Objective. - To assess the test-retest reliability and the construct validity of a modified version of the French-Canadian version of the WOMAC index.

Methods. - Open prospective study conducted in departments of Rheumatology, Rehabilitation and Orthopedic surgery of a tertiary care teaching hospital. Eighty-eight patients with symptomatic knee osteoarthritis (OA) fulfilling the revised criteria of the American College of Rheumatology were included. The French-Canadian version of the WOMAC index was completed twice at a 3-h interval. Impairment outcome measures, patients' perceived discomfort in walking and handicap were recorded. An item-by-item analysis was performed. Test-retest reliability was assessed using the intra-class correlation coefficient (ICC) and the Bland and Altman method. Construct validity was investigated using the Spearman rank correlation coefficient and a factor analysis was performed.

Results. - Eight questions assessing function (section C) and the two questions assessing stiffness (section B) had insufficient psychometric properties and were excluded. Although test-retest reliability of the questionnaire was fair to good (0.82, 0.85, for the WOMAC section A, and modified section C, respectively), construct validity could not be demonstrated. Factor analysis of the modified form of the WOMAC extracted four factors, which differed from the a priori stratification. However, factor analysis of the modified section C extracted two factors explaining 68.4% of the variance, which could be clinically characterized.

Conclusion. - Despite its good test-retest reliability, the modified WOMAC index is not valid for assessing pain and disability induced by knee OA in a French population. Section A and modified section C could be used separately to assess, respectively, pain and function.