Targeting cancer gene therapy to endothelial cells seems to be a rational approach, because (a) a clear correlation exists between proliferation of tumor vessels and tumor growth and malignancy, (b) differences of cell membrane structures between tumor endothelial cells and normal endothelial cells exist which could be used for targeting of vectors and (c) tumor endothelial cells are accessible to vector vehicles in spite of the peculiarities of the transvascular and interstitial blood flow in tumors. Based on the knowledge on the pharmacokinetics of macromolecules it can be concluded that vectors targeting tumor endothelial cells should own a long blood residence time after intravascular application. This precondition seems to be fulfilled best by vectors exhibiting a slight anionic charge. A long blood residence time would allow the formation of a high amount of complexes between tumor endothelial cells and vector particles. Such high amount of complexes should enable a high transfection rate of tumor endothelial cells. In view of their pharmacokinetic behavior nonviral vectors seem to be more suitable for in vivo targeting tumor endothelial cells than viral vectors. Specific binding of nonviral vectors to tumor endothelial cells should be enhanced by multifunctional ligands and the transduction efficiency should be improved by cationic carriers. Effector genes should encode proteins potent enough to induce reactions which eliminate the tumor tissue. To be effective to that degree such proteins should induce self-amplificating antitumor reactions. Examples for proteins which have the potential to induce such self-amplificating tumor reactions are proteins endowed with antiangiogenic and antiproliferative activity, enzymes which convert prodrugs into drugs and possibly also proteins which induce embolization of tumor vessels. The pharmacological data for such examples are discussed in detail.