Existing hypoxia-selective cytotoxins (HSCs) are designed to kill only the hypoxic subpopulation in tumours, and to be used in conjunction with other therapies (e.g., radiation). A new class of drugs, hypoxia-activated prodrugs of diffusible cytotoxins (HPDCs) are proposed. These are designed to exploit, rather than merely deal with, tumour hypoxia, by releasing diffusible cytotoxins on bioreduction in hypoxic regions. Such diffusible cytotoxins are required to be much more cytotoxic than the parent prodrug, to be sufficiently stable (half lives from 0.1 to 10 min) to allow them to diffuse up to 200 microns from the hypoxic regions, and to be equally effective against all major tumour cell subpopulations, including non-cycling cells. Nitrogen mustards, which show little cell cycle specificity, which kill cells by a well-understood mechanism (DNA cross-links), and which have stabilities and reactivities able to be predictably controlled by structural variations, are proposed as suitable candidates fur such diffusible cytotoxins. Design parameters for two classes of potential HPDCs are discussed; nitro-deactivated aromatic mustards, and cobalt (III) complex-deactivated aliphatic mustards. Examples of both classes show greater cell-killing activity against intact compared with dissociated multi-cellular spheroids. This suggests they may indeed function as HPDCs, by penetrating to the hypoxic core of the spheroid and there releasing potent cytotoxins which diffuse out to kill surrounding cells at lower oxygen tensions.