The 238U-230Th-226Ra and 235U-231Pa disequilibria have been measured by mass spectrometry in historic lavas from the Kamchatka arc. The samples come from three closely located volcanoes in the Central Kamchatka Depression (CKD), the most active region of subducted-related volcanism in the world. The large excesses of 226Ra over 230Th found in the CKD lavas are believed to be linked to slab dehydration. Moreover, the samples show the uncommon feature of (230Th/238U) activity ratios both lower and higher than 1. The U-series disequilibria are characterized by binary trends between activity ratios, with (231Pa/235U) ratios all >1. It is shown that these correlations cannot be explained by a simple process involving a combination of slab dehydration and melting. We suggest that they are more likely to reflect mixing between two end-members: a high-magnesia basalt (HMB) end-member with a clear slab fluid signature and a high-alumina andesite (HAA) end-member reflecting the contribution of a slab-derived melt. The U-Th-Ra characteristics of the HMB end-member can be explained either by a two-step fluid addition with a time lag of 150 ka between each event or by continuous dehydration. The inferred composition for the dehydrating slab is a phengite-bearing eclogite. Equilibrium transport or dynamic melting can both account for 231Pa excess over 235U in HMB end-member. Nevertheless, dynamic melting is preferred as equilibrium transport melting requires unrealistically high upwelling velocities to preserve fluid-derived 226Ra/230Th. A continuous flux melting model is also tested. In this model, 231Pa-235U is quickly dominated by fluid addition and, for realistic extents of melting, this process cannot account for (231Pa/235U) ratios as high as 1.6, as observed in the HMB end-member. The involvement of a melt derived from the subducted oceanic crust is more likely for explaining the HAA end-member compositions than crustal assimilation. Melting of the oceanic crust is believed to occur in presence of residual phengite and rutile, resulting in no 226Ra-230Th disequilibrium and low 231Pa excess over 235U in the high-alumina andesites. Consequently, it appears that high-alumina andesites and high-magnesia basalts have distinct origins: the former being derived from melting of the subducted oceanic crust and the latter from hydrated mantle. It seems that there is no genetic link between these two magma types, in contrast with what was previously believed.