Marimastat is a specific inhibitor of matrix metalloproteinases that has been shown to be effective in cancer models. A pilot, escalating-dose study of oral marimastat was performed in patients with recurrent colorectal cancer, in whom evaluation of serological response was made by measurement of carcinoembryonic antigen (CEA) levels. The study assessed the safety and tolerability of 4 weeks administration of marimastat, and determined a dose range producing detectable serological effects. Patients were recruited with a serum CEA level greater than 5 ng ml(-1), and rising by more than 25% over a 4-week screening period. Patients were treated for 28 days and entered into a continuation protocol if a serological response or clinical benefit was observed. Pharmacokinetic and safety data determined that groups of patients were recruited sequentially at 25 mg and 50 mg twice daily, and, thereafter, 10 mg twice daily, 10 mg once daily, 5 mg once daily and 20 mg once daily. A biological effect (BE) was defined as a CEA value on day 28 no greater than on day 0; a partial biological effect (PBE) was defined as a rise in CEA over the 28-day treatment period of less than 25%. Of 70 patients recruited, 63 completed the 28-day treatment period, and 55 were eligible for cancer antigen analysis. Examination of the dose-effect relationships provides evidence for a causal relationship between marimastat and biological effects: the proportion of patients with BE or PBE was higher with twice daily dosing (16 out of 25, 64%) than with once daily dosing (11 out of 30, 37%) (P = 0.043, chi2 test). Furthermore, the median rates of rise of CEA fell markedly during treatment compared with the screening period for patients receiving twice daily marimastat (P<0.0001), but not for patients receiving marimastat once daily (P = 0.25). Musculoskeletal adverse events emerged as the principal drug-related toxicity of marimastat, occurring in a dose- and time-dependent fashion. It was concluded that marimastat was associated with dose-dependent biological effects in cancer patients. The occurrence of musculoskeletal side-effects define 25 mg twice daily as the upper limit of the dose range for continuous use in further studies. Therefore, a dose range of 20 mg once daily to 25 mg twice daily seems appropriate for further studies, which should aim to demonstrate the efficacy of the drug in terms of conventional clinical end points and describe the long-term tolerability of this novel agent.