We assigned a total of 131 hemophiliacs infected with HIV-1 into four clusters by applying a 3-way data analysis method. Sequentially acquired CD4+ and CD8+ cell counts obtained longitudinally over an observation period from 1986 to 1992 were analyzed. During the successive observation in this interval, a clustering of patients is not always coincident over all the times, because the cell counts vary with time. Therefore, the 3-way data clustering is to obtain the optimal result of the classification of patients through all the interval of observation. Examining patients' survival after that period, the cumulative mortality rate was highest among the 36 hemophiliacs in Cluster 1. Less mortality was found in Cluster 2, consisting of 49 hemophiliacs and none was reported in Clusters 3 and 4, which included 33 and 13 hemophiliacs, respectively. However, a significantly lower blood viral copy number was found in Cluster 3 than in Cluster 4. A total of six long-term non-progressors was found, five in Cluster 3 and one in Cluster 4, while none was found in Cluster 1 or 2. As demonstrated in this analysis, 3-way data clustering may represent a good data mining technique for handling various types of clinical data.