Sorption of radionuclides on host rocks is a key process in the safe geological disposal of radioactive waste. For performance assessment (PA) calculations, the magnitude of sorption, expressed normally by a distribution coefficient (K_d), needs to be determined taking into account the rock types and geochemical conditions and associated uncertainty. The K_d setting approach is needed to apply for various rock types and geochemical conditions, and to evolve from site-generic to site-specific stages by considering site-specific information obtained in forthcoming site investigation stage. Such comprehensive K_d setting approach applicable for various conditions and situations was developed based on international state of the art knowledge by integrating three different methods; i) direct use of measured K_d data extracted from the sorption database, ii) transferring procedures by scaling differences between experimental and PA conditions, and iii) thermodynamic sorption models. This integrated approach was tested for crystalline host rock (granitic rock) by comparing example derivation of K_d values and their uncertainties of Cs and Am by the three different methods. The results indicated that the integrated K_d setting approach is effective, and that K_d can be quantitatively evaluated by all approaches when adequate data and models are available. The K_d dataset for key safety-relevant 25 elements in PA calculation for generic granitic rocks was developed based on the direct use of measured data, and compared with the recent K_d dataset in European PA projects. This K_d setting approach allows to estimate the K_d values and their uncertainties for various rock types and geochemical conditions in accordance with the amount of site-specific information in a stepwise manner.