There are good theoretical arguments for exploring the use of replicating gene-transfer vectors for human cancer therapy. Such vectors should be derived from weakly pathogenic human viruses with initially broad tissue tropism. Coat protein engineering and promoter engineering might be used successfully to narrow the tropism of the vector, enhancing its ability to target tumour cells. Killing of uninfected 'bystander' tumour cells could be achieved through prodrug activation by a vector-encoded enzyme. Rapid elimination of infused vector particles by circulating antiviral antibody would limit access to tumour deposits after repeated administration, but might be circumvented by the use of infectious nucleic acid which is poorly imunogenic [64]. This putative therapeutic strategy is illustrated in Figure 1.