One recent article, which has received a tremendous amount of public attention in recent weeks, purports to show no association between carpal tunnel syndrome and computer keyboard work .  Because of its widespread coverage in the lay news media, some detailed discussion of this paper is warranted.

One methodological concern involves subject selection and exclusions from the study population.  The authors stated that they surveyed all Mayo Clinic "employees who use a computer" but did not explain how these persons were identified nor how they could be sure that they correctly included all appropriate employees and no others.  An unknown number of subjects did not undergo NCS because "it was obvious that some other condition was responsible for the paresthesia" (other conditions not specified in the article, although in a recent issue of CTD News Dr. Stevens mentioned other potentially work-related disorders such as ulnar nerve entrapment).  Three subjects with possible CTS refused to undergo electrodiagnostic testing, and 8 had been previously diagnosed "but were not currently symptomatic" (whether because of treatment or episodic symptoms is not known).  If the number of cases was 35 (27 + 8), the prevalence would be 13.6%, and if it was 38 (27 + 8 + 3), then the true prevalence was 14.8%.  (None of these estimates includes the participants who were not offered NCS.)  It should also be noted that subjects who denied hand paresthesias on questionnaire "were assumed not to have CTS," even though the authors themselves cited a frequency of 18.4% of median neuropathy in asymptomatic persons.

Thus the CTS prevalence estimate from this study probably lies in the range from 10.5% to 14.8%.  To what reference value should these values be compared?  No unexposed group was surveyed, a serious weakness.  The authors cited prevalence ranges from several other articles, with little discussion of the populations studied, the varying study methods used, or why such a wide range of values was obtained.  Of particular concern here are the noncomparability of case definitions among these studies, including the present report, and lack of attention to the age‑gender distribution of each population, although Dr. Stevens himself earlier showed age and gender to be associated with CTS .

Of the studies listed, the only one with even generally comparable data collection methods (mailed survey, NCS offered to symptomatic cases) was that by Atroshi and colleagues.  This paper gave a prevalence of 3.8% for clinical CTS, which Stevens et al. astonishingly described as "similar," even though it was less then one-third as high as their estimate.  

Another serious weakness is that the authors conducted no formal assessment of exposure to computer use.  Job title, amount of keyboard and mouse use, and years of keyboard work were apparently obtained from questionnaire responses, although this was not stated.  The type and amount of use likely varied widely among the 314 persons to whom the questionnaire was sent.  Of note, CTS cases were 1.7 times more likely to use a computer mouse "frequently" than the non-cases (48.1% versus 27.9%).  In fact, mouse use has been suspected of increasing the risk of upper extremity disorders.

A substantial fraction of respondents gave a job title that was classified as "other," obscuring the nature of the work demands in these jobs.  Specific features of work such as type of task performed (data entry, interactive use, etc.) and work postures have been shown to affect the risk of upper extremity disorders.  Failure to assess these characteristics obscures potential exposure-response relationships and could have led to unmeasured confounding.  Examination of Table 2 shows that the reported prevalence of clinical CTS was 11.8% (15/127) among "secretaries," 13.8% (9/65) among those in "other" occupations, and 17.3% (13/75) among frequent mouse users—all higher frequencies than the overall prevalence.

Another issue is that the study included only current workers; any person out of work because of CTS would not have been surveyed.  This form of selection bias, the "healthy worker effect" (HWE) cited above, is particularly problematic in relation to a highly disabling disorder such as CTS, in which the symptoms are likely to interfere with and be aggravated by repetitive manual work.  The studies offered for comparison with the Stevens' data all sampled the general population and thus were unlikely to have been affected by HWE in the same way.  Another working population would have been a far more appropriate basis for comparison with these data.

Finally, it is of note that the authors do, in fact, describe a substantial proportion of computer workers with hand paresthesias.  These paresthesias are not in themselves inconsequential in their effect on people's lives.  Overall, this study is a description of the prevalence of CTS (one of many possible upper extremity health outcomes) among a group of computer users.  However, because of the lack of a control (unexposed) group, potential selection bias, inadequate exposure assessment, and likely confounded comparisons to other study populations, it is severely methodologically limited in its ability to assess the presence or absence of causal relationships.