The 84 isoform of apolipoprotein E (ApoE) has been proposed as a risk factor for Alzheimer's disease (AD), while the possible role of the epsilon2 allele in AD is controversial. We have studied the ApoE genotype in 38 patients with early-onset AD (EOAD) and in 43 patients with late-onset AD (LOAD). In the EOAD group we observed a significant increase of epsilon4 allele frequency as compared with normal controls, while there was a more than 3-fold decrease of epsilon2 allele frequency that did not reach statistical significance. In the LOAD group we found a highly significant increase of epsilon4 allele frequency as compared with normal controls, while there was a significant decrease of epsilon2 allele frequency. In both the EOAD and LOAD groups, no significant difference was observed between epsilon4 carriers and epsilon4 noncarriers as for age at disease onset, disease duration, and Mini-Mental State score at observation. However, in both EOAD and LOAD groups a statistical trend towards a longer disease duration was observed in epsilon4 carriers. In both the EOAD and LOAD groups, disease severity was compared in epsilon4 carriers versus epsilon4 noncarriers by means of analyses of covariance, with disease duration as covariate. No significant difference between epsilon4 carriers and epsilon4 noncarriers was observed in both EOAD and LOAD. The results of the present study confirm that epsilon4 allele seems to be associated with an increased risk for sporadic AD, while the significant decrease of epsilon2 allele frequency in the LOAD group supports the hypothesis of a possible protective role of epsilon2 allele in AD.