Evidence for Work-Related Musculoskeletal Disorders

To the Editor: The recent editorial submitted by Dr Hadler raises some important issues related to epistemology, bias, and specific analysis of repetitive strain injury (RSI) literature.¹

While dismissing the entity of RSI, Dr Hadler fails to reference many important findings that contradicted his view, including a study of degenerative joint disease and occupational hand and arm use. This study found important and statistically significant differences in degenerative joint disease in textile workers when the right hand (use hand) was compared with the left.² The authors stated, "These differences argue cogently for the rejection of the null hypothesis that pattern of usage does not influence structure and function of the hands." The authors went on to note that "the tasks involved usages that are extraordinary only because of their repetitive stereotyped nature."

In his editorial, Dr Hadler stated that "individuals who perform repetitive motion of the hands for decades suffer less degenerative disease of the hands, than those who do not." However, his own study directly contradicted this assertion, providing evidence that degenerative joint disease is associated with repetitive hand motions (also observed in keyboard use).

Hadler's causation analysis lacks both scientific rigor and any obvious connection to 50 years of epistemological writing.^3-6 Hadler claims to prove that keyboards do not cause injury. This cannot be done. Studies can fail to support the null hypothesis that keyboards do not cause injury. As Popper illustrated with swans, however, one need find only one black swan to disprove the hypothesis that all swans are white. No matter how many white swans are found, it does not follow that all swans are white. Hadler may critique studies like his own that find an association between injury and repetitive use but this does not prove an association does not exist.⁷

Computer industry scientists have never considered the presence or absence of epidemiological data as the essential criterion for determining the etiology of RSI. Since 1984, medical researchers have generated substantial evidence of the magnitude of the problem, the contribution of work factors as causes, and possible solutions. This evidence has been noted by manufacturers of keyboards. IBM recognized, before the publication of prospective epidemiological evidence in peer-reviewed journals, that tenosynovitis was increasing within the company and was "associated with sustained, high-speed, repetitive actions such as keyboarding."⁸ During the past few years, several keyboard manufacturers have taken this evidence to heart and added warnings to their keyboards.

The Computer and Business Equipment Manufacturers Association, an industry trade group, is so concerned about its own written track record of opinions, studies, and strategies for dealing with keyboard-induced RSIs that they have recently sought and received judicial authorization to suppress the release of their own documents on the topic.⁹ Why are they so concerned about the publication of their own pre-litigation opinions on the matter?

A new strategy for dealing with this injury has been developed of late. Deviating from the usual practices of medical decision-making, Dr Hadler asserts that the relationship between regional arm pain and repetitive stress cannot be evaluated until a controlled epidemiologic study is conducted and then states that there is "no reasonable way to design such a study" (despite his own past experience designing one). This is the type of novel scientific argument Daubert is concerned with.

Hadler demonstrates a lack of understanding of the underlying scientific criteria that have generally been applied to establish cause-effect relationships. Both Hadler and his responders comment almost exclusively on the epidemiologic literature. Epidemiology is not the sole, or in some cases the best, way of establishing cause-effect relationships. Epidemiology is one among several types of information. For some types of cause-effect relationships, primarily injuries, epidemiology is primarily valuable in trying to explain why injuries occur, but it is less important in determining whether or not a specific traumatic event resulted in an injury. For example, one does not require epidemiology to establish that getting hit on the head with a hammer may cause a fractured skull. Epidemiology may help clarify some of the factors that lead to head trauma (not using a helmet, alcoholism, seizure disorders, etc), but is not used to determine whether or not a particular traumatic event caused the injury in question.

Traditional medical causal analysis relies on all information that can be brought to bear on the question of cause-effect relationships. Medical inference is based on chemistry, physics, biochemistry, pathology, molecular biology, and other scientific disciplines. The types of data relied on include observation, experiment, analogy, computer models, animal experiments, epidemiologic studies, and bacteriologic experiments. Doctors consider the whole data set available to establish a causal nexus.

We find it curious that both Hadler and his responders fail to refer at all to Hill's guidelines for determining cause-effect relationships. Of Hill's nine guidelines, only the rate ratio (strength of association) is assessed from epidemiological evidence alone. Most of Hill's other "things to consider" rely on nonepidemiological evidence, mechanistic understanding, biologic plausibility, coherence, analogy, etc.

There is a great deal of animal, ergonomic, pathologic, molecular, and epidemiological evidence in support of the conclusion that repetitive use of keyboards can cause injury. The epidemiological evidence must be evaluated in the context of all of the relevant scientific knowledge.

Hadler appears to be genuinely concerned with socioeconomic issues such as the worker's sick role, the compensation system's interference with the rehabilitation process, and the adverse impact of a physician's financial incentives on treatment choices, although he does not appear to be familiar with the vast medical sociological literature. In his effort to impact on this problem, Hadler selectively picks studies that favor his position and biases his interpretation of those studies to favor his interpretation. The social problems and the medical problems are intertwined, but misrepresenting the science will not help to solve either.

David Egilman, MD

Brown University; Providence, RI

Laura Punnett, ScD

University of Massachusetts; Lowell, MA

Ewa Wigaeus Hjelm, PhD

Arbetslivsinstutet; Solna, Sweden

Laura Welch, MD

George Washington University; Washington, DC