Levamisole in rheumatoid arthritis--a multivariate analysis of a multicentric study.

Statistical analysis of the data from a multicentric study of six months of treatment with levamisole in rheumatoid arthritis involved two steps. In a first step, the differences between observations before and after treatment have been made independent of their initial values. Differences after treatment have been transformed into responses to treatment using a proportionality factor that is related to the initial values before treatment. The second step involved a reduction of the various responses into a single global response by means of a discriminant analysis between two classes. Class membership has been defined by the type of treatment (either placebo or drug) predicted by the investigators at the end of treatment and before breaking of the code of the double-blind study. This method of analysis allows for the combination of multiple clinical observations with a subjective evaluation. Its results can be easily represented graphically and tested for statistical significance. The use of global responses to treatment is illustrated with a comparison of placebo and drug treatments and can be extended to comparisons between investigators or between different dose and time regimens.