We have conducted a preliminary study of the optimum conditions for a therapeutic effect of ganglioside GM1 in Alzheimer's disease. Five patients with the early onset form of Alzheimer's disease (AD type I) received the ganglioside by intracerebroventricular administration for 12 months. Bilateral stereotactic punction of the frontal horns of the ventricular system was performed, and shunt catheters were implanted and connected to a programmable pump. The optimum GM1 dose varied between 20 and 30 mg/24 h. Neurological neuropsychological, psychiatric and neurochemical examinations were performed 7 days before surgery and on days 30, 90, 180 and 360. No patient found the surgery difficult and no patient or relative regretted that they participated in the study. The patients became more active and safer in relation to others and to performance of various activities from day 90. The cerebrospinal fluid level of the monoamine metabolites homovanillic acid and 5-hydroxyindoleacetic acid and the neuropeptide somatostatin increased.