The concept of generating cytotoxic agents from non-toxic prodrugs at tumour sites by antibody vectored enzyme introduces a wide range of opportunities. Various prodrug-enzyme combinations have been described and encouraging results reported in xenograft models. Whilst the mouse model is a valuable tool in this approach translation to the human patient may expose more complex issues. The objective of restricting drug action to tumour sites and thus allowing greatly increased cytotoxic action requires more precise restriction of enzyme activity to tumour sites than has been achieved with an antibody vector and natural clearance alone. Assisted clearance mechanisms have been found effective. Alternatively, or additionally, the difference between prodrug and active drug creates the opportunity to degrade active drug selectively in blood and thus protect normal tissues. In order to give more than one cycle of treatment it will be necessary for the antibody-enzyme conjugate to be nonimmunogenic or for the concurrent administration of immunosuppressive agents. A pilot scale clinical trial with a prototype prodrug indicated the feasibility of antibody directed enzyme prodrug therapy (ADEPT).