Objective: To develop criteria for the selection and application of molecular markers for the study of osteoarthritis (OA).
Methods: Statistical criteria for marker selection for OA are developed.
Results and conclusions: After studying more than 20 different molecular markers for monitoring OA, procedures for choosing markers for clinical studies have been developed. For a particular study, the process starts with the markers showing 'face-validity' for monitoring OA. They are next required to successfully distinguish OA patients from controls. This necessitates definition of the distribution of marker values in OA patients and controls. So far, they have been consistently log-normal. The difference (Delta) in marker values between OA and controls defines the opportunity for marker improvement. The between-visit standard deviation (S) in patients puts limits on the detection of marker changes. The two variables can be combined to estimate the practicality of a marker using a modified power analysis. The number of patients (N*) required to observe a 50% improvement with an alpha level of P=0.05 and with 80% certainty is estimated as 50(S/Delta)(2). N*, S and Delta should be used to characterize and compare markers. Marker efficiency can be refined by regressing on secondary variables, such as age, sex, BMI, severity, etc. Finally, the use of two or more markers may be required to improve marker prediction of clinical outcome. Correlated markers can be used to reinforce conclusions by essentially adding replicative data. Independent, complementary markers can be used to develop associations with clinical parameters, and perhaps diagnose and monitor disease status, activities that so far have not been possible with single markers.