difference could result in different outcomes.
In this review we focus on gender differences in the effect of risk factors. The aim was to determine whether there are gender differences in the relations between work-related physical and psychosocial risk factors and musculoskeletal complaints of the back, neck-shoulder, hand-wrist, and lower extremities.  (1990( -January 2002, were checked in order to identify relevant studies. The databases were searched with the following search string: (risk factor OR predictor OR determinant or causality OR (a)etiology OR causal factor) AND (gender (difference) OR sex (difference)) AND (work(-)(related) OR work environment OR job OR employment OR workplace OR occupation(al)) AND (back (pain) OR musculoskeletal (disorder) OR upper extremity (disorder) OR lower extremity (disorder) OR shoulder OR wrist OR elbow OR neck OR knee OR RSI OR repetitive strain injury OR cumulative strain disorder OR hand OR arm OR leg OR foot OR feet). In addition, a snowball search was performed, and the references of some recent reviews (20)(21)(22)(23)(24)(25) were checked for relevant publications. Finally, articles from personal databases were included.

Selction of the literature
Articles were included if they met the following criteria: (i) the study design was cohort, case-control or cross-sectional, (ii) the study population included both men and women who came from a working or community-based population, (iii) the study addressed a musculoskeletal complaint, (iv) the exposure to relevant risk factors was measured separately for men and women and, for example, not based on job title or a job exposure matrix, (v) separate analyses were performed for men and women or an interaction effect for gender was calculated, and (vi) the study was published in a peerreviewed journal in English. Two reviewers (WH and MP) read the titles and abstracts of all the studies to decide whether the inclusion criteria were met. If no abstract was present or if, based on title and abstract, it still was unclear whether an article should be included or excluded, the complete article was retrieved and checked.

Quality assessment
The quality of the studies was assessed using a quality assessment list (table 1), based on lists used in earlier reviews of observational studies (22,25). The items on the list were rated as + (minimal requirements met), -(minimal requirements not met) or ? (unclear whether the minimal requirements were met). For all the studies, the number of positive items was calculated. Studies were rated as high in quality if they scored positive on at least 50% of the relevant items. Two reviewers (WH and MP) separately evaluated the quality of the studies. A consensus meeting was arranged to sort out differences between the reviewers.

Data extraction
From all the studies, information on design, population, response rate, exposure, outcome, and the risk estimates [relative risk (RR), odds ratio (OR)] were extracted for the men and women. When risk estimates were not presented, but enough data were given, the risk estimates were calculated. When multiple outcome measures were presented, for example, pain and sick leave, the outcome closest to the complaint level was used in the analysis.
When it is being determined whether there was a gender difference for a risk factor, it is not sufficient for a risk estimate to be statistically significant in one group and not in another. It is also not correct to say that, if confidence intervals overlap, the risk estimates are not significantly different (26). Therefore, we divided the risk for women by the risk for men in order to calculate a gender ratio. A ratio higher than 1.25 (ie, women had a higher risk) or lower than 0.75 (ie, women had a lower risk) was regarded as a relevant gender difference.
It was anticipated that a wide variety of risk factors would be found in the various studies. Therefore, on the basis of the results of several recent reviews (20)(21)(22)(23)(24)(27)(28)(29)(30)(31)(32), we selected certain risk factors and used them in the analysis. The following tasks and job characteristics were considered physical risk factors: (i) lifting, manual materials handling, patient handling, awkward posture, bending and twisting, heavy physical workload, and whole-body vibration for the back; (ii) repetition, hand-arm vibration, arm posture, arm force, and head posture for the neck-shoulder region; (iii) repetition, vibration, wrist posture, and use of force for the handwrist region; and (iv) heavy physical work, kneeling or squatting, walking, and climbing for the lower extremities. Job demands, job control, social support, and job satisfaction were considered psychosocial risk factors for all the regions of the body under study.

Levels of evidence
Based on the reviews of Ariëns et al (22) and Hoogendoorn et al (25), the following four levels of evidence were constructed to determine the strength of evidence for a gender difference: (i) strong evidence, comprised of consistent gender differences in multiple high-quality cohort or case-control studies; (ii) moderate evidence, consistent gender differences found in one high-quality cohort or case-control study and at least one low-quality cohort or case-control study or consistent gender differences found in multiple low-quality cohort or casecontrol studies or consistent gender differences found in multiple high-quality cross-sectional studies; (iii) inconclusive evidence, consistent gender differences found in multiple low-quality cross-sectional studies or inconsistent results found in multiple studies or results based on one study; and (iv) no evidence of a difference, consistently no gender differences found. The results were regarded as consistent if at least 75% of the results were in the same direction.

Selection of the literature
The search resulted in a total of 1653 articles. After the exclusion of doubles, 1473 titles and abstracts were reviewed for their relevance. Initially, there was a 7% disagreement between the reviewers about whether a paper met the inclusion criteria. After these disagreements were resolved, the full text of 185 articles was retrieved. On the basis of the full text, we included 31 studies. Another nine studies were included on the basis of the snowball search, the reference check, and perusal of personal databases.
Eight studies (33)(34)(35)(36)(37)(38)(39)(40) were excluded after the data extraction because they did not present a risk estimate or there were not enough data to calculate one. Two studies (41,42) were excluded because they did not report musculoskeletal complaints for a specific region. Finally, four studies (43)(44)(45)(46) that met all the inclusion criteria and presented their data in a usable way could not be used in the analysis because they did not report findings in respect to any of the predetermined risk factors. Therefore, 14 studies (47-60) on back complaints, 9 studies (58-66) on neck-shoulder complaints, 4 studies (65-68) on hand-wrist complaints and 4 studies (69-72) on lower-extremity complaints were used. A description of the studies that were used is given in the appendix. Only relevant outcome and exposure measures are presented.

Back complaints
A summary of the determination of the levels of evidence for back complaints can be found in table 3. Eight studies (47, 52-54, 56-58, 60) concerned lifting. The high-quality cohort study (58) found a gender ratio of 0.18, while, in the high-quality case-control study (60), gender ratios of 0.57 and 0.80 were found for heavy lifting and manual materials handling, respectively. The low-quality cohort and case-control studies (52, 54) and a high-quality cross-sectional study (47) found gender ratios between 1.35 and 2.27. The second high-quality cross-sectional study (56) and a low-quality cross-sectional study (57) found no difference between men and women, while, in another low-quality cross-sectional study (53), a ratio of 0.55 was found. On the basis of the results of the high-quality cohort and case-control studies, we concluded that there is strong evidence that men have a higher risk of back complaints due to lifting than women do. Posture was investigated in four studies (47,53,58,60). The high-quality cohort study (58) and the Level of evidence high-quality cross-sectional study (47) found no difference between men and women. The high-quality casecontrol study (60) and the low-quality cross-sectional study (53) showed risk ratios of 0.67 and 0.40, respectively. Since the results of the high-quality cohort and case-control studies were not consistent, there is inconclusive evidence for a gender difference for posture. Four case-control (50,55,59,60) and two cross-sectional (48,49) studies reported on heavy physical work as a risk factor for back pain. One high-quality (60) and two low-quality case-control (50, 55) studies found a larger risk for women (gender ratios ranging from 1.36 to 3. 43). No difference in the risk estimate between men and women was found in the other high-quality casecontrol study (59) and the cross-sectional studies (48,49). Since these results were not consistent, there is inconclusive evidence for a gender difference for heavy physical workload.
Whole-body vibration, measured as vibration or driving, was investigated in six studies (52,54,(56)(57)(58)60). The high-quality case-control (60) and the lowquality cross-sectional (57) studies found gender ratios of 3.11 and 1.40, respectively. However, the high-quality cohort study (58) found a gender ratio of 0.58. In the high-quality cross-sectional study (56), gender ratios of 0.24-0.67 for driving, and a gender ratio of 3.80 for exposure to vibration machinery, were found. Finally, the low-quality cohort (54) and case-control (52) studies did not find a gender difference. Since these results were not consistent, there is inconsistent evidence of a gender difference for whole-body vibration.
Job demands were assessed in five studies (48,49,51,58,59). One high-quality cohort study (58) and one high-quality cross-sectional study (48) found gender ratios of 1.90 and 1.35 respectively. The second high-quality cohort study (51), the high-quality case-control study (59), and a low-quality cross-sectional study (49) did not find a gender difference. Due to the inconsistency of these results there is inconclusive evidence of a gender difference for job demands.
Five high-quality studies (47,48,51,58,60) examined job control. One cohort study (58) found a gender ratio of 1.35, while, for the other cohort study (51), a gender ratio of 0.70 was calculated. The case-control study (60) and both cross-sectional studies (47, 48) did not find a gender difference. Because of the inconsistency of these results, there is inconclusive evidence for a gender difference for job control.
A gender ratio of 1.41 for social support as a risk factor was found in a high-quality cohort study (58). However, the other high-quality cohort study (51), both high-quality case-control studies (59,60), and the highquality cross-sectional study (48) did not find a gender difference. The conclusion, therefore, is that there is no evidence of a gender difference.
A gender difference in the relation between job satisfaction and back pain was only found in one high-quality case-control study (60), with a gender ratio of 0.33. No gender difference was found in two high-quality cohort studies (51,58) and one low-quality cross-sectional study (49). Due to the inconsistency in the high-quality studies, there is inconclusive evidence for a gender difference for job satisfaction. Table 4 provides an overview of the determination of the levels of evidence for neck-shoulder complaints. A total of five studies (59)(60)(61)63) assessed the relation between repetition and neck-shoulder complaints. One high-quality case-control study (60) found a gender ratio of 1.33, while the second high-quality case-control study (59) did not find a gender difference. The highquality cohort study (61) found a gender ratio of 1.44 for the exposure at baseline, but no difference for exposure before baseline. The results of the low-quality cross-sectional studies (63,65) were not consistent either, with gender ratios of 0.53-2.34, depending on the exact outcome and exposure. Because of these inconsistent results, there is inconclusive evidence for a gender difference for repetition.

Neck-shoulder complaints
The relation between hand-arm vibration and neckshoulder complaints was measured in four studies (59,60,62,63). Both high-quality case-control studies (59,60) and one low-quality cross-sectional study (63) found a larger risk for men (gender ratios 0.50, 0.54 and 0.73, respectively). The second low-quality cross-sectional study (62) found a gender ratio of 0.22 for pain in the past 7 days, but no difference for pain in the past 12 months. Because the case-control studies (59, 60) consistently showed a higher risk estimate for men, it is concluded that there is strong evidence that exposure to hand-arm vibration is a larger risk for men.
Arm posture was investigated in one high-quality cohort study (58), one high-quality case-control study (60), and three low-quality cross-sectional studies (62,63,66). The cohort and case-control studies found larger risk estimates for women, with gender ratios of 6.39 (58) and 1.44 (60). The cross-sectional studies found no difference between men and women (62,66) or a larger risk for men (63). The results of the cohort (58) and case-control (60) studies indicate that there is strong evidence that exposure to awkward arm postures is a larger risk factor for women than for men.
Arm force, measured as lifting, was measured in one high-quality cohort study (58), one high-quality casecontrol study (60), and two low-quality cross-sectional studies (62,63). The case-control study (60) and one of the cross-sectional studies (63) found a larger risk for men (gender ratios from 0.20 to 0.67). No gender difference was found in the second cross-sectional study (62) and the cohort study (58), in which men and women with a heavy lift index had a lower risk of neckshoulder complaints. Therefore, the evidence is inconclusive.
Job demands were investigated in seven studies (58-61, 64-66). One high-quality cohort study (58) found a gender ratio of 0.64, but the two high-quality case-control studies (59, 60) found gender ratios from 1.57 to 4.50. No gender difference was found in the second high-quality cohort study (61) and the cross-sectional studies (64)(65)(66). Since these results were not consistent, there is inconclusive evidence of a gender difference for job demands.
Three high-quality studies (58-60) measured job control. One case-control study (60) found no gender difference, but the second case-control study (59) found a gender ratio of 5.0. The gender ratio in the cohort study (58) was 1.33. Due to the inconsistency of the results, there is inconclusive for a gender difference for job control.
One high-quality cohort (58), one high-quality casecontrol (60), and one low-quality cross-sectional (66) study concerned social support. Since none of them found differences between men and women, it is concluded that there is no evidence for a gender difference.

Hand-wrist complaints
Two low-quality cross-sectional studies (65, 67) concerned repetitive movements. Since only one of them (67) found a gender difference (gender ratio 1.29), there is inconclusive evidence for a gender difference. One study (68) focused on the relation between vibration and hand-wrist complaints (gender ratio 0.49), but, since this was a low-quality cross-sectional study, there is inconclusive evidence of a gender difference.
Three cross-sectional studies (66-68) concentrated on wrist postures. One of them (67) found gender differences, with ratios of 0.71 and 1.29 depending on the exact exposure, but the other two studies found no gender differences. Since these results were inconsistent and the studies were low in quality, there is inconclusive evidence for a gender difference.
Job demands were measured in two studies (65,66), but only one of them (66) found a gender difference. Due to the inconsistency and the low quality of the studies, there is inconclusive evidence of a gender difference.
One low-quality cross-sectional study (66) reported the relation between social support and hand-wrist complaints. No gender difference was found, but, since the results were based on only one study, there is inconclusive evidence for a gender difference.

Lower-extremity complaints
Only one low-quality study (72) reported a relation between heavy physical workload and lower-extremity complaints. This study found a gender ratio of 1.33. With only one study, there is inconclusive evidence of a gender difference.
Four low-quality case-control studies (69-72) reported on kneeling or squatting. Two studies (69,72) used exposures that combined kneeling and squatting. Neither study found a gender difference. Two studies (70,71) found a gender difference for kneeling (gender ratio 0.33-0. 64), and, in one study (69), a gender ratio of 1.27 was found for squatting. Since the results of these studies were not consistent, there is inconclusive evidence for a gender difference for kneeling or squatting.
Much walking was a larger risk factor for men in two out of four low-quality case-control studies (71,72), with gender ratios from 0.36 to 0.72. The third study showed no difference between men and women, while, in the fourth, gender ratios of 1.36 and 1.88 were found. Due to the inconsistency of the results, there is inconclusive evidence for a gender difference for walking.
Climbing was measured in four low-quality casecontrol studies (69)(70)(71)(72). Two studies (69-72) found a larger risk for men (gender ratios 0.30-0.54). In one study (71) the direction of the gender difference depended on the outcome (gender ratio 0.18 for hip complaints and 2.04 for knee complaints), and in one study (70) the direction was dependent on the duration of the exposure (gender ratios 0.57-1.28). Since these results were not consistent, there is inconclusive evidence of a gender difference for climbing.

Discussion
The purpose of this review was to examine gender differences in the effect of exposure to work-related physical and psychosocial risk factors. Considering the gender differences in prevalence, we expected that women would have higher risks. The results show evidence of a gender difference for a few risk factors, but in most cases men had the higher risk.

Back complaints
Before the study, we presumed that women would have a higher risk of back complaints due to lifting than men, but we found strong evidence that men have a higher risk. However, it could be argued that, since the weight of the average larger male torso has to be added to the weight of the lifted object, men in fact have a higher exposure than women when lifting an equal object. This difference may be one factor leading to a higher risk of back complaints among men. Several studies (18,73,74) indeed found that men have a greater absolute exposure, due to their greater body mass. However, these same studies also showed that women are not merely scaled-down versions of men, but, in fact, use different techniques while lifting. In the end, this difference resulted in a greater relative workload for women and, therefore, in a greater risk of complaints. Another remarkable point is that Vingard et al (60) found a (not significant) relative risk of 0.8 for women, while Barnekow-Bergkvist et al (58) found odds ratios of <1 for both men and women. This evidence is clearly in contrast to the generally accepted view that lifting is a risk factor for back pain (23,32). It should be mentioned, however, that the study population in this last study was relatively young [mean age 34 (SD 0.74) years]. Together with the possible selection bias of this study, the young age may explain the unexpected result. Finally, although the high-quality cohort and case-control studies found gender ratios below 0.75, the low-quality cohort and case-control studies consistently found ratios above 1.25. Therefore, the conclusion that men have a higher risk than women due to lifting should be considered with due caution.

Neck-shoulder complaints
As for back complaints, it was expected that women would have a higher risk. This was indeed the case for arm posture, but for hand-arm vibration men had the higher risk. The studies in our review used a rather low cut-off point for exposure (30 minutes and 16% of the time); hence a large range of exposures within the highest exposure category was possible. Total daily exposure to vibration has been found to be much higher for men than for women (75), and, therefore, men may still have had a higher exposure than women within the same exposure category. Furthermore, the effect of vibration on complaints may be rather small for women, since the 1-week prevalence of exposure was found to be only 6% for working women, but 32% for working men (75).

Hand-wrist complaints
Very few studies on hand-wrist complaints were found. Although initially nine studies were identified, four were excluded because they did not report findings for the selected risk factors. Three studies (44,65,66) considered the duration of computer use as a risk factor, but the results were not consistent. While, in the study by Blatter et al (44), the risk was larger for women (gender ratios ranging from 1.05 to 1.38), Jensen et al (65) and Karlqvist et al (66) found larger risks for men (ratios ranging from 0.55 to 0.99). Nevertheless, only a few studies reported risk factors for men and women separately, and the reason for the inconclusiveness should primarily be sought in the lack of (high-quality) studies. Furthermore, since all these studies were cross-sectional, no causal relation could be established. It is recommended that more, preferably prospective, studies on hand-wrist complaints make separate analyses for men and women.

Lower-extremity complaints
Due to the inconsistency and the small number of lowquality studies, inconclusive evidence was found for all the risk factors for lower-extremity complaints. As for hand-wrist complaints, we would like to emphasize the need for more (high-quality) studies.

Selection of the literature
To our knowledge this is the first review that systematically examined gender differences in the relation between work-related risk factors and musculoskeletal complaints. In spite of our extensive literature search, it is likely that both selection and publication bias influenced the results. Most studies on risk factors do not aim at examining gender differences and do not use key words referring to such differences. By including the terms gender (difference) and sex (difference) in the search string, we may have missed these studies. Another potential source of bias is publication bias. While some studies tested for all possible interactions or made separate analyses for all risk factors, most of the studies only did this for a few variables. It could very well be that such an approach was only used because (significant) gender differences were found for these risk factors. The results of this review may therefore overestimate gender differences.

Analysis
We chose to use a percentage difference in risk estimates rather than an absolute number or a significant difference to identify relevant differences. However, we could not find theoretical grounds for the point of cut-off. By using the percentage difference, we had to exclude studies that did not present risk estimates and those which only reported a nonsignificant difference. Four of these studies did mention that there was no difference between men and women, or no significant interaction with gender was found (37)(38)(39)(40). One study (37) assessed the relation between lifting and back complaints, three studies concerned job demands and neck-shoulder complaints (38)(39)(40), while job control, social support, and work with hands above shoulder level were each assessed in one study (39,40). Considering these studies did not change the strength of the evidence.

Methodological quality and levels of evidence
The combination of a quality scale and levels of evidence is often used, but not without criticism (76,77). Our quality list was very similar to lists used earlier (20,22,23,25). One of these lists (22) was rated by West et al (78), and it scored positive on six and partially positive on one out of nine domains for assessing study quality. A point of criticism on this and similar lists is that all the items have the same weight, and studies that have only a few, but very important, flaws can still be regarded as high in quality (21,22). In our review, the three studies with the highest quality (59 ,60, 64) scored positive on all items regarding validity of outcome and exposure measures. Another three high-quality studies (48,51,58) scored positive on at least one of these items, while none of the lowquality studies scored positive on these items. Therefore, these items are important in discriminating between highand low-quality studies. Another point of criticism is that, when different levels of evidence are compared, their agreement is poor and may result in differences in the conclusion (76). Unfortunately, to our knowledge, no other levels of evidence for observational studies have been published, and no comparison can be made with our levels.

Concluding remarks
Strong evidence of a gender difference was found for only three risk factors, but for two out of three factors the difference was not in the expected direction. These findings seemed fairly insensitive to the limitations of our study, but are likely to be an overestimation of gender differences. Therefore, the results should be interpreted with some caution. For hand-wrist and lowerextremity complaints only a few low-quality studies were found, and it is recommended that more studies make separate analyses for men and women.
Since gender differences in the effect of risk factors do not seem to provide an explanation for the higher prevalence of musculoskeletal complaints among women, alternative explanations have to be considered, such as gender differences in the number of workers exposed, in exposure within the same exposure category, or the expression of pain (8)(9)(10)(11). In terms of prevention, until more clarity is achieved, the focus should remain on the reduction of exposure among female workers.