Delivery of cytokine genes at the tumor site in pre-clinical models has been shown to recruit host inflammatory cells followed by inhibition of tumor growth. This local effect is often accompanied by systemic protection mediated by the immune system, mainly by CD8(+) T and NK cells. On this basis, cytokine gene-transduced tumor cells have widely been used as vaccines in clinical trials, which have shown good safety profiles and some local responses but substantial lack of systemic efficacy. Are these findings the end of the story? Possibly not, if major improvements will be attained in the coming years. These should be directed at the level of gene selection and delivery, in order to identify the optimal cytokine and achieve efficient and durable cytokine expression, and at the level of improving immune stimulation, i.e. by co-administration of co-stimulatory molecules including B7 and CD40, or boosting the expression of tumor antigens or MHC class I molecules. Interestingly, some of the cytokines which have shown encouraging anti-tumor activity, including IFNs, IL-4, IL-12 and TNF-alpha, are endowed with anti-angiogenic or vasculotoxic effects, which may significantly contribute to local tumor control. Therapeutic exploitation of this property may result in the design of novel approaches which, by maximizing immune-stimulating and anti-angiogenic effects, could possibly lead to starvation of established tumors in patients.