A placebo controlled, randomized, double blind, multicentric clinical trial was carried out in Italy to test the effectiveness of a drug in the therapy of Alzheimer's disease. Informed consensus was gained for the inclusion of 130 patients, whose mental performance was assessed before allocation and after 3, 6, 12 months of continuing assumption of treatment. 14 different outcome variables were used. The main analysis was performed on measures adjusted for baseline, using both univariate and multivariate techniques for each of the three assessing times. The former is mainly descriptive in its scope and allows an analytical view of the results; the latter protects from a too high type I error probability, without implying the loss of power produced by the Bonferroni correction. The techniques are respectively Mann-Whitney test, and analysis of deviance of logistic models, which was suggested by Cupples et al. (1984). In the latter we used the allocation of the patients as response variable, and compared the predictor containing all outcome measures with that having just a constant term. It is shown that a significant difference of fit implies that treatment significantly affects outcome. For the 3 and 6 month follow-up assessment, univariate tests do not show clearcut trends (only 1 outcome measure is significantly better for the drug, none is for the placebo). After 12 months, a trend shows: 13 out of 14 variables are in favour of the drug and 4 of them are significant. Of the three multivariate test carried out, that relative to the 12-month follow-up is significant at the 0.05 level, which is consistent with the univariate results.(ABSTRACT TRUNCATED AT 250 WORDS)