The interest in TNF, discovered at the interface between inflammation and cancer, as an anti-cancer agent, has phased out in recent years. Indeed, despite its profound cytostatic and cytotoxic effects in primary tumors, the cytokine's systemic toxicity in general and its hepatotoxic and pro-metastatic nature in particular, prevent its routine use in cancer patients. An elegant approach to circumvent these problems consists in the local application of TNF in an isolated limb or organ setting, preferentially in the presence of cytostatic and alkylating agents, such as melphalan. However, this treatment, when locally applied during the perfusion of liver tumors, results in hepatotoxicity in a significant number of the patients, by means of a still unknown mechanism. The hemorrhagic necrosis of the tumors induced by TNF seems to be predominantly mediated by an induction of apoptosis as well as by an anti-angiogenic effect in the endothelial cells of the microvasculature supplying the tumor. These cells therefore represent a prime target for the action of anti-cancer drugs. In this review, we discuss preclinical studies which elucidated the mechanism of melphalan- and TNF-associated hepatotoxicity and, as a consequence, provided insights for preventing the adverse reactions of the drug. Moreover, we review recent findings aimed at improving the TNF molecule by means of specific mutations, or searching for alternative factors lacking the systemic toxicity of TNF.