The ability to alter human tumour cells geneticallyin vivoprovides a variety of new opportunities to selectively destroy malignant cells.The herpes simplex thymidine kinase gene (HS-tk) confers a sensitivity to the antiherpes drug ganciclovir. Insertion of HS-tk into tumours and subsequent treatment with ganciclovir has successfully eliminated tumours in experimental animal models, despite a less than 100% gene transfer efficiency. This phenomenon, the ‘bystander effect’, allows the destruction of neighbouring tumour cells not transduced with HS-tk. Since there is no gene transfer method that is 100% efficient, the bystander effect makes the possibility of using gene therapy for the treatment of brain tumours a reasonable approach in patients with recurrent or metastatic CNS tumours.Human experimentation with this approach began in December 1992. Results from an initial trial have shown that the HS-tk/ganciclovir system can selectively destroy tumour cells with minimal toxicity. Despite the bystander effect, the magnitude of the antitumour effect is currently limited by insufficient gene delivery. Since the HS-tk system is a potent method for tumour cell destruction that is not limited by toxicity, further improvements in gene transfer efficiency may allow the development of a clinically useful therapy for the treatment of CNS malignancies.