Drug resistance is the main cause of therapeutic failure and death in patients with cancer. However, there have been surprisingly few studies designed specifically to investigate the mechanisms underlying poor treatment response in vivo, compared with the number of phase II and III trials investigating treatment effects. We can now analyse the expression patterns of multiple genes by use of microarrays, rapid gene sequencing, and proteomics, and so need to reassess the way we design clinical trials to take full advantage of these new opportunities. I discuss the concept of clinical studies of chemoresistance in terms of the collection of tumour samples for biological studies, the use of appropriate clinical settings, and the importance of trial design. Ideally, such studies should investigate specific biological features in relation to measurable antitumour effects of single drugs.