Objective/background: Efficacy in osteoarthritis (OA) is principally measured using subjective visual analogue (VAS) and/or Likert scale responses. The relationship between these two scales and their relative precision in discriminating active from placebo treatment in OA patients was determined.
Design/methods: Patient overall pain assessment, and patient and investigator global assessments were each measured on a 100mm VAS and on a 0 to 4 point Likert scale in a 6-week OA study of rofecoxib vs placebo. The relationship between the VAS and Likert responses was examined graphically and via summary statistics. Analysis of variance was used to assess consistency of the VAS/Likert relationship over time and across the different endpoints. Precision was compared using effect size, and normality of VAS scale of measurement was assessed using the Shapiro-Wilk test.
Results: Mean VAS scores and changes from baseline at individual time points were generally highly correlated with corresponding Likert responses (r-values generally approximately 0.7-0.8). The magnitude of VAS values and changes varied depending on endpoint, on the associated magnitude of increment of Likert score, and on the Likert baseline value (i.e., where on the Likert scale the change was occurring). Precision of VAS and Likert responses to detect difference between treatments was generally similar with effect sizes approximately 1. Normality and homogeneity of variance of VAS scores was most closely approximated by actual changes in comparison to percent change or log-transformed measures.
Conclusions: VAS and Likert responses are highly correlated and yield similar precision for discriminating treatments in OA patients. Since Likert responses are easier to administer and interpret, they may be preferable to measure OA response.