The release rate of an active pharmaceutical ingredient (API) from a dosage form, measured by a dissolution test, is one of the key parameters governing the formulation viability. A number of dissolution tests have to be performed during the life cycle of a dosage form. This number can be significantly reduced by prediction of the API dissolution behavior. Thus the aim of this work was to find if, and how precisely, it is possible to predict dissolution behavior of final dosage forms by carrying out the experiments with their precursors. In this purpose the rate of release of a model API – caffeine –from a powdered, granulated and compacted form was observed. The experiments were performed in a flow-through cell dissolution apparatus (USP 4) with agglomerates and in a paddle dissolution apparatus (USP 2) with tablets and capsules. It was found that the caffeine release from the agglomerates was retarded by the excipient matrix and that compacted caffeine dissolved faster than granulated caffeine. However, the dissolution profiles were changed when model forms were compressed and then ground up. Consequently, the release rate of the granulated caffeine exceeded the release rate of the compacted form. Facts which were found during the measurement of the agglomerates and final dosage forms were in compliance, hence it was shown that dissolution behavior of final dosage forms can be predicted from their precursors