Direct gene transfer to tumor cells in vivo may have considerable therapeutic potential, provided that a high enough efficiency of gene delivery can be achieved. All other things being equal (e.g. dose, route of administration), a replicating vector should give a higher efficiency of gene transfer than a nonreplicating vector due to local spread from the initially transduced cells. Almost any virus could provide the basis for a replicating gene transfer vector but, for the treatment of human cancer, the choice is greatly constrained by issues of public safety and potential efficacy. Public safety considerations point to viruses of low pathogenicity which are already prevalent in the human population whilst efficacy considerations point to viruses which replicate efficiently in the tumor cells to be targeted. Additional factors influencing the suitability of individual viruses as a basis for replicating vectors include the availability of an infectious molecular clone, the capacity for foreign gene insertion, genomic stability, potential for targeting by coat protein engineering or promoter engineering, cell-cycle dependence and availability of an effective vaccine.