An assessment of the reliability of three methods used in evaluating the status of multiple sclerosis patients.

IN THE past few years, increasing effort has been given to the development of quantitative methods for the evaluation of neurologic function so that more objective methods could take the place of subjective methods in the evaluation of therapeutic response. Attempts to estimate the extent and degree of deficits in neurologic function by numerical methods have raised questions about the reliability of such scores. Do they accurately reflect degrees of neurologic dysfunction? More specifically, can such scores be depended upon to portray accurately the changes in degree of a patient’s dysfunction from 1 examination to another? Do several examiners, scoring dysfunction for the same patient, obtain the same score? It may also be asked whether several examiners will observe the same increment of change in a numerical score, representing a change in function between successive examinations of the same patient. These questions are particularly pertinent in collaborative studies where appropriate patient evalution is frequently based on numerical scores and changes in scores as obtained at several investigating centers. The issues of accuracy and repeatability were particularly relevant to a recent cooperative study evaluating ACTH therapy in Multiple Sclerosis [I]. Four different methods of evaluating patient dysfunction were used, and the separate items of each method were recorded as numerical scores or were changed to numerical scores when the items were recorded as “slight”, “moderate”, or “severe”, etc. Since patients were evaluated by several investigators who submitted data to one statistical center, it was considered essential that a test giving an indication of the reliability of the clinical evaluation methods be performed prior to the analysis of the data. This report gives the results of the tests of the reliability of 3 of the 4 evaluation methods and presents a statistical design that is considered efficient relative to other