Radiographic scoring is generally accepted as the "gold standard" to follow the course of rheumatoid arthritis (RA). Several methods have been developed to quantify radiographic changes due to RA. The most widely used methods have been proposed by Larsen and Sharp. Problems of the interpretation of the radiograph and the design of the scoring methods are not resolved. The design of established scoring methods lacks correct correlation to the amount of change on radiographs, leading to confounding of the results of the radiographic analysis. Concerns arise about the correct interpretation of findings on plain radiographs that are caused by having a 2-dimensional picture of a 3 dimensional object. There is no agreement which of the different changes caused by RA on the radiograph have to be considered and which can be neglected. Examples for these problems are given. We compared the established scoring methods of Larsen and Sharp on the same set of radiographs with longterm followup using the correct data analysis to provide the information needed to design an improved scoring method. Details of the study design are reported.