Impact of occupations and job tasks on the prevalence of carpal tunnel syndrome.

HAGBERGM, MORGENSTERNH, KELSHM. Impact of occupationsand job taskson the prevalence of carpal tunnel syndrome. Scand J Work Environ Health 1992;18:337-45. In thisinvestigationreported epidemiologic studies on carpal tunnel syndrome (CTS)(15 cross-sectional studies involving32 occupa tionalor exposure groupsand sixcase-referent studies) werereviewed. Theprevalence of crs in thedifferent occupationalgroupsvariedbetween0.6and 61 0/0. Thehighest prevalence was notedfor grinders, butchers, grocerystore workers, frozen food factoryworkers, platers, and workerswith high-force, high-repetitive manual movements. Odds ratios greater than 10were reported for exposed groups in three studies. On the basisof epidemiologic and other evidence, it wasconcluded that exposure to physicalwork load fac tors, such as repetitive and forceful gripping, is probably a major risk factor for CTS in several types of worker populations. At least 50%, and as much as90%, of allof the CTS casesin theseexposedpopu lations appeared to be attributable to physical work load.

The association between occupational activities and carpal tunnel syndrome (CTS) has been addressed by several investigators. Despite the reported association between occupation or occupational exposure and CTS , a causal relationship between usage and CTS has been refuted. Recently an editorial in the Journal of Occupational Medicine claimed that "CTS is not a cumulative trauma disorder [p 39]." In the Newsletter, Occupational Problems In Medical Practice the same author claimed that "I am convinced that some forms of usage can exacerbate the symptoms of some regional musculoskeletal illness, but not all-for example, no t CTS [p 7]" (2). On the other hand Gerr et al (3), in a review of upper-extremity musculoskeletal disorders of occupational origin, concluded that " carpal tunnel syndrome is etiologic related to occupational exposures [p 562]." Evidence for a causal relationship between workplace ergonomic factors and nerve entrapment of the median nerve at the wrist was recently reported in a meta-analysis by Stock (4). To infer a causal relationship between occupational exposure and CTS, epidemiologic studies would have to show a statistical association between occupation and CTS that is not due entirely to estimation errors (chance or bias).
The aim of the present investigation was to review Reprint requests to: Professor M Hagberg, National Institute of Occupational Health, IFM, S-171 84 Solna, Sweden. the epidemiologic literature for the possible effect of occupation, specifically physical work load, on the occurrence of CTS and to consider the overall evidence for a causal relationship.

Materials and methods
A survey of the Medlars documentation system was performed for the years 1966 to 1990 (November); 164 references were obtained with the use of the key words carpal tunnel syndrome, CTS, occupation, incidence, and prevalence. Most of the articles dealt with case descriptions, diagnosis, or treatment (60% of the references). Reviews or educational articles came second (20010 of the references). Only 15% of the articles were population-based studies of CTS in occupational groups. Furthermore, we studied the reference lists of these articles to assess all the literature of interest for epidemiologic aspects of CTS. Only articles or official reports in which CTS was defined by both symptoms and signs were considered. These signs included an electrodiagnostic test of median nerve block, a pos itive Phalen's test, or a positive Tinel test. Surgical release of the median nerve at the wrist was also accepted as CTS, since both symptoms and signs are a general requirement for carpal tunnel surgery (5). Furthermore we included a study of occupational groups in which only nerve conduction was measured (6).
Prevalence odds ratios with 95% confidence intervals were estimated for the different exposures or occupational groups in the studies in which the authors did not perform these computations (7). The prevalence odds ratio approximates the incidence rate ratio if the mean duration of the disease is the same for the expo sed and the unexposed and no other sources of bias are present (8,9).

337
Scand J Work Environ Health 1992, vol 18, no 6 The impact of the exposure (occupational group or job title) was estimated by the attributable fraction in the exposed population -ie, the proportion of exposed cases that would not have developed the disease in the absence of exposure. In the reviewed studies the reference group had a different occupation or job task than the study group. The reference groups could not be regarded as unexposed but rather as less exposed than the study groups. Therefore the preferred terminology is reference group and not control group since exposure was not under control. The attributable fraction was estimated by (OR-l)/OR, where OR was the estimated odds ratio (8). The confidence limits for the attributable fraction were calculated from the upper and lower limits of the odds ratio.
In one case-referent study, the authors presented differences in mean exposures between the cases and referents. These differences were converted to odds ratios by the method of Greenland (10).
The prevalence of crs in the different occupational groups varied between 0.6 and 61% (table 1). The lowest prevalences were noted for industrial workers with low-force, low-repetitive jobs (0.6%) and slaughterhouse workers (1%) (table 1). The highest prevalences in the cross-sectional studies were noted for grinders, butchers, grocery workers and frozen food factory workers with high-force, high-repetitive gripping jobs, and platers. The estimated effects on CTS in the cross-sectional studies were high for industrial workers with high force and high repetition in their jobs with an odds ratio of 16 and for platers with an odds ratio of 11.0 (table 2). In the case-referent studies (table 3) the highest odds ratios were observed for exposures defined as high-level or long-lasting vibration exposure, three different studies showed odds ratios greater than 4 (table 4). An exposure-response relationship for vibration exposure was indicated in both cross-sectional and case-referent studies (11,25). Other occupational factors related to CTS were wrist flexion and extension postures and keying work (table 4).
The impact of occupation or occupationally related exposure on CTS was substantial (table 5). High-force, high-repetition vibration exposure or manual work or both indicated attributable fractions of 80% or higher in the exposed populations (table 5). Keying had an attributable fraction of 74% among women working at least 20 h a week with a keyboard (table 5).

Discussion
There are numerous sources of estimation error (bias) in epidemiologic studies. In the following discussion     we evaluate the relevant CTS literature in terms of the following four major threats to (internal) validity: (i) selection bias and temporal ambiguity, (ii) problems of case definition and identification, (iii) problems of exposure definition and measurement, (iv) and con-founding. In addition, we consider several other criteria that reflect our ability to make scientific generalizations about causal associations (8,(31)(32)(33)(34). Although we recognize the limitations in defining any set of causal criteria (31), we discuss the five criteria that  (12) we believe are the most relevant to generalizing about the possible effect of physical work exposure on the occurrence of CTS: strength of association, consistency of results , coherence of evidence, experimental and laboratory support, and literature availabilit y.

Selection bias and temporal ambiguity
A potential limitation of prevalence studies is thatthe time of disease onset among study cases is usually not known. Consequently, we cannot be sure that the exposure preceded disease occurrence, especially when the exposure is measured at the same time the disease status is observed. Even retrospective measurement of the exposure may not rectify this problem of •'temporal ambiguity," since we seldom know the exact time of disease or symptom onset. The problem of temporal ambiguity is the most likely to threaten the validity of a prevalence stud y when the disease or related symptoms can affect exposure status or when the disease can affect subject selection differentially through exposure status. In occupational studies, the latter phenomenon is a secondary form of the well-known "healthy-worker effect "; that is, exposed workers who develop symptoms of CTS or other conditions (in part because they are exposed) leave their jobs and are not selected for future studies of working populations. This type of selection problem, there-342 fore , will probably lead to a negative bias in the estimation of the effect of physical work exposures on CTS. Thus prevalence findings are likely to underestimate the true exposure effect.

Problems of case definition and identification
In all of the studies selected for this review, the diagnosti c criteria for CTS included both symptoms and signs. In most of the studies it was not clear what quality of symptoms or combination of symptoms was required to meet the case criteria. Median nerve distribution of symptoms was a case criterion used by most of the studies. The criteria for intensity and duration of hand symptoms to meet the case criteria may have differed among the reviewed stud ies. In studies in which the sign was a positive Phalen 's or Tinel test, however, there may have been undetected cases in comparison with studies in which nerve conduction measurements were performed, and the sign was defined as median nerve conduction block. It has been reported that the sensitivity of the Phalen's or Tinel test is only 60-67 070 in comparison with electro physiological nerve conduction block as the "gold standard" (35). De Krom et al (36) claimed that provocative tests do not distinguish between CTS and other causes of nocturnal hand complaints. This possibility could explain some of the variability between similar occupational groups. We would urge scientists to use electrodiagnostie tests for CTS for case definition in future studies of CTS.

Problems of exposure definition and measurement
In the cross-sectional studies, exposure was defined by job title according to the employer's records . In the case-referent studies, the exposure was usually obtained by questionnaire or interview. There was no study in which direct exposure measurements were performed in the workplace for all of the study persons. Job titles and questionnaire data are crude indicators of exposure (37,38). The variability of exposure definition (eg, job title criteria, posture and movement criteria) is probably great across different studies.

Confounding and individual susceptibility
Individual risk factors (individual indicators of susceptibility) can act as confounders of exposure effects. The most important confounders are age, gender, anthropometry, and other diseases.
The stress capacity of different tissues decreases with age, causing a shift of the stress-strain curve. The normal reparative and wound healing process is also slower with age. The duration of exposure is related to age. There is yet no study that has shown age as a risk factor for CTS when the duration of exposure is controlled.
The incidence rate of CTS is related to gender. A male-to-female ratio of 1:3 was described in a population study (39). There is yet no evidence for an increased female susceptibility for work-related CTS when exposure is controlled for. In a population-based incidence study of occupational CTS, the male-tofemale ratio was 1.2:1 (40). In the study, by Silverstein et al (12), of CTS among industrial workers, no difference was observed between genders when exposure factors were controlled.
Carpal canal size is a controversial risk factor for CTS. There are different studies linking CTS with both small and large areas (15,41). Other disorders associated with CTS are diabetes mellitus, myxedema, acromegaly, amyloidosis, fracture of the forearm, and the like. There is one report of a CTS prevalence of 11% for diabetes mellitus patients (42). For pr egnant women an incidence of CTS of up to 25% has been reported (43,44).
Most of these studies have taken into account possible confounding by age and gender. Would the other individual risk factors influence the inference of occupation and job tasks as risk factors for CTS? For example, Juntunen et al (45) questioned the association between vibration exposure and CTS, claiming that patients with neuropathic diatheses, who are at greater risk for CTS, tend to be selected into groups of patients with vibration syndrome. This phenomenon could be due to a detection problem (ie, workers with vibration exposure are likely to get more medi-Scand J Work Env iron Health 1992, vol 18, no 6 cal attention because of white fingers than unexposed workers do) . If so, the frequency of CTS would be higher among exposed workers because of earlier diagnosis or detection in this group. However, this detection bias was not present in the cross-sectional studies in which the same diagnostic procedures were used for both the study and reference groups . Nevertheless, it is not likely that workers with neuropathic diatheses would be exposed to physical work load such as repetitive forceful gripping and vibration to a greater extent than would workers without this susceptibility. It is more likely, we believe, that the most susceptible workers would seek jobs with less physical workload exposure. Although exposure to physical workload factors can result in CTS as an early manifestation of a neurologic disorder in some workers (46), the frequency of such disorders is probably very low in worker populations (17,47,48).

Strength of association
The strength of association in the reviewed studies was generally high. There were three different studies reporting odds ratios greater than 10 for (i) high repetition and high force in the hands, (ii) vibration exposure, and (iii) occupation as a plater. Such large effects cannot be easily explained by any sources of bias (eg, a confounder), especially biases that were not evident to the investigators. Nevertheless it should be noted that even strong observed associations are not incompatible with spurious results.

Consistency of results
There was a surprising consistency of observed effects across the different cross-sectional studies, as well as between the case-referent studies and the cross-sectional studies. Occupational tasks or job titles associated with vibration exposure, repetitive hand movements , and forceful grip s were reported in both the cross-sectional and case-referent studies as risk factors for CTS. Vibration exposure is probably an indicator of exposure to forceful repetitive gripping. In the study by Silverstein et al (12), the crude odds ratio for exposure to vibration was 5.3, but the odds ratio adjusted for high force and high repetition was only 1.9 (table  3). One inconsistency was noted, however. Butchers were reported to have a CTS prevalence of 53% in the study by Falck & Aarnio (14), but the 113 slaughterhouse workers (including 38 butchers) studied by Viikari-Juntura (18) had a reported prevalence of only 1% (one case with the job title cutter). One explanation for this inconsistency is that CTS was defined by symptoms and nerve conduction measurements, which may be a more sensitive indicator of CTS, in the Falck & Aarnio study. If one considers only symptoms and a positive Phalen's test in the Falck & Aarnio study, the prevalence would drop from 53 to 12% (exact 95% confidence interval 1-36) (14). Furthermore, the secondary healthy worker selection may have been stronger in the Helsinki area, it being a metropolitan area with more job opportunities than in the smaller town of Pori where Falck & Aarnio performed their study. Thus the inconsistency could have been due to different diagnostic criteria, sociodemographic factors, or chance.
The issue of temporality, which means that the exposure precedes the onset of the disorder, was not demonstrated in all of the studies. In the investigation by Nilsson et al (11)the observed latency (duration between onset of exposure and symptoms) for CTS was reported. In the study by Silverstein et al (12), only disorders that developed after the onset of exposure were considered. In most of the other studies it was not clear whether all registered exposure preceded the onset of symptoms. In future studies of CTS we would urge investigators to explore the temporality of CTS and to establish exposure-latency time relationships.

Coherence oj evidence
An exposure-response relationship was reported in severalstudies. For example, Silversteinet al (12)found that both repetition and force were predictors of CTS and, if both exposure factors were present, the risk of CTS was even greater . Repetition and force can be regarded as two components of biomechanical stress exposure. In the study comparing platers with office workers , the risk of CTS increased with the number of years exposed to vibration (11). In the case-referent study of Wieslander et al (25), there were progressive trends in the odds ratios with the number of years of use of handheld vibrating tools and repetitive movements of the wrist. In the studies of Silverstein et al (12), Cannon et al (26), and Chiang et al (24), however, the number of years on the job was not a predictor of CTS. One explanation for this inconsistencymay be that only certain individuals are at risk of developing crs and the latency for these individuals is relatively short when they are exposed to hazardou s physical work load. Thus cumulative exposure may not be an important predictor of CTS.

Experimental and laboratory support
There are experimental studies on humans that support the causal relationship between ergonomic factors and CTS. Exposures are likely to affect the median nerve both directly through mechanical stress (eg, stretching and compression) and indirectly through ischemia causing paresthesia or a nerve conduction block . A localized pressure greater than 50 mm Hg ( =:: 6666 Pa) in the carpal tunnel or a tourniquet with a pressure higher than systolic around the upper arm will cause a conduction block in the median nerve at the wrist (44). Extreme flexingor extension of the wrist causes an increase in pressure in the carpal tunnel that can affect the blood perfusion of the median nerve (49,50). The histopathology of CTS shows coherence with the epidemiologic evidence for physical work load as 344 a risk factor . Microscopic studies of tissues in the carpal tunnel in wrist specimens reveal changes (eg, an increasein epineurium density) suggesting that repeated exertions with a flexed or extended wrist are involved in the etiology of CTS (51), These changes relate to mechanical pressure and perfusion of the median nerve.

Literature availability
Most of the studies considered in this review were cross-sectional investigations . All of the case-referent studies dealt with prevalent cases. There were no prospective cohort investigations. Furthermore, a review of the association betweenoccupation and crs is likely to be distorted by the fact that there is a tendency to publish positivefindings rather than negativeor equivocal ones. However, since data concerning the epidemiology of work-related musculoskeletal disorders is sparse, reports of negative or equivocal findings of CTS epidemiology are probably welcome in the journals. An example of an equivocal report is the study of slaughterhouse workers by Viikari-Juntura (18), who found only one case of CTS among 117workers.

Concluding remarks
On the basis of epidemiologic and other evidence, we conclude that exposure to physical workload factors such as repetitive and forceful gripping is probably a major risk factor for CTS in several types of worker populations. At least 500/0, and as high as 90%, of all CTS cases in these exposed populations appear to be attributable to physical work load.