Painful shoulder conditions are common and often persistent, and tend to cause prolonged disability. According to recent population-based studies, every sixth adult suffers from chronic shoulder pain, and more than half of those with current shoulder symptoms still have symptoms 3 years later.1^–3 Prevalence and persistence rates have been even higher in studies among occupational populations exposed to shoulder loading work.4 5 In 2004, shoulder disorders caused the longest absences from work, compared with other occupational musculoskeletal injuries and illnesses in the United States.6

Work is a major source of hazardous exposures affecting the shoulder, and also the most feasible target of preventive measures. Work-related physical exposures linked to shoulder disorders are heavy workload, working in awkward postures, repetitive movements and vibration, and particularly their combination.7^–10 The current knowledge of risk factors is, however, mainly based on cross-sectional studies in highly exposed occupational populations. Prospective studies that provide more reliable information on the temporal relations between the exposures and the onset of the disorder are very few to date. Little is also known of the other temporal aspects in the exposure–disorder relationship, for example, whether the health effects of physical work are acute or cumulative, or reversible or permanent. Moreover, in the vast majority of earlier papers, shoulder disorders have merely been defined using self-reported symptoms, despite the fact that the shoulder is one of the few anatomical areas in which consensus-driven criteria for specific diagnoses have been available for several years.11 12

The aim of this prospective study was to assess the long-term incidence of a clinically diagnosed shoulder disorder in a general population and to investigate whether work-related physical exposures predict the risk of a subsequent shoulder disorder diagnosed 20 years later.

METHODS

Baseline data collection in 1977–80, the Mini-Finland Health Survey, and baseline results have been described in detail elsewhere.13 In brief, a two-stage cluster sample was drawn from the population register to represent the adult population aged 30 years or older in Finland. The original sample comprised 8000 Finnish adults, and 7217 subjects (90%) participated in the screening phase of the study that consisted of questionnaires, interviews and standard laboratory and functional tests. The consistency in the participants’ responses to the questionnaires was checked by the study nurse at the beginning of the baseline physical examination. Subjects with shoulder pain during the previous month, or limited performance in the functional screening tests, for example, active shoulder abduction, were clinically examined by a specially trained physician. The overall sensitivity of the screening was 92.6%.13 14

Twenty years later, in 2000–2001, a new extensive national health examination survey corresponding to the Mini-Finland Health Survey was conducted.15 As a part of this Health 2000 Survey, a sample of 1286 subjects from the original cohort (all those who were alive in 2000 and living in or around the five regional capitals of Finland) were invited to be re-examined and 909 (71%) participated in a comprehensive follow-up examination (mean age 64.2 (SD 9.5) years, men 42%). To assess the long-term incidence of a chronic shoulder disorder at follow-up, those with a chronic shoulder disorder at baseline (similar diagnostic criteria used in both occasions) were excluded from the analyses (n = 26). A total of 883 subjects were available for the analyses.

Outcome

At follow-up, the examining physician performed a systematic musculoskeletal evaluation of all 883 participants and identified those with chronic shoulder disorder. The diagnosis was based on a standard clinical examination protocol and symptoms, as well as detailed medical history from interview and medical records provided by the examinee on previous shoulder diagnosis, treatments, sick leave and x-rays. The specific diagnoses were rotator cuff tendinitis, biceps tendinitis, frozen shoulder, inflammatory arthritis, post-traumatic condition, or other non-specified shoulder disorder. In order to establish chronicity of the disorder, a minimum of 3 months of symptoms preceding the time of the examination was required.

Two pilot studies were carried out 7 and 3 months before the follow-up field work started, in order to test and improve the methods. All staff members, including physicians, attended a 3-week training course. Quality control measures included observation, video recording with feedback on examination technique, and repeated and parallel measurements. To study the repeatability of diagnoses, a subsample of 94 subjects underwent the standard clinical examination performed by two field physicians. The agreement between the two examiners was good (κ coefficient 0.71) for the diagnosis of a chronic shoulder disorder.

Potential predictors and confounders

Occupational exposures based on the self-administered baseline questionnaires were separately assessed in the present or last occupation and in the previous occupation of longest duration (if present occupation was not the longest one). Individual physical work-related factors were identified by asking which of the following tasks are (were) typical in their job: lifting or carrying heavy loads; working in forward flexed, twisted or otherwise awkward position; shaking of the whole body or use of vibrating equipment (eg, working in a vibrating vehicle or operating a power saw); a constantly repeated series of movements; or work paced by machine (all dichotomous variables: no/yes). A sum index of physical workload was formed with a possible range of 0 (exposed to none of the factors) to 5 (exposed to all).

The potential confounders of the association between physical exposures and the outcome were considered to be: work-related psychosocial factors (uninteresting or monotonous work, a hurried or tight work schedule, and worry about making mistakes); perceived psychological well-being (from General Health Questionnaire 12); somatisation (13 questions derived from SymptomCheckList-90 excluding three questions related to musculoskeletal or generalised pain); age; gender; body mass index (BMI); smoking; leisure-time physical activity; prior shoulder injury; and diabetes or elevated fasting plasma glucose according to a blood test taken at the baseline examination.

Data analysis

Both gender-combined and gender-specific analyses were performed using logistic regression. The risk of being diagnosed with a chronic shoulder disorder at follow-up is expressed as OR with 95% CI. Considerable positive correlations were found among the individual physical workload factors (eg, κ coefficient for lifting heavy loads and working in awkward position was 0.5); consequently, collinearity prohibited us from inspecting the simultaneous effects of these variables. Selection of confounders (other than age or gender) was performed for each physical exposure separately. If the coefficient of the physical exposure changed by at least 10% due to the variable it was designated a confounder. The main results are therefore derived from multivariable models containing one physical exposure at a time with age, gender and other confounders as appropriate. Multivariable analyses were performed for 852 subjects with no missing values on physical exposures or potential confounders.

As gender was expected to modify the effect of work exposures on the outcome, the analyses were stratified by gender. Other variables (work-related mental stress, individual psychological factors and other individual factors, such as health behaviours) were also tested for possible effect modification. If a statistically significant interaction (Wald χ² statistic, p<0.05) was found, the regression analyses were stratified on that variable.

In addition to excluding those with a diagnosed shoulder disorder at baseline, also subjects with shoulder pain during the preceding month before baseline were excluded from the additional analyses (number of excluded subjects from subcohort 1 was 179). In subcohort 2, the exclusion concerned also those with any self-reported history of shoulder pain at baseline (n = 363). All statistical analyses were performed with the statistical software package SAS (version 9.1; SAS Institute, Inc., Cary, NC, USA).

RESULTS

Participation in the study

Women were more likely than men to meet the criteria for re-examination and be invited (OR 1.31; 95% CI 1.15 to 1.50), whereas older subjects (OR 0.58; 95% CI 0.54 to 0.63; per increment of 1 SD) and those with a higher sum index of physical workload (OR 0.79; 95% CI 0.74 to 0.83; per increment of one item) were less likely to be invited. Among those invited, the corresponding ORs (with 95% CIs) of participation in the re-examination were: sex 0.94 (0.72–1.22), age 0.60 (0.52–0.68) and workload 0.86 (0.78–0.96).

Chronic shoulder disorder at follow-up

In the follow-up health examination, a physician diagnosed a chronic shoulder disorder in 63 subjects (7%). Men were diagnosed more often than women (9% vs 6%). For the majority of the subjects with shoulder disorder, either it was a new diagnosis at the follow-up examination, or medical records indicated that the disorder had occurred within the 5 years preceding the follow-up. All variables associated with the outcome (p<0.2) are presented in table 1.

Physical exposures at work

In the gender-combined multivariable analyses, all physical workload factors except work paced by machine were statistically significant predictors of a subsequent shoulder disorder (table 2). The highest ORs were found for repetitive movements and vibration. Being exposed to several physical exposures increased the risk so that the OR was nearly 4 for the combination of three to five exposures. In general, work involving repetitive movements had the most robust individual association with the outcome; even when all the physical exposures were placed in the same model, work involving repetitive movements held its significance (OR 1.9; 95% CI 1.0 to 3.4; p = 0.05; for men: OR 3.7; 95% CI 1.6 to 8.7; p = 0.004), despite its correlation with other exposures, particularly with working in an awkward position and work paced by machine.

Gender difference

Predictors of shoulder disorder differed for men and women (table 2). In men, work involving vibration and repetitive movements significantly increased the risk of a shoulder disorder at follow-up, whereas in women, an increase in the risk was seen for lifting heavy loads and working in awkward postures (p = 0.07). Although the OR for vibration for women was as high as 3.0, it was not statistically significant because so few were exposed. Women with several physical exposures had considerably higher ORs for developing a chronic shoulder disorder than the similarly exposed men.

Effect modification by age and body mass index

Increased risks of shoulder disorder related to physical exposures were seen in both baseline age groups (30–45 and >45 years) (table 3). Even when the analyses were restricted only to those older than 55 years at baseline, the risk effects for repetition and working in awkward postures remained considerably elevated (OR’s varying from 3 to 5, with wide CIs due to few subjects). Lifting was a strong predictor of a subsequent disorder among those older than 45 years at baseline, whereas among the younger, it showed no effect. Age also similarly modified the effect of the sum index of exposures.

BMI modified the effect of two individual physical exposures: working in awkward postures and work involving repetitive movements, as well as the sum index of the work exposures (table 4). These exposures had a strong increasing effect on the risk of a shoulder disorder, but only among those with BMI lower than 25.

Asymptomatic subcohort analyses

In the subcohort analyses restricted to subjects with no clinical shoulder disorder or shoulder pain at baseline, the proportions of new cases did not differ from those found in the main analyses. Also, the effects of most of the physical exposures (working in awkward postures, work involving vibration or repetitive movements) remained essentially the same, although the CIs widened. Only the OR of lifting heavy loads decreased, from 2.0 to 1.5 (subcohort 1) and to 1.3 (subcohort 2).

DISCUSSION

Our primary goal in this prospective population-based study was to investigate whether exposure to physical factors at work that particularly load the shoulder leads to an increased risk of developing a chronic shoulder disorder. We found that even after a follow-up interval as long as 20 years, occupational physical exposures predicted future shoulder disorders. Of the five exposures measured, four (lifting heavy loads, working in awkward postures, work involving vibration or repetitive movements) increased the risk by 80–150%.

Our results are well in line with the literature; according to four reviews, all of the four exposures and their combinations have been linked to shoulder disorders, the association being the strongest for posture and repetitive movements.7^–9 16^–18 Using an exceptionally long follow-up period in this study as well as extending the study cohort beyond the working population to include older subjects gave us an unique opportunity to investigate whether the effects of work-related physical exposures can still be seen even after the work life has ended. If the deteriorating effects of physical work on the health of the shoulder was only short-term and reversible, we would see a clear decline in the risk estimates among the older subjects, who had left work several years, if not decades, before the diagnosis of the shoulder disease. (The average retirement age in Finland is 59 years.) No such decline occurred, and in fact the ORs for many exposures increased. Even among those older than 75 years at follow-up the risks remained elevated. To our knowledge this is the first prospective study in a general population showing that physical loading at work increases the risk of a subsequent clinically diagnosed shoulder disorder and the damaging effects of physical work seem to be long-term and perhaps irreversible.

In this study, men developed shoulder disorder more often than women. There is some inconsistency in the literature regarding the possible gender difference19 but in general, in epidemiological studies using self-reports, women have consistently reported a higher prevalence of musculoskeletal pain than men at every anatomical site,20 whereas in other population-based studies using clinical examination, no major difference has been detected.14 21 22 Neither have the objective radiographic findings in the shoulder differed between asymptomatic men and women.23 24 The gender-specific risk factors in our 20-year follow-up study are similar to those in a 3-year follow-up study by Leclerc et al25, which showed that repetitive use of a handheld tool (indicating repetitive forceful movements) was the strongest predictor of shoulder pain in men, whereas for women, the predictors were working in awkward postures (arms above the shoulder level) and vibration. Women are traditionally more often exposed to repetitive movements at work, as they were also in our study, but the characteristics of repetitive work often differ between the genders; for example, repetitive work in men may involve more use of force. Despite the higher long-term incidence among men, the ORs were approximately twice as high for women than for men when several physical exposures were present. Our results in concordance with some other studies may indicate that women are more susceptible to the adverse health effects of physical work.26

Our results are also consistent with several cross-sectional, case–control and longitudinal studies in which higher body weight has been associated with higher occurrence of shoulder disorders varying in severity from self-reported pain to surgically repaired rotator cuff tendinitis.4 14 27 28 In our study, overweight men had a particularly high risk of shoulder disorder. Obesity presumably affects the shoulder mainly through impaired metabolism (glucose, lipids, collagen). It is possible that metabolic factors initiate or enhance degenerative process in the shoulder structures, especially the rotator cuff tendons. Once initiated, there may not be regeneration or healing. In our study, however, the effect of BMI was independent of diabetes.

A new finding in this study is the role of BMI as an effect modifier: for the overweight subjects, working in awkward postures or work involving repetitive movements did not increase the already elevated risk of shoulder disorder, as they did for those with normal body weight. One potential explanation is that overweight people may perceive physical exposures differently compared with people with slimmer body structure, for example, by experiencing and reporting more awkward work postures. We detected a significant increase in exposure reporting among those with BMI >25, which was not dependent on age, gender, education or the presence of shoulder symptoms at baseline. Over-reporting of exposures may lead to non-differential misclassification causing the associations between the exposures and the disorder to be deflated. Then again, subjects with higher BMI, even if unexposed to workload, had much higher incidence rates than subjects with lower BMI. This may imply that obesity is such a strong risk factor that the additional contribution from the physical exposures is minimal. Our results indicate that not only physical exposures at work but also obesity are important targets for the prevention of shoulder disorders.

This Mini-Finland cohort is exceptional in many ways. The original baseline population represented the whole population aged 30 years and older in Finland.13 The participation rate in this particular study is high considering the long follow-up. The follow-up health examination was designed to generate specific clinical diagnoses with well-defined criteria. In order to increase the validity and repeatability of the diagnostic testing, only the best-documented tests were used. The physicians were especially trained using video-feedback technique. In general, the benefits of investigating work-related exposures in a cohort of a normal population are that all segments of the population are present; also, in addition to achieving information on the magnitude of the particular health problem at the population level, important information on the distribution of the potential risk factors and their risk effects can be collected across a wide variety of people. Hence, the results are better generalisable to other populations.

Our study has limitations. The statistical power was rather weak because there were only 63 incident cases. The assessment of physical exposures, which formed only a small part of the extensive health survey, was based on self-reports and the questions did not include more detailed information on the potentially harmful features of physical work that were not well known 20 years ago. For instance, whole-body and hand–arm vibration were not separated. Hence, our results may be affected by misclassification of potential non-differential exposure, meaning that the risk estimates would have presumably been even higher if current knowledge of the harmful qualitative and quantitative exposure limits had been available during the assessment of exposure. Also, symptoms may cause some subjects to overestimate their exposures.29 We took this into account by performing additional asymptomatic subcohort analyses. All of the risk estimates, except for heavy lifting (which decreased), were similar in this subset. The term “heavy” may have involved more subjectivity than other items and hence its estimation may have been affected by present symptoms. Confounding due to unmeasured exposures, such as more specific work organisational or psychosocial features, may also have affected the results.

We did not have any information on the subjects’ work history after the baseline but we assumed that for the majority of people, exposure to specific features of physical work did not markedly change during work life. This assumption was based on our additional analyses using the population-based Health 2000 Survey data, which showed that for the great majority (77–88%) of workers who were 50 years or older, being exposed or not to shoulder-specific physical work factors (repetitive movements, working in awkward postures, or with a vibrating tool) did not change during their entire work life (unpublished data). Accordingly, it is likely that in the present study, most of the subjects who were 30–40 years at baseline and exposed to physical work were also exposed till the end of the follow-up time. If selection out of exposed jobs during the follow-up due to shoulder disorders occurred it would have likely deflated the associations towards the null.

The two-staged selection from the baseline examination to the invitation and further from that to the participation in re-examination may also have attenuated the risk estimates, as physical workload predicted participation at both stages. At the first stage, all factors associated with mortality, such as male sex, older age and lower socio-economic status, and all factors that are more common in rural than urban settings, led to a decreased likelihood of re-examination. At the second stage, it was not possible to determine why workload 20 years earlier was associated with non-participation. As a result, the prevalence of and power for the effects of strenuous work may have been negatively affected. Any severe bias in the other direction could only result if individuals in heavy work with shoulder symptoms were more likely to participate than similarly exposed people without symptoms. While this cannot be ruled out, continuous follow-up of both mortality and the incidence of chronic shoulder disorder would be needed to control all the effects of health-related selection in this study.

Our results give further evidence that work-related physical exposures are risk factors for shoulder disorders, and the effects also seem to be long-term so that the accumulation of damage in shoulder tissues potentially induced or aggravated by these factors can be seen several years after work life has ended. Early preventive measures at the workplace may have long-lasting health benefits for the shoulder, although scientific evidence of the effectiveness of preventive interventions is still lacking perhaps partly because of too short follow-up periods. Combining exposure interventions with comprehensive workplace health promotion programmes with the particular focus on body weight management may provide an additional contribution to the prevention of shoulder disorders.