Carpal tunnel syndrome and the use of computer mouse and keyboard: A systematic review

In the literature search on the association between computer work and CTS 4661 references were identified (Figure 1). Eight epidemiological studies met the criteria for inclusion [6‐14] (two of the papers were from the same population [11, 12]). Four of these studies were prospective in design [6‐8, 12], one was a case-referent study [10], one was cross-sectional but with a case-referent approach [9], and two were cross-sectional[13, 14]. The studies are listed in Table 2 with information on design, population, response rate, control group, exposure, CTS case definition, confounders controlled for, results and strengths and weaknesses [see Additional file 1].

The large population-based study by Atroshi et al. included both physical examination and NCT [14]. It showed a significant protective effect of keyboard work, i.e. the prevalence of CTS increased with decreasing hours/day with self-reported computer work. The study was carefully conducted with adequate blinding of the interviewer regarding exposure and of the technician regarding the symptom status.

An Indian cross-sectional study, on the contrary, showed a significant effect of both years and hours/day with computer work [13]. The participation rate was 100% which raises a question of the selection process. There was not the expected effect of gender (men had twice the frequency of CTS compared to women), age or BMI. Blinding was not described and NCT was not used.

In the follow-up study of Andersen et al. (The NUDATA study), two case definitions relevant for this review were used [see additional file 1] [8]. When applying case definition 1, analysis of baseline data showed odds ratios of 2.3 (95% CI 1.2–4.5) among participants reporting 5–9 h/w of mouse use increasing to 3.6 (95% CI 1.8–7.1) among participants reporting 20–24 h/w of mouse use. There was no further increase of risk among participants reporting more than 24 h/w of mouse use. Case definition 2 showed only borderline significance at 30+ hours and a very irregular pattern of association. In the follow-up analyses, a significantly increased risk was found among participants reporting working more than 20 h/w with the mouse.

In the extensive prospective study of computer users by Gerr et al. the prevalence of CTS was too low to allow any analyses of association [6]. It is worth noting, however, that with the use of an outcome definition including both symptoms and NCT the prevalence and incidence of the disease was low in this working population.

As a substudy of the Danish PRIM study (Project on Intervention and Research in Monotonous work), 731 participants from three companies, one bank and two postal centres, were included. [7]. Data entry was the main repetitive task. Electrogoniometer measurements of wrist movements showed that this work task was highly repetitive. Blinding of physician and technician was observed. The overall prevalence of CTS was 1.1% (8 cases) on the working hand and 0.3% (2 cases) on the contralateral hand. The risk of CTS was statistically significantly increased for every 10 hours of repetitive work (OR = 1.86 (95% CI 1.06–3.19)) after adjusting for forceful work and personal characteristics. The annual incidence of CTS was 0.62% (4 cases) and thus no further analyses of incidence data could be made.

Nathan et al. had followed a cohort established in 1984 for 11 years [11, 12]. Originally, the cohort included 471 participants from 4 industries representing a wide variety of hand activities. It was not stated if the at-risk population had normal NCT at baseline nor if the examiners were blinded to exposure or health status. In multivariate analyses adjusting for various potential baseline confounders (but not age and gender), neither an effect of keyboard work nor of repetitive work (as ascertained in 1984) was found.

In the study by Stevens et al., all participants were identified as "frequent computer users" working in a medical facility [9]. Exposure and demographic characteristics were compared between the two groups but without formal statistical testing. Hours of daily keyboard use and years with keyboard did not differ between the groups. Frequent mouse use was more prevalent in the CTS group than the non-CTS group (48.1% vs. 27.9%). This difference was not tested statistically. However, testing the reported figures with a Mantel-Haenszel chi-square test (1 df) showed a statistically significant difference (p = 0.04).

In another study with a case-referent approach de Krom et al. identified 156 CTS cases from a population based survey and from an outpatient department of neurology [10]. Referents were persons without CTS symptoms from the population based survey (n = 473). Care was taken to keep the purpose of the study blinded to the participants in order to avoid information bias. One question documented hours per week of typing during the last 5 years. The odds ratios were all below 1 and not statistically significant.

A large population based study with main focus on hand-arm vibration and CTS did not meet the inclusion criteria because of insufficient outcome definition [15]. There was one questionnaire item concerning keyboard use more than 4 hours per day. No excess risk for self reported tingling/numbness was found (PR = 1.1, 95% CI 0.8–1.3) after adjusting for age, smoking, headaches and tiredness/stress.

In summary, of the eight studies identified, four studies were designed as follow-up studies. One of these studies was well performed with a short follow-up period, a large study population and thus sufficient statistical power. The study found some positive associations. However, the possibility of information bias in combination with an outcome definition not involving NCT made inferences difficult [8]. Two of the follow-up studies had too few CTS cases to perform analyses [6, 7]. The last follow-up study identified had a very long follow-up period, only baseline exposure information and no adjustment for age and gender. They found no association but because of the important limitations the result was difficult to interpret [12]. One of the follow-up studies reported positive associations from baseline data but here the computer exposure was mixed with other kinds of repetitive work [7]. One of the two cross-sectional studies was very well performed and had enough statistical power and found the opposite association of what was expected [14]. The other cross-sectional study showed a positive association but had some confusing results, e.g. a much higher prevalence among men compared to women. Two of the eight studies were case-control studies. One study had limited statistical power and showed no association [10]. The other showed a weak association but only in a crude analysis with no adjustment for age [9].

In the literature search on repetitive work and CTS, 229 studies were retrieved and three studies fulfilled the inclusion criteria [16‐18]. Two of the three studies were based on one population [16, 17]. Werner et al. identified 49 asymptomatic participants with an abnormal NCT and 59 with a normal NCT and assessed them after 17 months and after 70 months. Work tasks were video-filmed and categorized according to repetition rate on a scale from 1–10. The incidence of developing CTS symptoms was equal in the two groups after 17 months but significantly higher after 70 months in the group with initially positive NCTs. Repetitive work was a risk factor for developing symptoms after 17 months (OR 1.35, 95% CI 1.03–1.77) and after 70 months (risk estimate not provided).

In a well designed follow-up study by Gell et al. no association between level of repetitive hand tasks and development of CTS was found [18]. The case definition included both symptoms as well as NCT. Each job was assessed and rated for ergonomic exposures.

In all, results from studies on other kinds of repetitive, low force work and CTS did not add to evidence of to an association between computer use and CTS.

Seven studies comparing median nerve function in computer users with groups without computer use were found [19‐25]. Two of the studies assessed median nerve function with NCT [23, 25] whereas the other five studies used vibration perception threshold testing. Vibration sense perception, however, is not a good indicator of CTS [26]. The selection of participants was not described in the five studies using vibration perception threshold testing and selection bias may have affected the results.

The study of Murata et al[23] used nerve conduction tests among 27 female life insurance company employees entering data for six hours or more per day and 24 female students. Significant differences in median nerve sensory conduction velocities were found for measurements across the carpal tunnel whereas values proximal and distal to the wrist did not differ. The two groups differed in symptom profile. The findings of Murata et al[23] were in contrast to a recent study by Sandén et al[25] in which 82 secretaries with a median of 6 hours of daily computer work were compared to 35 nurses with very limited computer work. No statistically significant differences were found in the median nerve conduction velocity or in the vibration threshold between the two groups in t-test analyses. Doezie et al. compared vibration thresholds among transcriptionists with symptoms to a control group [21]. Thresholds of the second and fifth fingers were significantly elevated in the transcriptionists compared to the control group but only for the high frequencies (125–500 Hz). Very little information about the control group was shown.

Greening et al. conducted two studies examining associations between vibrotactile thresholds and computer use [19, 20]. In both studies, they found that patients with musculoskeletal symptoms in the upper limbs had higher thresholds than healthy individuals. The studies did not compare computer users without symptoms to non-computer users. Thus, the design of the studies did not allow conclusions considering an effect of computer work per se.

Similar methods were used in a Danish study and similar results were observed [24]. Finally, another Danish study studied vibration thresholds in computer users but with a focus on symptoms and not levels of computer use [22].

It is widely believed that biomechanical factors (e.g. forceful exertions, repetition, and awkward postures) increase the risk of CTS by increasing carpal canal pressure with subsequent nerve ischemia [27]. Therefore, in addition to epidemiological evidence of associations between computer work and CTS, insight into the role of computers on the development of CTS may be found in studies examining wrist biomechanics or carpal canal pressures during computer use.

Wrist positions and forces exerted by computer users have been measured in several studies. Keir et al[28], reported that wrist extension ranged from 23° to 30° and that ulnar deviation from -3.2° to 5.2° during mouse work. In a study of wrist position in keyboard work (entering of data), electrogoniometer measurements showed wrist extension of 14° and 20° at the 50^th and 90^th percentile, respectively [7]. In another study of keyboard work, mean ulnar deviation using a conventional keyboard was 18.9° (SD 6.8°) [29]. Gerr et al. reported wrist postures observed among 379 computer users. Mean wrist extension was 24.3° (SD 9.6°) during keyboard use and 23° (SD 8.8) during mouse use. Mean ulnar deviation was 5.0° (SD 7.3) during keyboard use and 1.0° (SD 7.7°) during mouse use [30].

Several investigators have measured finger tip forces among computer users. The finger tip force exerted while keying varied from less than 1 N to 7 N but in most studies was between 1 and 4 N [31‐34].

In the literature search, 253 studies were retrieved and nine studies were found with measures of carpal tunnel pressure in relation to finger, wrist or arm use. Several studies have measured the carpal tunnel pressure (CTP) among persons free of CTS and among those with CTS [28, 35‐42]. In aggregate, these studies suggest that CTS development is associated with elevated CTP. The resting CTP with the wrist in neutral position among persons free of CTS ranges from 3 to 13 mmHg (results from 7 studies are summarized in [37] and [39]). CTP in CTS patients varies between 10 and 43 mmHg [39] though higher values have been found [38]. In an often cited study by Lundborg et al[35], CTP was increased experimentally among 16 human volunteers. In four participants the CTP was increased to 60 mmHg and in four other participants the CTP was increased to 90 mmHg. In these two groups, the sensory and subsequently the motor response were blocked within an hour. In a third group of 4 participants CTP was increased to 30 mmHg. This produced minor and varying effects but "pins and needles" was reported in 2 of 4 subjects. This was further studied by Gelberman et al. who found some functional loss at 40 mmHg and complete motor and sensory block at 50 mmHg among healthy subjects [36].

Several studies have measured the CTP profile associated with different wrist angles, finger flexion and forearm position [39, 40, 42, 43]. The studies show that CTP is dependent on the position of the forearm, wrist and metacarpophalangeal joint (MCP). In particular, supination showed higher CTPs than pronation and MCP flexion increased CTP [28, 39, 42]. With wrist positions between 40° flexion and 40° extension the CTP did not exceed 20 mmHg regardless of MCP angle. Ulnar and radial deviation had only small effects [40].

CTP has been studied among persons engaged in actual work tasks. Rempel et al. measured CTP among 19 healthy subjects engaged in a number of hand intensive tasks [37]. CTP increased from 8 (SD 6) mmHg at rest to 18 (SD 13) mmHg after lifting 0.5 kilogram cans for 5 minutes at a rate of 20 cans per minute. Keir et al. conducted a study on the effect of computer tasks on CTP. Among 14 healthy subjects the mean CTP rose from 5.3 mmHg during rest to 16.8–18.7 mmHg (varying between different kinds of computer mice) with the hand static on the computer mouse and to 28.8–33.1 mmHg while dragging or pointing and clicking with the mouse [28]. This was the only study of computer work and CTP that was found.

To summarize, measurements of CTP under conditions commonly observed among computer users showed modest increases generally believed to be below potential harmful levels. However, one study showed an increase of CTP during actual mouse use to levels where possible neurological effects were seen experimentally. These studies have not been repeated in other studies and nothing is known about the effects of prolonged or repeatedly increased pressures to this level.