A Critical Review of Epidemiologic Evidence for Work-Related Musculoskeletal Disorders of the Neck, Upper Extremity, and Low Back

Chapter 3. Shoulder Musculoskeletal Disorders: Evidence for Work-Relatedness

Summary

There are over 20 epidemiologic studies that have examined workplace factors and their relationship to shoulder musculoskeletal disorders (MSDs). These studies generally compared workers in jobs with higher levels of exposure to workers with lower levels of exposure, following observation or measurement of job characteristics. Using epidemiologic criteria to examine these studies, and taking into account issues of confounding, bias, and strengths and limitations of the studies, we conclude the following:

There is evidence for a positive association between highly repetitive work and shoulder MSDs. The evidence has important limitations. Only three studies specifically address the health outcome of shoulder tendinitis and these studies involve combined exposure to repetition with awkward shoulder postures or static shoulder loads. The other six studies with significant positive associations dealt primarily with symptoms. There is insufficient evidence for a positive association between force and shoulder MSDs based on currently available epidemiologic studies. There is evidence for a relationship between repeated or sustained shoulder postures with greater than 60 degrees of flexion or abduction and shoulder MSDs. There is evidence for both shoulder tendinitis and nonspecific shoulder pain. The evidence for specific shoulder postures is strongest where there is combined exposure to several physical factors like holding a tool while working overhead. The association was positive and consistent in the six studies that used diagnosed cases of shoulder tendinitis, or a constellation of symptoms and physical findings consistent with tendinitis, as the health outcome. Only one [Schibye et al. 1995] of the thirteen studies failed to find a positive association with exposure and symptoms or a specific shoulder disorder. This is consistent with the evidence that is found in the biomechanical, physiological, and psychosocial literature.

There is insufficient evidence for a positive association between vibration and shoulder MSDs based on currently available epidemiologic studies.

Introduction

Shoulder MSDs and their relationship to work risk factors have been reviewed by several authors [Hagberg and Wegman 1987; Kuorinka and Forcier 1995; Sommerich et al. 1993; Winkel and Westgaard 1992]. Hagberg and Wegman [1987] attributed a majority of shoulder problems occurring in a variety of occupations to workplace exposure. Kuorinka and Forcier [1995] looked specifically at shoulder tendinitis and stated that the epidemiologic literature is “most convincing” regarding work-relatedness, especially showing an increased risk for overhead and repetitive work.

The focus of this review is to assess evidence for a relationship between shoulder tendinitis and workplace exposures to the following: awkward postures, forceful exertions, repetitive exertions, and segmental vibration. Also included are studies relevant to shoulder disorders—as defined by a combination of symptoms and physical examination findings or by symptoms alone, but not specifically defined as tendinitis—and those studies for which the health outcome combined neck and shoulder disorders, but where the exposure was likely to have been specific to the shoulder. Chapter 2 discusses studies involving neck-shoulder disorders where assessment of exposure was likely specific to the neck region.

Pertinent information about the 39 reviewed studies is presented in several ways. Detailed descriptions of the studies are provided in Table 3-5. The text of this section on shoulders is organized by exposure risk factor. The discussion within each risk factor is organized according to criteria presented on Pages 1-1 to 1-10 of the Introduction. Conclusions are presented with respect to the specific MSD of concern, shoulder tendinitis.

Back to top

Repetition

Definition of Repetition for Shoulder MSDs

Studies that addressed the physical factor of repetition and its relation to shoulder MSDs were included in this review. Studies usually defined repetition, or repetitive work, for the shoulder as work activities that involved cyclical flexion, extension, abduction, or rotation of the shoulder joint. Repetitiveness was defined in four different ways in the reviewed studies: (1) the observed frequency of movements past pre-defined angles of shoulder flexion or abduction, (2) the number of pieces handled per time unit, (3) short cycle time/repeated tasks within cycle, and (4) a descriptive characterization of repetitive work or repetitive arm movements. Some of the studies that examined repetition as a risk factor for shoulder MSDs had several concurrent or interacting physical work load factors. Therefore, repetitive work should not be considered the primary exposure factor, particularly independent of posture. Some studies indirectly inferred shoulder repetition by characterizing hand, wrist, and forearm movements.

Studies Reporting on the Association of Repetition and Shoulder MSDs

Three of the reviewed studies reported results on the association between repetition and shoulder tendinitis [English et al. 1995; Ohlsson et al. 1994, 1995]. For all three studies, some or all of the results were for associations with a combined exposure to repetition and awkward posture. Six additional studies reported results on the association between repetition and non-specific shoulder disorders [Sakakibara et al. 1995], non-specific shoulder symptoms [Andersen and Gaardboe 1993a; Ohlsson et al. 1989], combined neck-shoulder disorders [Bjelle et al. 1981; Chiang et al. 1993] or combined neck-shoulder symptoms [Kilbom et al. 1986; Kilbom and Persson 1987].

Studies Meeting the Four Evaluation Criteria
Four studies met all four of the criteria [Chiang et al. 1993; Kilbom et al. 1986; Ohlsson et al. 1994, 1995] (Table 3-1, Figure 3-1). Chiang et al. [1993] studied workers in the fish processing industry in Taiwan. The health outcome of “shoulder girdle pain” was defined as self-assessed symptoms of pain in the neck, shoulder or upper arms, and signs of muscle tender points or palpable hardenings upon physical examination. Pain referred from a nerve root or other spinal source was included in the case definition. The force requirements of the jobs were estimated by surface electromyographs (EMGs) in the forearm flexor muscles. This is not a direct measure of shoulder muscle activity. There may be no relationship between the level of activity in the forearm and shoulder girdle muscles. Three categories, based on both force and repetitiveness, were used as the exposure outcome: Group I (low force, low repetitiveness), Group II (high force or high repetitiveness), and Group III (high force and high repetitiveness). Force was also evaluated independently in multivariate analyses.

Kilbom et al. [1986] performed a prospective study in which female employees in the electronics manufacturing industry were observed for a 2-year period. The health outcome in the neck, shoulder, or arm regions was based on symptoms and physical findings. Symptom severity was coded on the basis of its character, frequency, and/or duration. Changes in severity status at follow-up evaluations were used as the dependent variables in multiple regression analyses. Neck, shoulder, and upper arm posture was determined by VIRA. Although the health outcome combined symptoms from different body regions, knowledge of biomechanical theory can be used to identify significant predictors related to the shoulder symptom severity.

For the two Ohlsson et al. [1994, 1995] studies, the authors reported that the examiners could not be completely blinded to exposed versus referent status, but that a standard protocol was followed and observer bias was likely to have been minimal. As examiners were blinded to objective exposure measures, analyses testing associations between neck-shoulder disorders and specific postures would not have been biased [Ohlsson et al. 1995].

In the first of the Ohlsson et al. studies, a cross-sectional study, women in the fish industry were compared to a control population of women employed in municipal workplaces in the same towns [Ohlsson et al. 1994]. Diagnoses of shoulder disorders (e.g., tendinitis, acromioclavicular syndrome, frozen shoulder) were made on the basis of symptoms determined by interview and a physical exam. Exposure evaluation of each work task held by the fish industry population was evaluated with ergonomic workplace analysis (EWA). Ten different factors were rated on a scale from 1 to 5 and the combined ratings were used as a profile of the work task. Based on this profile, the authors reported that fish industry work was found to be “highly repetitive” and to include “poor work postures.”

Ohlsson et al. [1995] compared a group of women who performed industrial assembly work to a referent group of women from a nearby town who were employed in jobs characterized as having varied and mobile work tasks. One examiner assessed signs and symptoms. The examiner was blinded to specific exposure information, but not completely blinded to factory worker versus referent group status. Shoulder tendinitis included supraspinatus, infraspinatus, and bicipital tendinitis. Another health outcome combined neck and shoulder disorders (tension neck, cervical syndrome, thoracic outlet syndrome, frozen shoulder, tendinitis, acromioclavicular syndrome). In a descriptive assessment, it was reported that the work tasks in the study group involved repetitive arm movements with static muscular work of the neck and shoulder muscles. The percentage of time spent in specific upper arm postures was determined from videotaped observation of 74 (out of 82) workers. The average result from two independent videotape analyses was used. Posture category demarcations included 0, 30, and 60 degrees for arm elevation, and 30, 60, and 90 degrees for arm abduction.

Studies Not Meeting the Four Evaluation Criteria
Bjelle et al. [1981] compared cases with acute, non-traumatic shoulder-neck pain to age- and sex-matched, paired controls. To determine exposure, each case and control was filmed and a biomechanical analysis was performed to determine the frequency and duration of shoulder abduction or forward flexion > than 60 degrees.

In the study by English et al. [1995], cases were determined by medical diagnosis and controls were selected from patients evaluated at specified orthopedic clinics. For statistical analyses, all diagnoses were grouped by anatomical site. The diagnoses for shoulder cases were rotator cuff injury, rupture of long head of biceps, shoulder capsulitis, and symptomatic acromioclavicular arthritis. It is assumed that shoulder tendinitis is included in this group. Exposure measures were determined by a standardized interview conducted by an interviewer who was “unaware of the case-control status of the individual wherever this was possible.”

In a study by Sakakibara et al. [1995], the health status of a group of women farm workers was assessed during the performance of two different tasks, with a 1-month interval between the tasks. The health outcome was defined by self-assessed symptoms of shoulder stiffness and pain and a physical examination for muscle tenderness and joint pain on movement. Whether the examining physician was aware of the prior hypothesis regarding differing exposures between the two tasks (bagging pears versus bagging apples) was not stated. Exposure was based on self-report of the number of hours per day spent bagging, the number of pears or apples bagged per day, and the total number of days spent bagging each fruit. One worker was observed for 3 hours while performing each bagging job, with repeated goniometric measures of shoulder forward flexion angles done each minute. While there was no difference in the total number of days or number of hours per day spent bagging each fruit, significantly more pears than apples were bagged per day. The proportion of time spent with the angle of shoulder forward flexion greater than 90 degrees was significantly larger when bagging pears (75%) than when bagging apples (41%).

One study did not meet any of the criteria. In a cross-sectional study by Ohlsson et al. [1989], the exposed population was factory employees who produced and assembled plastic components. Work exposure was characterized as “repetitive arm and hand movements in constrained work postures.” The referent population was composed of women randomly sampled from the general population in a nearby area. The health outcome was determined by self-reported symptoms of shoulder pain during the previous seven days. The exposure measure was the self-reported number of items completed per hour. The range was from less than 100 items completed per hour (slow category) to more than 700 items per hour (very fast category). Self-reporting was believed to be accurate because workers were paid by the piece.

Strength of Association: Repetition and Shoulder MSDs

Using the data presented in the study by Ohlsson et al. [1994], for supraspinatus, infraspinatus, or bicipital tendinitis the odds ratio (OR) for working in the fish industry (repetitive work, poor posture) was calculated as 3.03 (95% CI 2.5–7.2). For shoulder tendinitis alone, the PRR was calculated as 3.5 (95% CI 2.0–5.9). For clinical diagnoses of the neck and shoulder, the OR for working in the fish industry versus the referent population was 3.2 (95% CI 2.0–5.3).

Using data presented in the study by Ohlsson et al. [1995] for supraspinatus, infraspinatus, or bicipital tendinitis, the OR for being an assembly worker (repetitive arm movements with static load on shoulders) versus the referent population was 4.2 (95% CI 1.35–13.2). For neck-shoulder disorders, the OR for being an assembly worker versus the referent group was 5.0 (95% CI 2.2–11.0).

Using multiple logistic regression analysis with age, gender, and force as covariates, Chiang et al. [1993] found that highly repetitive upper extremity movements were associated with shoulder girdle pain (OR 1.6, 95% CI 1.1–2.5). When tested in the same model with force and repetition, the interaction term for force and repetition was also significant (OR 1.4, 95% CI 1.0–2.0). Several factors could have resulted in an underestimation of the strength of association: no requirement that symptoms had begun on current job means that some symptomatic workers may have transferred to lower risk jobs. Relative to shoulder MSDs, the major limitation of this study was that the exposure assessment was not specific to movement at the shoulder joint and may therefore have either over- or underestimated repetition at the shoulder. In some cases the exposure assessment may have been a measure of repetitive upper arm movements, but it may also have been a measure of repetitive hand and distal upper extremity activity occurring in the context of a static load on the shoulder muscles.

For the shoulder diagnoses used to form their group of cases, English et al. [1995] found an association with repeated shoulder rotation with an elevated arm (OR 2.30, p< 0.05). They also found what appeared to be a protective effect associated with elbow flexion (OR 0.4, 95% CI 0.2–0.8). This effect was greatest at low amounts of daily cumulative exposure to elbow flexion; the protective effect decreased (RR increased) as the number of hours of total daily elbow flexion increased. In a laboratory study of shoulder muscle activity in relation to different combinations of shoulder and elbow joint postures (a total of 21 different postures), Herberts et al. [1984] found that humeral rotation and elbow flexion had insignificant effects on shoulder muscle activity. However, the postures tested by that study were stationary, whereas the associations reported by English et al. [1995] appear to be related to repetitive movements.

For symptoms of shoulder pain within the previous 7 days, the OR for assembly workers versus the referent group was 3.4 (95% CI 1.6–7.1) [Ohlsson et al. 1989]. A significantly higher proportion of the farm workers studied by Sakakibara et al. [1995] had signs of shoulder muscle tenderness while bagging pears than while bagging apples. There was no way to analyze the relative contribution to risk of repetitive shoulder exertions (increased number of pears picked per day) and awkward posture (greater portion of each day spent with extreme forward flexion when picking pears).

Consistency of Association

Repetitiveness was defined in four different ways in the reviewed studies: (1) the observed frequency of movements past pre-defined angles of shoulder flexion or abduction, (2) the number of pieces handled per time unit, (3) short cycle time/repeated tasks within cycle, and (4) a descriptive characterization of repetitive work or repetitive arm movements.

Repetition Characterized as Frequency of Movements Past Pre-Defined Shoulder Angles
Bjelle et al. [1981] and Ohlsson et al. [1995] found a significant positive association between the prevalence of neck-shoulder disorders and the frequency of upper arm movements past 60 degrees of flexion or abduction. English et al. [1995] found a significant association between diagnosed cases of shoulder disorders and repeated shoulder rotation with an elevated arm posture.

Repetition Characterized as the Number of Pieces Handled per Time Unit
A significant positive association was found between both nonspecific shoulder symptoms [Ohlsson et al. 1989] and nonspecific shoulder disorders [Sakakibara et al. 1995] and the number of pieces handled per hour or per day.

Repetition Characterized as Short Cycle Time
Chiang et al. [1993] found a significant association between a very short or repetitive cycle (<30 seconds or >50% spent repeating same task) and shoulder girdle pain.

Repetition Characterized Descriptively
Three studies by Ohlsson et al. found a significantly higher proportion of shoulder MSDs in exposed populations with work characterized as involving repetitive arm and hand movements than in referent populations [Ohlsson et al. 1989, 1994, 1995].

Repetition Combined with Static Shoulder Load
Except for the study by Sakakibara et al. [1995], in which the increased number of pears bagged per day was associated with an increased proportion of the work day spent with extreme shoulder flexion, the studies using measures of piece work or repetitive arm movements as the exposure outcome did not specify which joints or body regions participated in the repetitive action. Ohlsson et al. [1995] described the assembly work performed by the exposed population as combining repetitive arm movements with a static shoulder load. It is possible that the association between piece work, short cycles, or repetitive hand-arm movements and shoulder disorders reported by the other authors is related to a sustained, static load on the shoulder muscles as the upper arm is stabilized in a posture of mild to severe flexion or abduction, while repetitive movements are performed by the hand-wrist-forearm.

Temporal Relationship

In the prospective study by Kilbom et al. 1986; Kilbom and Persson 1987; and Jonsson et al. 1988 the number of shoulder elevations per hour was a strong predictor for a change to severe status at the 1- and 2-year follow-up evaluations. Although the change in status included problems in the neck and arm, as well as the shoulder, it is reasonable to assume that repetitive shoulder elevations would have had the greatest effect on disorders of the shoulder.

Several studies with a cross-sectional design used techniques to determine whether the health outcome of interest had occurred since, or was present during, exposure to hypothesized risk factor(s) of interest. Case definitions which required a positive physical examination finding [Chiang et al. 1993; Ohlsson et al. 1994, 1995] or where symptoms had occurred within the recent past [Chiang et al. 1993; Ohlsson et al. 1989, 1994] were designed to focus on disorders most likely to have been caused or aggravated by current work exposures.

Exposure-Response Relationship

Chiang et al. [1993] found a significant increasing trend in the prevalence of shoulder girdle pain from Group I (low force, low repetitiveness) to Group III (high force, high repetitiveness). However, the health outcome was not specific to shoulder disorders, and the exposure categories combine increasing repetitiveness—as defined by either less than a 30-second cycle time or a repeated task within the job cycle—and increasing forearm flexor muscle activity. Ohlsson et al. [1995] found that neck and shoulder disorders among assembly workers were significantly associated (p<0.05) with both the number of arm elevation movements from less than to greater than 60 degrees and the number of arm abduction movements from less than to greater than 60 degrees. Bjelle et al. [1981] found that the frequency of shoulder abduction or forward flexion (past 60 degrees) was significantly greater (p<0.005) for cases with neck-shoulder disorders than for controls.

In the study of assembly workers by Ohlsson et al. [1989], the number of pieces completed per hour was categorized as follows: slow: <100, medium: 100–299, fast: 300–699, very fast: >700. In this study, the ORs are shown in a figure, rather than reported in the text. Compared with the slow-paced group, the odds for symptoms of shoulder pain is approximately seven times that for those workers in the medium-paced group and approximately nine times that for those in the fast-June 26, 1997 pace group. While adjusting for age and length of employment, the OR for shoulder pain was significantly higher for the medium- and fast-paced groups than for the slow-paced group (p=0.0006). The OR for the very fast-paced group compared to the slow-paced group was between 1.0 and 2.0 and was not significantly different from the slow-paced group. The authors hypothesized that symptomatic workers may have self-selected out of the very fast paced jobs or that other unknown factors may have mitigated the effects of work pace.

When comparing fish industry workers to the reference population, Ohlsson et al. [1994] found that among those workers younger than age 45, the ORs for disorders of the neck and shoulders were significantly elevated and increased with duration of employment [0–5 years, OR 3.2 (95% CI 1.5–7.0); >5 years, OR 10 (95% CI 4.5–24)]. In their study of assembly workers, Ohlsson et al. [1989] found a statistically significant increase in the odds for pain in the shoulder with duration of employment (p=0.03) which was dependent on age. The increase with duration of employment had a steeper slope for younger (<35 years) assembly workers than for the older subgroup (i.e., among those workers employed for short durations, older women had more symptoms, and among those workers employed for long durations, younger women had more symptoms). This was thought to be a reflection of both survivor bias as well as the possibility that older new hires may have experienced a relatively more rapid onset of symptomatic problems than do younger women.

Coherence of Evidence

Repetitive movements of the upper extremity involving flexion or abduction of the glenohumeral joint would increase the frequency of effects such as fatigue and tendon circulation disruption hypothesized to occur as a result of such postures. These effects could be magnified by the addition of a hand-held load. Repetition may also be solely related to the development of tendinitis. In a laboratory study, Hagberg [1981] induced acute shoulder tendinitis in female subjects performing repetitive shoulder elevations for one hour. Six female students, ages 18–29, all developed shoulder tenderness (two with tendinitis) when exposed to 15 shoulder flexions (from 0 to 90 degrees) per minute for 60 minutes while holding up to 3.1 kg (6.4 lb) of weight.

Some of the significant associations reported may have been related to exposure to repetitive work in the distal upper extremity while the shoulder and upper arm were maintained in a static posture [Chiang et al. 1993; Ohlsson et al. 1989, 1994, 1995]. Winkel and Westgaard [1992] have pointed out that, “It is not possible to use the arm/hand without stabilizing the rotator cuff girdle and the glenohumeral joint. Therefore, work tasks with a demand of continuous arm movements generate load patterns with a static load component.”

The finding that the supra- and infraspinatus muscles were particularly prone to fatigue when subjects performed overhead work led Herberts et al. [1984] to hypothesize that the rotator cuff muscles may develop high intramuscular pressures at relatively low contraction levels. These high intramuscular pressures could lead to an impairment of intramuscular circulation, which could contribute to the early onset of fatigue. Intramuscular pressure increases with the muscle contraction level, and impaired circulation has been demonstrated at levels of contraction as low as 10–20 percent of maximal voluntary contraction (MVC). [Hagberg 1984].

The increased pressure in rotator cuff muscles and increased pressure on the supraspinatus tendon may trigger two different events that are both related to impaired microcirculation. The impaired microcirculation in the tendon may also result from tension within the tendon produced by forceful muscle contractions [Rathburn and Macnab 1970]. An inflammatory infiltrate with increased vascularity and edema within the rotator cuff tendons, especially the supraspinatus tendon may be a result of or a contributor to the process. If the inflammation process is sufficiently intense, then shoulder tendinitis may occur. If the process is less intense, and more chronic, then it may contribute to a degenerative process in the tendons of the rotator cuff. In the muscles of the rotator cuff, the impaired microcirculation may lead to small areas of cell death. A reasonable hypothesis is that repeated or sustained episodes of muscle ischemia result in localized cell death and persistent inflammation.

Neither of these proposed models for shoulder muscle pain or tendinitis suggest that all muscle activity is potentially harmful. Both muscles and tendons are strengthened by repeated activity if there is sufficient recovery time. However, the models present plausible mechanisms by which work tasks with substantial shoulder abduction could contribute both to shoulder pain and tendinitis.

There is evidence of a relationship between shoulder tendinitis and highly repetitive work. However, there are several limitations to the evidence. In the three studies for which the health outcome was shoulder tendinitis, the exposure combined repetition with awkward shoulder posture and/or a static shoulder load [English et al. 1995; Ohlsson et al. 1994, 1995]. Five out of the eight studies reviewed used either nonspecific shoulder disorders, nonspecific shoulder symptoms or combined neck-shoulder disorders as the health outcome.

Despite the limitations of the evidence, significant and positive relationships between repetitiveness, regardless of the measurement method, and shoulder MSDs or symptoms were found in all studies. Of the eight studies in which the effect of repetition was examined, three studies found ORs above 3.0 [Ohlsson et al. 1989, 1994, 1995] and three studies found ORs from 1.0 to 3.0 [Chiang et al. 1993; English et al. 1995; Sakakibara et al. 1995]. The remaining studies were prospective studies [Jonsson et al. 1988; Kilbom and Persson 1987] or studies that reported risk indicators other than OR [Bjelle et al. 1981].

In none of these studies is it likely that age, the most important personal characteristic associated with shoulder tendinitis and other shoulder disorders, or non occupational factors such as sports activities, caring for young children, or hobbies explained these associations. There is evidence of a relationship between shoulder tendinitis and highly repetitive work.

Back to top

Force

Definition of Force for Shoulder MSDs

Studies that examined force or forceful work or heavy loads to the shoulder, or described exposure as strenuous work involving the shoulder abduction, flexion, extension, or rotation that could generate loads to the shoulder region were also included. Most of the studies that examined force or forceful work as a risk factor for shoulder symptoms or tendinitis had several concurrent or interacting physical work load factors. However, there is still a need to summarize present knowledge about the relationships between forceful work and shoulder MSDs. This section summarizes that knowledge, while acknowledging that other factors can modify the response. Neck-shoulder disorders are discussed in Chapter 2.

Studies Reporting on the Association of Force and Shoulder Tendinitis

There are five studies which reported results on the association between force and adverse shoulder health outcomes (Table 3–2, Figure 3–2). The epidemiologic studies that addressed forceful work and shoulder MSDs tended to compare working groups by classifying them into broad categories based on an estimated amount of resistance or force of exertion and a combination of estimated rate of repetition [Andersen and Gaardboe 1993a; Chiang et al. 1993] or in terms of overall load [Herberts et al. 1984; Stenlund et al. 1992; Wells et al. 1983].

Studies Meeting the Four Evaluation Criteria
Chiang et al. [1993] studied workers in the fish processing industry. (This study was described in detail in the section on shoulder MSDs and repetition.) Chiang et al. [1993] did not report an exposure specific to the shoulder.

Studies Not Meeting the Four Evaluation Criteria
Andersen and Gaardboe [1993a] performed a cross-sectional study in which a cohort of sewing machine operators was compared to a random sample of women in the general population of the same region. Chronic shoulder pain was defined as a having experienced a continuous pain episode lasting more than 1 month and either daily pain or pain lasting more than 30 days in the same location within the previous year (per self-administered questionnaire). In order to compare the current exposure of sewing machine operators and those in the control group, the authors’ experience and knowledge of the jobs were used to assign job titles to exposure categories based on crude assessments of force and repetitiveness. High exposure was characterized as a combination of high repetitiveness (activity repeated several times per minute) and low or high force, or medium repetitiveness (activity repeated many times per hour) and high force. Medium exposure was characterized as medium repetitiveness and low force, or low repetitiveness (jobs with more variation) and high force. Those in teaching, academic, self-employed, or nursing professions were classified as low exposure. The exposure classification scheme in this study does not allow separation of the effects of force from those of repetition. More sewing machine operators than referents were considered to have high exposure (41% versus 15%), but more in the referent population were considered to be in the medium exposure group (44% versus 22%). Because the outcome of interest was duration of historical exposure, current exposure was included as an independent variable in multivariate regression analyses.

Herberts et al. [1984] added to the 1981 study by comparing the prevalence of supraspinatus tendinitis between plate-workers and office clerks. Tendinitis in welders was determined by a combination of self-reported symptoms and positive physical examination findings. The only information given regarding plate-work is that it is dynamic in character. It is presumed that plate-workers handled heavy loads more frequently than office clerks.

In a cross-sectional study, the prevalence of osteoarthrosis in the acromioclavicular joint, as determined by radiography, was compared among three groups of workers in the construction industry [Stenlund et al. 1992]. The three groups were bricklayers, rock blasters, and construction foremen. The foremen did not perform manual work currently, or in the past, and were considered the control population. A standardized interview was used to determine exposure factors,including job title and the sum of loads lifted during all working years (expressed in tonnes). Analyses were performed separately for right and left sides.

In a study of letter carriers, Wells et al. [1983] evaluated the effect of a load carried on the shoulder. Letter carriers, who carry a load and walk, were compared to gas meter readers (who walk without carrying a load) and postal clerks. Utilizing information from telephone interviews, points were assigned to symptom characteristics such as frequency, length of episodes, and interference with work ability. Case definition required a report of recurrent shoulder pain with greater than 20 points. A subset of letter carriers had experienced an increased load during the previous year. (The Postal Service had increased maximum weight carried from 25 to 35 pounds, but not all locations had implemented this change.)

Strength of Association—Force and Shoulder MSDs

The studies are presented in alphabetical order in Table 3-2. Results of studies where ORs, or other measures of association, were specifically associated with a measure of exposure, are presented in the section on Exposure-Response Relationship.

Andersen and Gaardboe [1993a] found that current work as a sewing machine operator was associated with chronic shoulder pain (OR 1.72, 95% CI 1.17–2.55). Using multiple logistic regression analysis with age, gender, and repetitiveness as covariates, Chiang et al. [1993] found that high force exertions measured in the forearm were associated with shoulder girdle pain (OR 1.8, 95% CI 1.2–2.5). When tested in the same model with force and repetition, the interaction term for force times repetition was also significant (OR 1.4, 95% CI 1.0–2.0). Two factors could have resulted in an underestimation of the strength of association: (1) no requirement that symptoms have started on current job meant that some symptomatic workers may have transferred to lower risk jobs, and (2) no matching of health status and exposure status by side (left, right, or both) may have caused non-differential misclassification. For supraspinatus tendinitis, Herberts et al. [1984] calculated a prevalence rate ratio (PRR) for plate-workers versus office clerks of 16.2 (90% CI 10.9–21.5) “under the assumption that missing data had the same characteristics as those considered.” The absence of specific exposure information was a major limitation of this study.

The age-adjusted OR associated with osteoarthritis of the acromioclavicular joint was 2.16 (95% CI 1.14–4.09) (right side) and 2.56 (95% CI 1.33–4.93) (left side) for manual construction workers versus foremen [Stenlund et al. 1992]. Because there was a lower participation rate among bricklayers and blasters, self-selection into the study because of having symptoms could have resulted in overestimation of the strength of association. While some of the items handled required a bilateral lift (e.g., jackhammer), other loads may have been specific to the right or left hand. Because the exposure measure did not separate load by sides, non-differential misclassification may have caused underestimation of the strength of association.

Consistency of Association: Force and Shoulder MSDs

Despite different outcome and exposure measures, all of the studies had positive associations. Each study used a different case definition, ranging from relatively mild symptoms to radiographic evidence of osteoarthritis, and a different measure of exposure. Chiang et al. [1993] used EMG measures of forearm flexor muscle activity. Wells et al. [1983] evaluated the effect of a direct load on the shoulder. Stenlund et al. [1992] used an estimate of the cumulative, lifetime load carried. Andersen and Gaardboe [1993a] compared sewing machine operators to a referent population. However, positive and significant associations were found, regardless of the measure of health outcome or exposure.

Temporal Relationship: Force and Shoulder MSDs

All of the studies of forceful exertions used a cross-sectional study design. To increase the likelihood that shoulder symptoms were caused or aggravated by current exposure, Chiang et al. [1993] required that symptoms had occurred within the previous 30 days.

Wells et al. [1983] used several analytical methods to increase confidence in a relationship between carrying the increased load and having shoulder disorders. The use of age, the number of years on the job, and previous heavy work experience as covariates when performing analysis of covariance helped ensure that the difference in the proportion of shoulder disorders between letter carriers with and without the increased load was related to current exposure rather than past peak exposures or cumulative duration. Although baseline symptom status in the group with the increased load could not be obtained, there was no significant difference in the prevalence of shoulder problems between the two groups when results were adjusted for the amount of weight currently carried. Therefore, the difference in symptom prevalence was likely related to the load increase rather than prior differences in symptom status. The cross-sectional studies are consistent with exposure occurring before the onset of the shoulder MSDs.

Exposure-Response Relationship

When sewing machine operators were compared with an external control population, there was a trend of increasing ORs for chronic shoulder pain with increasing duration of work as a sewing machine operator [Andersen and Gaardboe 1993a]. The OR for 0–7 years was 1.38 (95% CI 0.86–2.39), for 8–15 years it was 3.86 (95% CI 2.29–6.50), and for >15 years it was 10.25 (95% CI 5.85–17.94), while controlling for other factors including age and current exposure.

Chiang et al. [1993] found a significant increasing trend in the prevalence of shoulder girdle pain from Group I (low force, low repetitiveness) to Group III (high force, high repetitiveness). However, the health outcome is not specific for shoulder tendinitis and the exposure categories combine increasing force, as measured in the forearm flexor muscles, and increasing repetitiveness.

In the study of bricklayers and blasters, and acromioclavicular osteoarthritis, Stenlund et al. [1992] found that for the left side, ORs increased with the level of lifetime load lifted. For a lifetime load of 710–24,999 tonnes versus less than 710 tonnes, the left side OR was 7.29 (95% CI 2.49–21.34), and for greater than 25,000 tonnes versus less than 710 tonnes, the left side OR was 10.34 (95% CI 3.10–34.46).

For severe, but not disabling, shoulder pain, the OR for letter carriers versus postal clerks was 3.6 (95% CI 1.8–7.8) [Wells et al. 1983]. For those letter carriers who had experienced a weight load increase within the previous year, versus postal clerks, the OR was 5.7 (95% CI 2.1–17.8). Furthermore, letter carriers who had experienced the weight load increase had significantly more shoulder problems than those whose bag weight had not been increased. If letter carriers tend to keep the mail-bag strap on one shoulder, the fact that the side of the load was not matched with the side of the shoulder problem could have resulted in non-differential misclassification and an underestimation of the strength of association. However, some of the health effects may have been related to activation of contralateral muscles involved in stabilizing the shoulder girdle [Winkel and Westgaard 1992].

Coherence of Evidence

High shoulder muscle force requirements can cause increased muscle contraction activity, which may lead to an increase in both muscle fatigue and tendon tension, and may possibly impair microcirculation as well.

Force may also be related to a static load on shoulder muscles. Sjígaard et al. [1988] found that muscular fatigue will occur at EMG levels as low as 5% of maximal voluntary contraction (MVC) if sustained for 1 hour. Other studies have demonstrated that when the period of muscle contraction is extended to more than an hour, the endurance limit of force may be as low as 8% MVC [Jonsson 1988]. Workers performing repetitive work with the hands and wrists, while maintaining static upper arm elevation may experience fatigue even at low load levels. Jonsson [1988] reported that many constrained work situations are characterized by static load levels near or exceeding 5% MVC, even when characterized by a fairly low mean muscular load.

Because the five studies reviewed had a considerable diversity of exposure assessment approaches and health outcomes, there is insufficient epidemiologic evidence to conclude that forceful exertions are associated with rotator cuff or bicipital tendinitis. The one study that used shoulder tendinitis as the health outcome reported a strong association related to job category (OR for plate-workers versus clerks: 16.2 (95% CI 10.9–21.5), but did not describe or measure specific exposure risk factors [Herberts et al. 1984]. One of the reviewed studies did present evidence for an association between acromioclavicular osteoarthrosis and cumulative, lifetime load on the shoulder muscles [Stenlund et al. 1992]. Another study reported a significant association between severe shoulder pain and a direct shoulder load [Wells et al. 1983].

Back to top

Posture

Definition of Awkward Posture for Shoulder MSDs

For the shoulder, a relaxed, neutral posture is one in which the arm hangs straight down by the side of the torso. As the arm is flexed, abducted, or extended, the included angle between the torso and the upper arm increases. In one study, postures in which the included angle was equal to or greater than 45 degrees required substantial supraspinatus muscle activity, while deltoid muscle activity underwent a pronounced increase as the angle of shoulder flexion or abduction increased from 45 to 90 degrees [Herberts et al. 1984]. As the arm is elevated, the space between the humeral head and the acromion narrows such that mechanical pressure on the supraspinatus tendon is greatest between 60 and 120 degrees of arm elevation [Levitz and Iannotti 1995]. While there is a continuum of severity from an included angle of 30 degrees to a maximally abducted arm, postures with shoulder abduction or flexion past 60 degrees are considered awkward.

Studies Reporting on the Association of Awkward Postures and Shoulder MSDs

Six of the reviewed studies reported results on the association between awkward postures and shoulder tendinitis [Baron et al. 1991; Bjelle et al. 1979; English et al. 1995; Herberts et al. 1981; Ohlsson et al. 1994, 1995] (Table 3-3, Figure 3-3). Seven additional studies reported results on the association between awkward postures and non-specific shoulder disorders [Sakakibara et al. 1995], non-specific shoulder symptoms [Hoekstra et al. 1994; Milerad and Ekenvall 1990; Schibye et al. 1995] combined neck-shoulder disorders [Bjelle et al. 1981; Jonsson et al. 1988; Ohlsson et al. 1995] or combined neck-shoulder symptoms [Kilbom and Persson 1987].

Studies Meeting the Four Evaluation Criteria
Four studies met all four of the evaluation criteria.

Using a prospective study design, Jonsson et al. [1988] assessed the health and exposure status of 69 electronics manufacturing plant employees at the beginning of the study and after one and two years. Employees who dropped out before completion of the study were compared to those who fully participated; there was no significant difference in medical status, working technique, or work history. Employees who had upper extremity disorders resulting in a physician visit or sick leave were excluded from the initial study group. The dependent variables related to health status were of two types: a change in symptom severity and being symptom free. Symptom status was assessed by interview and a physical examination by a physiotherapist. The symptoms severity index compiled data from the five body regions combined and was not specific for the shoulder region. Because the exposure was determined by direct observation for each individual, and clearly separated ergonomic risk factors by body region, it was still possible to evaluate associations likely to specifically involve the shoulder.

Kilbom and Persson [1987] and Kilbom et al. [1986] performed a study in which female employees in the electronics manufacturing industry were observed for a 2-year period. The health outcome of fatigue, ache, or pain in the neck, shoulder, or arm regions was based on symptoms information. Symptom severity was coded on the basis of its character, frequency, and/or duration. Changes in severity status at follow-up evaluations were used as the dependent variables in multiple regression analyses. Neck, shoulder, and upper arm posture was determined by computerized analysis (VIRA) of videotapes of individuals. Although the health outcome combined symptoms from different body regions, knowledge of biomechanical theory can be used to identify significant predictors related to the shoulder symptom severity.

Two of the reviewed studies in which tendinitis was the health outcome are Ohlsson et al. [1994, 1995]. For both studies, the authors reported that the examiners could not be completely blinded to exposed versus referent status, but that a standard protocol was followed and observer bias was likely to have been minimal. Because examiners were blinded to objective exposure measures, analyses testing associations between neck-shoulder disorders and specific postures would not have been biased [Ohlsson et al. 1995].

In a cross-sectional study, women in the fish industry were compared to a control population of women employed in municipal workplaces in the same towns [Ohlsson et al. 1994]. Diagnoses of shoulder disorders (e.g., tendinitis, acromioclavicular syndrome, frozen shoulder) were made on the basis of symptoms determined by interview and a physical exam. Exposure evaluation of each work task held by the fish industry population was evaluated with ergonomic workplace analysis (EWA). Ten different factors were rated on a scale from 1 to 5 and the combined ratings were used as a profile of the work task. Based on this profile, the authors reported that fish industry work was found to be “highly repetitive” and include “poor work postures.”

Ohlsson et al. [1995] compared a group of women who performed industrial assembly work to a referent group of women from a nearby town who were employed in jobs characterized as having varied and mobile work tasks. One examiner assessed signs and symptoms. The examiner was blinded to specific exposure information, but not completely blinded to factory worker versus referent group status. Shoulder tendinitis included supraspinatus, infraspinatus, and bicipital tendinitis. Another health outcome combined neck and shoulder disorders (tension neck, cervical syndrome, thoracic outlet syndrome, frozen shoulder, tendinitis, and acromioclavicular syndrome). In a descriptive assessment, it was reported that the work tasks in the study group involved repetitive arm movements with static muscular work of the neck and shoulder muscles. The percentage of time spent in specific upper arm postures was determined from videotaped observations of 74 (out of 82) workers. The average result from two independent videotape analyses was used. Posture category demarcations included 0, 30, and 60 degrees for arm elevation, and 30, 60, and 90 degrees for arm abduction.

Studies Not Meeting the Four Evaluation Criteria
Summaries of studies that specifically evaluated associations with shoulder tendinitis are presented next [Baron et al. 1991; Bjelle et al. 1979, 1981; English et al. 1995; Herberts et al. 1981]. Summaries of other studies are presented in alphabetical order.

In the study by Baron et al. [1991], grocery store workers who performed the job of checker were compared to a non-checker group that performed a variety of other jobs (e.g., general stocking, working in the produce section, the bakery, salad bar, pharmacy, and courtesy counter). There was a low participation rate among non-checkers (55%), which could have resulted in an underestimation of the OR for checkers if symptomatic non-checkers were more likely to participate than those non-checkers without symptoms. The authors evaluated this possibility by performing a sufficient number of telephone interviews with non-participants to raise the non-checker participation rate for interviews to 85%. The OR for shoulder symptoms among the full participant population was similar to the OR for the full participant plus telephone interview population. The case definition was shoulder symptoms lasting at least one week or occurring at least once per month during the previous year that began while the worker was performing her current job and positive physical examination findings consistent with a shoulder tendinitis. Detailed descriptions of the checker jobs were presented based on both on-site and videotape analyses of a few representative workers per workstation. No videotaping of non-checkers was performed. Shoulder flexion and/or abduction (>90 degrees) was observed during a variety of different tasks performed by the checkers. The exposure measures used in statistical analyses were: (1) checker versus non-checker and, (2) for exposure-response assessment among checkers, the total number of months and the number of hours per week working as a checker.

Bjelle et al. [1979] compared cases with persistent shoulder pain to controls employed as manual workers. After an extensive medical evaluation, a diagnosis of bicipital and/or supraspinate tendinitis was made for a majority (12/17) of the cases. Physical workload was categorized in relation to sitting or standing posture, weight lifting, and carrying. The work height of the hands was categorized based on position relative to the acromion height, per individual. Placement of workers into exposure categories was determined by the combined efforts of each study participant and a physician.

Bjelle et al. [1981] compared cases with acute, non-traumatic shoulder-neck pain to age- and sex-matched, paired controls. An extensive physical examination was performed and workers with inflammatory rheumatoid diseases were excluded. To determine exposure, each case and control was filmed and a biomechanical analysis was performed to determine the duration and frequency of shoulder abduction or forward flexion greater than 60 degrees.

In a study by English et al. [1995], cases determined by medical diagnosis, and controls were selected from patients evaluated at specified orthopedic clinics. For statistical analyses, all diagnoses were grouped by anatomical site. The diagnoses for shoulder cases included rotator cuff injury, rupture of the long head of the biceps, shoulder capsulitis, and symptomatic acromioclavicular arthritis. It is assumed that shoulder tendinitis was included in this group. Exposure measures were determined by a standardized interview conducted by an interviewer who was, “unaware of the case-control status of the individual wherever this was possible.”

In a study by Herberts et al. [1981], the prevalence of supraspinatus tendinitis was compared between welders and office workers. Tendinitis cases were based on a combination of symptoms reported on a nurse-administered questionnaire and a positive physical examination done by a physiotherapist. For welders, an “experienced physiotherapist” rated work-load on the shoulder as low, high, or very high; no description of the classification scheme was given.

Hoekstra et al. [1994] evaluated government office workers at two locations. The case definition for shoulder symptoms was symptoms that began after starting current job, lasting greater than one week, or occurring at least once per month during the past year with an intensity greater than two on a five point scale, and no preceding acute, non-occupational injury. A self-administered questionnaire was used to determine exposure to factors such as “perceived adequacy of adjustment of video display terminal (VDT).” Walk-through ergonomic evaluations of factors such as workstation surface height and furniture adjustability were used to provide descriptive differences between the two office locations.

Milerad and Ekenvall [1990] compared the prevalence of self-reported, non-specific shoulder symptoms between dentists and pharmacists. Dentistry, as a profession, was described as work “with the arms abducted and unsupported” whereas, pharmacists had “physically light and varied work.”

In a prospective study by Sakakibara et al. [1995], the health status of a group of women farm workers was assessed during the performance of two different tasks, with a 1-month interval between the tasks. The health outcome was defined by self-assessed symptoms of shoulder stiffness and pain and a physical examination for muscle tenderness and joint pain on movement. Whether the examining physician was aware of the prior hypothesis regarding differing exposures between the two tasks (bagging pears versus bagging apples) was not stated. Exposure was based on self-report of the number of hours per day spent bagging, the number of pears or apples bagged per day, and the total number of days spent bagging each fruit. One worker was observed for 3 hours while performing each bagging job, with repeated goniometer measures of shoulder forward flexion angles done each minute. While there was no difference in the total number of days or number of hours per day spent bagging each fruit, significantly more pears than apples were bagged per day. The proportion of time spent with the angle of shoulder forward flexion greater than 90 degrees was significantly larger when bagging pears (75%) than when bagging apples (41%).

Schibye et al. [1995] performed a prospective study of a population of sewing machine operators in which the change in self-reported shoulder symptom status was compared with those sewing machine operators who continued to work and those operators that moved into other occupations (e.g., shop assistant, health care worker, and fishing industry worker).

Strength of Association—Awkward Posture and Shoulder MSDs

Results are presented in the section on Exposure-Response Relationship (Table 3-3, Figure 3-3) for studies where ORs, or other measures of association, were specifically associated with a measure of exposure.

Using data presented in the study by Ohlsson et al. [1994], for supraspinatus, infraspinatus, or bicipital tendinitis, the PRR for working in the fish industry (repetitive work, poor posture) versus the referent population was calculated as 3.03 (95% CI 2.0–4.6). For shoulder tendinitis alone, the PRR was calculated as 3.5 (95% CI 2.0–5.9). In the same study, the authors also interviewed a large group of former fish industry employees and found that a quarter of those workers who left employment had done so because of problems with their neck or upper limbs. This proportion increased with age and also occurred after a shorter duration of employment among the oldest workers. This evidence of a survivor bias highlights the importance of controlling for age. Higher risks were found for the workers less than 45 years old and these risks may be a more accurate assessment of the true risk.

Using data presented in the study by Ohlsson et al. [1995], for supraspinatus, infraspinatus, or bicipital tendinitis, the OR for being an assembly worker (repetitive arm movements with static load on shoulders) versus the referent population was 4.2 (95% CI 1.35–13.2). For neck-shoulder disorders, the OR for being an assembly worker versus the referent group was 5.0 (95% CI 2.2–11.0).

For shoulder disorders consistent with tendinitis, Baron et al. [1991] found that the OR for being a checker versus a non-checker was 3.9 (95% CI 1.4–11.0). Because non-checkers also performed work requiring awkward postures, the reported OR may underestimate the risk for checkers. Short stature (< 5'2") was associated with an elevated, but not statistically significant, OR for shoulder disorders (2.1, 95% CI 0.7–6.9). Because work-station height was fixed, it is likely that short stature workers experienced more frequent and/or more severe episodes of shoulder flexion and/or abduction.

The OR for work performed at or above acromion height (i.e., hands above the shoulder) versus work performed below acromion height was 10.6 (95% CI 2.3–54.9) [Bjelle et al. 1979]. In this study, all cases were patients who had been examined by the same physician. Placement of cases and controls into exposure categories was performed by each subject in collaboration with a physician who “had personal knowledge of the work involved in each case.” Whether or not the physician who performed the clinical examinations is the same person as the physician involved in exposure classification is not stated. If this was the same person, a potential bias towards assigning cases to higher exposure categories could have resulted in overestimation of the strength of association. However, two other factors could have resulted in an underestimation of the strength of association. The exposure outcome was based on current work load without any stated restriction that cases’ symptoms had started on their current job. If some of the cases, defined as having problems non-responsive to therapy lasting longer than 3 months, had transferred to a lower risk job, the strength of association may have been underestimated. Location of the disorder and exposure were not matched by side (left, right, or both) and this would have caused non-differential misclassification, resulting in some underestimation of the strength of association.

English et al. [1995] found that the risk of having a medically diagnosed shoulder condition was increased by repeated shoulder rotation with an elevated arm (OR 2.30, p<0.05). Non-differential misclassification due to a combination of complicated exposure definitions using a questionnaire, and the fact that analyses did not relate health outcomes and exposure on a temporal basis, or by left/right side, may have caused an under-estimate of the strength of association.

For supraspinatus tendinitis, Herberts et al. [1981] found that the PRR for welders (characterized as using awkward postures to perform overhead work) versus clerks was 18.3. However, in determining this PRR, the authors performed extrapolation based on an assumption that, “the drop-out group does not deviate from the examined group,” without any data to support this assumption. To determine a more reliable indicator of risk, unextrapolated data presented in the study were used to calculate a crude OR=8.3 (95% CI 0.63–432). The office clerks were older than the welders, so that confounding by age may have caused an under-estimation of the strength of association.

In a study of teleservice employees, there was an association between reporting shoulder symptoms and working at one location versus another location; the OR was 4.0 (95% CI 1.2–13.1) [Hoekstra et al. 1994]. Descriptive differences between workstation design at the two locations provided a plausible explanation for this finding. At the higher risk location, the workstation surface was too high to serve as a keyboard support, there were nonadjustable chairs, and it was observed that “nonadjustable furniture universally promoted undesirable postures (i.e. elevated arms, hunched shoulders).” Having shoulder symptoms was also positively associated with using a non-optimally adjusted desk height (OR 5.1, 95% CI 1.7–15.5) and a non-optimally adjusted VDT screen (OR 3.9, 95% CI 1.4–11.5). Because exposure was self-reported without any indication of whether or not study participants had received education regarding good VDT workstation design, the phrase, “non-optimally adjusted,” may have had various meanings to the study participants. This could have caused non-differential misclassification of exposure and an under-estimation of the strength of association. On the other hand, a possible reporting bias related to self assessment of both symptoms and exposure could have resulted in an overestimation of the strength of the association. A plausible explanation for the association between shoulder symptoms and these workstation design factors is that the non-optimally adjusted workstation components forced the employees to abduct the upper arms and/or hunch the shoulders.

For shoulder symptoms without concomitant neck symptoms, Milerad and Ekenvall [1990] found that the OR for being a dentist (work with both arms abducted) versus being a pharmacist was 3.8 (95% CI 1.2– 10.3). As with most cross-sectional studies, the survivor bias may have resulted in underreporting of the strength of exposure. Conversely, the exposed group may have had better recall of self-reported symptoms with a resultant overestimation of the OR.

In the study of farm workers by Sakakibara et al. [1995], the point prevalence of muscular tenderness in the shoulder regions (per physical examination) was significantly higher when performing pear bagging (48%) than when performing apple bagging (29%). The proportion of time spent with the shoulder in forward flexion greater than 90 degrees was significantly larger when bagging pears (75%) than when bagging apples (41%). Whether or not there was a recovery period between pear and apple bagging is not stated. If there was insufficient recovery after pear bagging, persistent muscle tenderness or increased susceptibility may have caused underestimation of the difference in shoulder disorder prevalence between these two work tasks.

With the exception of the study by English et al. [1995], in which the strength of association may have been underestimated, for the studies in which the health outcome was shoulder tendinitis [Baron et al. 1991; Bjelle et al. 1979; Herberts et al. 1981; Ohlsson et al. 1994, 1995], the magnitude of association was strong. ORs ranged from 2.0 to 10.6. In none of these studies is it likely that nonoccupational factors such as sports activities or personal characteristics such as age explain these associations.

Consistency of Association

All but one of the reviewed studies relevant to posture and shoulder disorders found a positive association between shoulder disorders or shoulder symptoms and awkward shoulder posture. Awkward postures were consistently described as overhead work, arm elevation, and specific postures relative to degrees of upper arm flexion or abduction. This association was found in cross-sectional, case-control, and prospective studies among a great variety of types of work performed.

Temporal Relationship

It is important to determine whether symptoms or MSDs occur as a consequence of work-related exposures. This can be done most clearly with a prospective study design.

In the study by Jonsson et al. [1988], the percent of the work cycle spent with the shoulder elevated was negatively associated with remaining healthy (symptom free). Because workers with pre-existing shoulder conditions were excluded from study participation, the onset of new symptoms may have been associated with the daily and/or cumulative duration of exposure to elevated shoulder postures. In the study by Kilbom and Persson [1987], three of the work exposure variables that were strong predictors for a change to severe status at the 1- and/or 2-year follow-up evaluations were related to shoulder posture: (1) percent of work cycle time with arm abduction greater than 30 degrees, (2) percent of work cycle time with arm abduction greater than 60 degrees, and (3) percent of work cycle time with arm extension.

A few studies utilized techniques to improve the ability to detect possible relationships despite a cross-sectional study design. The case definition used by Baron et al. [1991] required that symptoms began while the worker was on the currently held job. Bjelle et al. [1979] filmed and analyzed the job held at the time the worker/case became symptomatic. The results of the prospective studies are similar to the cross-sectional studies. There is no evidence that shoulder disorders predicted the onset of exposure.

Exposure-Response Relationship

The level of an exposure can be described in two different ways. It may be related to the amount of exposure over a relatively short time period, such as a day or week, or it may be related to cumulative or life-time exposure over a number of years. Studies that tested associations related to daily or weekly variation in exposure are presented first, followed by studies that evaluated cumulative exposure by using independent variables, such as duration of employment or estimated lifetime exposure.

Four studies have some evidence of exposure-response relationships. Baron et al. [1991] found a significantly larger OR for shoulder disorders among employees working greater than 25 hours/wk as a checker compared to those working less than 20 hours/wk. Bjelle et al. [1981] found that the duration of hours worked per day with the shoulder flexed or abducted >60 degrees was significantly higher (p<0.025) for cases with neck-shoulder disorders than for controls. Ohlsson et al. [1995] found that neck and shoulder disorders among assembly workers were significantly associated (p<0.05) with the percent of time spent with the shoulder abducted or elevated >60 degrees. Although it is more difficult to detect associations with homogenous exposure, this association was significant despite very little variability in exposure to arm abduction greater than 60 degrees. While the analysis among assembly workers was performed without controlling for age, there is no evidence to suggest that older workers were more likely to be on high exposure jobs, and therefore a substantial bias is unlikely.

When comparing fish industry workers to the reference population, Ohlsson et al. [1994] found that among those workers younger than 45 years, the ORs for disorders of the neck and shoulders were significant and increased with duration of employment (0–5 years, OR 3.2; 95% CI 1.5–7.0) (>5 years, OR 10; 95% CI 4.5–24). Ohlsson et al. [1995] found a decreasing trend when they compared OR after stratifying the factory workers by employment duration (<10 years, OR 9.6; 10–19 years, OR 4.4 and >20 years: 3.8). Given the cross-sectional study design, this finding could be an artifact caused by the survivor bias (i.e., workers with disorders left, while symptom-free ‘survivors’ stayed; see Table 3-5). The assumption of a survivor bias is based on the finding that 28% of a group of former assembly workers reported pain in the musculoskeletal system as their reason for leaving employment at the factory. In the study by Schibye et al. [1995], improvement in shoulder symptoms among those who were no longer sewing machine operators appeared greater at follow-up, but was not significant. The fact that many of those who left sewing jobs moved into industries such as health care and fishing, where awkward postures and high force loads may occur, might explain why a large difference between sewing machine operators and non-sewing machine operators was absent. These four studies provide some support for the relationship between shoulder abduction and shoulder MSDs.

Coherence of Evidence

Discussions of the probable influence of workplace exposure factors in the pathophysiology of localized muscle fatigue, myalgia, and tendinitis have been presented by a number of authors [Bjelle et al. 1981; Hagberg 1984; Herberts and Kadefors 1976; Herberts et al. 1984; Levitz and Iannotti 1995]. Posture is important: when the arm is raised or abducted, the muscle activity in supraspinatus and other muscles increases, and the supraspinatus tendon comes in contact with the undersurface of the acromion. The mechanical pressure on the tendon from the acromion is greatest between 60 and 120 degrees of arm elevation. [Levitz and Iannotti 1995]. The degree of upper arm elevation is also important in the onset and intensity of localized muscle fatigue in the trapezius, deltoid, and rotator cuff muscles. [Hagberg 1981; Herberts and Kadefors 1976; Herberts et al. 1984]. In a laboratory study, EMG signals from these muscles were analyzed. The supraspinatus muscle was found to be highly active at ò45 degrees of abduction. The deltoid muscle underwent a pronounced increase in activity as shoulder flexion or abduction increased from 45 to 90 degrees [Herberts et al. 1984]. The earlier sections on Coherence of Evidence also discussed the rate of fatigue and role of impaired micro-circulation in shoulder tendinitis.

Overall, there is epidemiologic evidence for a relationship between repeated or sustained shoulder postures with more than 60 degrees of flexion or abduction and shoulder MSDs. There is evidence for both shoulder tendinitis and nonspecific shoulder pain. The evidence for increased risk of MSDs due to specific shoulder postures is strongest when there is a combination of exposures to several physical factors such as force and repetitive work. An example of this combination would be holding a tool while working overhead. The strength of association was positive and consistent in the six studies that used diagnosed cases of shoulder tendinitis, or a combination of symptoms and physical findings consistent with tendinitis, as the health outcome [Baron et al. 1991; Bjelle et al. 1979; English et al. 1995; Herberts et al. 1981; Ohlsson et al. 1994, 1995]. Only one [Schibye et al. 1995] of the thirteen studies failed to find a positive association with exposure and symptoms or a specific shoulder disorder. However, in this study discontinuing employment as a sewing machine operator was associated with a reduction in neck and shoulder symptoms. While most of the studies that considered specific shoulder postures as an exposure variable were cross-sectional, the two prospective studies found that the percent of work cycle spent with the shoulder elevated [Jonsson et al. 1988] or abducted [Kilbom et al. 1986; Kilbom and Persson 1987] predicted change to more severe neck and shoulder disorders. While there is insufficient evidence to develop a quantitative exposure-disorder relationship, three studies reported a significant association with shoulder flexion or abduction greater than 60 degrees [Bjelle et al. 1981; Kilbom and Persson 1987; Ohlsson et al. 1995]. Among the studies for which shoulder tendinitis was the health outcome, the largest ORs were associated with work above acromion height [Bjelle et al. 1979;

Herberts et al. 1981]. These results are consistent with the current models for the pathophysiology of shoulder tendinitis and stressful shoulder muscle activities. In none of these studies does “age,” an important personal characteristic associated with shoulder tendinitis, explain the positive results. Most of the studies controlled for a variety of confounders, such as occupational sports activities in their analyses. In summary, there is evidence that repeated or sustained shoulder abduction or flexion is associated with shoulder tendinitis, and the evidence is stronger for highly repetitive, forceful work.

Back to top

Vibration

Three of the studies evaluated exposure to low-frequency vibration found in industrial settings (Table 3-4, Figure 3-4). Because of the small number of studies, the full outline used for the sections on repetition, force, and posture will not be repeated here. The study by Stenlund et al. [1992] is summarized in the section on force. Vibration exposure occurred in one of the three job categories: rock blaster. The exposure outcome, lifetime exposure to vibration expressed in hours, was determined from a weighted summary of the number of self-reported hours using specific tools. However, because the rock blaster job category was also the only one where workers performed heavy lifts several times per day, the authors concluded that, “vibration exposure is indivisible from static load and heavy lifting in the present data.” When both cumulative lifting exposure and cumulative vibration exposure were included in the same multivariate model of an association with acromioclavicular osteoarthrosis, the OR for lifting and right- side osteoarthrosis remained significant while the weaker ORs for vibration became non-significant.

In the study by Stenlund et al. [1993], the same population of bricklayers, rock blasters, and foremen described in Stenlund et al. [1992] were evaluated to determine whether signs of tendinitis or muscle attachment inflammation in the shoulders were related to lifetime work load, years of manual work, lifetime exposure to vibration, or job title. The case definition for “signs of shoulder tendinitis” was pronounced (i.e., grade 3 out of 3) pain upon palpation of the muscle attachment or pronounced pain in response to isometric contraction of any of the rotator cuff muscles or the biceps muscle. The case definition of “clinical entity of tendinitis” was “signs of shoulder tendinitis” plus the subject’s report of shoulder pain during the past year. Using multivariate models that included age and hours spent in arm intensive sports activities, a significant association with cumulative vibration exposure was found when it was tested in isolation from the other exposure variables. For “clinical entity of tendinitis” the OR for the left side was 1.86 (95% CI 1.00–3.44) and the OR for the right side was 2.49 (95% CI 1.06–5.87).

For “signs of shoulder tendinitis” the OR for the left side was 1.66 (95% CI 1.06–2.61) and the OR for the right side was 1.84 (95% CI 1.10–3.07). When cumulative vibration exposure was tested in the same model with cumulative lifting load, significant associations were not found for either variable. Several factors could have resulted in an underestimation of the strength of association: (1) bricklayers or rock blasters with tendinitis may have been more likely to leave their jobs than foremen, (2) subjects may have had difficulty recalling exposure throughout their

lifetimes, (3) the inability to separate exposure by left and right sides. These factors may have caused non differential misclassification. Most important is the authors’ observation that vibration exposure occurred through the used of hand-held, heavy tools (e.g., jack-hammers) and thus is intertwined with exposure to a static load on the shoulders (from stabilizing the upper extremity while using the tool) as well as being associated with the heavy lifting tasks performed by rock blasters.

In a cross-sectional study by Burdorf and Monster [1991], riveters and control subjects in an aircraft company were investigated for vibration exposure and self-reported symptoms of pain or stiffness in the shoulder. Riveters were exposed to hand-arm vibration from working with hand drills, riveting hammers, bucking bars, and grinders. Controls were manual workers selected from the machine shop, maintenance, and welding departments in the same factory. In order to focus on the effect of vibration alone, a walk-through survey was performed to confirm that there were “no striking differences in dynamic and static joint loads during normal working activities.” Participation was 76% among riveters and 64% among controls. An analysis of non-respondents revealed that controls with health complaints were more likely to have participated than those without, while riveters with health complaints were less likely to have participated. The health outcome, determined by a self-administered questionnaire, was shoulder pain or stiffness occurring for at least a few hours during the prior year. Only subjects who reported having no symptoms before starting their present work were included in logistic regression analyses. The vibration transmitted by hand-tools was measured and weighted according to International Standards Organization (ISO) standards. Tool vibration profiles and time-work studies of riveters and controls were used to determine daily vibration exposure for each group. For riveters, on the basis of daily tool operating time, the equivalent frequency-weighted acceleration for a period of 4 hours was 2.8 m s ^-2. For controls, it was 1.0 m s ^-2. Using a multiple logistic regression model that included age, there was a weak association between shoulder symptoms and the number of years riveting (0.05< p<0.10). When the age-adjusted ORs for riveters compared to controls were plotted by the duration (in years, from 0 to 20) of riveting, the slope for shoulder symptoms was very gradual, with ORs ranging from 1.0 to 2.0. While the results of the analysis of non-respondents described above suggest that the strength of association may have been underestimated, the reported associations are weak and it is unlikely that the response bias would have resulted in a large increase in the magnitude of association.

There is insufficient evidence for an association between shoulder tendinitis and exposure to segmental vibration. In four separate evaluations, stratified by “signs of tendinitis” (positive physical examination findings), “clinical entity of tendinitis” (signs plus symptoms), left and right side, Stenlund et al. [1993] found an association between shoulder tendinitis and vibration exposure to segmental vibration; the range of ORs was from (OR for right side 1.66, 95% CI 1.06–2.61) (OR for left side 1.84, 95% CI 1.10–3.07). However, work with vibration exposure also placed a large, static load on shoulder muscles so that the effects of forceful shoulder muscle exertions could not be separated from vibration.

Back to top

Role of Confounders

Shoulder MSDs are multifactorial in origin and may be associated with both occupational and non-occupational factors. The relative contributions of these covariates may be specific to particular disorders. For example, the confounders for non-specific shoulder pain may differ from those for shoulder tendinitis. Two of the most important confounders or effect modifiers for shoulder tendinitis are age and sport activities. Most of the shoulder studies considered the effects of age in their analysis. Some studies considered sport activities [Baron et al. 1991; Stenlund et al. 1993; Jonsson et al. 1988; Kilbom et al. 1986]. Some studies also used multivariate methods to simultaneously adjust for several confounders or effect modifiers. For example, Ohlsson et al. [1995] found that for shoulder/neck diagnoses, repetitive work was the strongest predictor 4.6 (95% CI 1.9–12); age, muscle tension, and stress/worry tendency were also significant predictors. It is unlikely that the majority of the positive associations between physical exposures and shoulder MSDs are due to the effects of non-work confounders.

Back to top

Conclusions

There are over 20 epidemiologic studies that have examined workplace factors and their relationship to shoulders (MSDs). These studies generally compared workers in jobs with higher levels of exposure to workers with lower levels of exposure, following observation or measurement of job characteristics. Using epidemiologic criteria to examine these studies, and taking into account issues of confounding, bias, and strengths and limitations of the studies, we conclude the following:

There is evidence for a positive association between highly repetitive work and shoulder MSDs. The evidence has important limitations. Only three studies specifically addressed the health outcome of shoulder tendinitis and these studies investigated combined exposure to repetition with awkward shoulder postures or static shoulder loads. The other six studies with significant positive associations dealt primarily with symptoms. There is insufficient evidence for a positive association between force and shoulder MSDs based on currently available epidemiologic studies. There is epidemiologic evidence for a relationship between repeated or sustained shoulder postures with greater than 60 degrees of flexion or abduction and shoulder MSDs. There is evidence for both shoulder tendinitis and nonspecific shoulder pain. The evidence for specific shoulder postures is strongest where there is combined exposure to several physical factors like holding a tool while working overhead. The strength of association was positive and consistent in the six studies that used diagnosed cases of shoulder tendinitis, or a combination of symptoms and physical findings consistent with tendinitis, as the health outcome. Only one [Schibye et al. 1995] of the thirteen studies failed to find a positive association with exposure and a specific shoulder disorder or symptoms of a shoulder disorder.

This is consistent with the evidence that is found in the biomechanical, physiological, and psychosocial literature.

There is insufficient evidence for a positive association between vibration and shoulder MSDs based on currently available epidemiologic studies.

Back to top