The validity of different definitions of radiographic worsening for longitudinal studies of knee osteoarthritis

Abstract 

Several definitions have been used to characterize radiographic worsening of knee osteoarthritis in longitudinal studies, yet a valid definition with maximal power to detect differences between groups is not known. We used serial radiographs from the Framingham Osteoarthritis Study to compare five dichotomous definitions according to construct validity (strength of association) and discriminant power (power to reject null hypotheses of no difference) for 1) known risk factors for knee osteoarthritis, and 2) development of new knee pain. For risk factors: definitions that included scores for osteophytes (bone spurs) showed good construct validity and discriminant power; a definition using the Kellgren and Lawrence grade of overall knee osteoarthritis was conservative with good construct validity but low discriminant power; a definition based solely on ordinal assessment of joint space narrowing had weak construct validity and low discriminant power. All definitions had comparably strong associations with the development of new knee pain. Similar associations with new knee pain were found when the analysis was confined to either knees with no osteoarthritis at baseline or knees with prevalent osteoarthritis, with increased standard errors for prevalent osteoarthritis. Use of any of these definitions, other than joint space narrowing alone, would permit detection of associations with most known risk factors. Definitions incorporating both osteophytes and joint space narrowing offer the most precise estimation of the association of risk factors with disease worsening.

Keywords:  Construct validity, Discriminant power, Framingham Study, Longitudinal data, Progression, Responsiveness