This paper describes a Petri Net-based model for workflow management in distributed cancer therapy, a domain which is characterized by the involvement of several geographically distributed departments and a long-term treatment. The approach uses hierarchical Petri-Nets; tokens, places and transitions are objects conform to the CORBA/IDL-model to facilitate implementation in a heterogeneous environment. During treatment, generic organizational nets are dynamically refined with disease-specific subnets obtained from an oncological knowledge base to adapt the net in a patient-specific way. This refinement task is mainly processed by a Knowledge Refinement Manager which has been added to a commercial workflow architecture. Currently, a prototype implementing the model, called HEMATOWORK, is developed at Leipzig University.