Introduction: Common statistical methods for the analysis of the results of RCTs assume independency of observations. There are, however, circumstances in which this assumption is not met, for example in research regarding carpal tunnel syndrome (CTS). If "hands" instead of "patients" are the units of analysis, observations in both hands of patients with bilateral CTS are not independent. In such cases, the analysis should be directed at patients rather than hands, or methods should be applied that deal with the dependencies of the observations.

Objective: To determine the prevalence of dependent observations in published RCTs regarding the treatment of CTS and to assess the appropriateness of the statistical methods that have been used for the management of those dependencies.

Methods: Simple descriptive statistical methods were used.

Results: In 4 out of 13 identified RCTs "patients" were the units of analysis, thus avoiding the problem of dependent observations. In 1 RCT it was not clear whether the dependency problem existed. In the remaining 8 RCTs, CTS patients with bilateral involvement had been included and "hands" were the units of analysis. In the latter 8 RCTs conventional statistical methods had been applied that failed to address the existing dependencies of the observations.

Discussion: Eight out of 13 RCTs regarding the treatment of CTS involved dependent observations, hut inappropriate statistical methods had been applied. These flaws induce a spurious increase of precision and cause bias. In addition, they complicate the interpretation of the results and impede meta-analysis. During the colloquium possible solutions to this problem will be presented.