We evaluate measurement properties of common rheumatoid arthritis (RA) assessments. Included are a comprehensive literature review and new data on the reliability and smallest detectable difference (SDD) for different classes of these measures. We found that certain common measures such as joint counts, pain, and patient global all had poor reliability and showed large SDD compared to multi-item measures of physical/psychological function or compared to radiographic measures. We discuss the implications of these findings on the use of composite endpoints such as the ACR20 or the EULAR responder index in RA clinical trials, particularly the introduction of misclassification bias that arises from differential measurement error. Finally, we consider generically how the concept of the SDD might or might not relate to the concept of the minimal clinically important difference.