Summary Coumarin anticoagulants inhibit metastasis in several animal models, but the mechanism of this effect is uncertain. In order to determine the role of cytotoxic and/or cytostatic actions of coumarins on the tumour cells, we have studied the effects of warfarin on tumour cell growth in a model in which tumour metastasis is inhibited by this drug. Clonogenic assay, growth curve analysis and thymidine labelling index revealed that warfarin had no effects on Mtln3 mammary carcinoma cell growth in vitro at concentrations below 1 mm. The growth rate of subcutaneously implanted Mtln3 tumour deposits in female F344 rats, assessed by weight and by stathmokinetic analysis of the tumour tissue, was identical in warfarin-treated and control animals. Spontaneous metastasis from such tumours to the lungs was, however, significantly reduced in warfarin-treated animals (median 0 pulmonary tumours per animal in warfarin treated, eight tumours per animal in control animals; P<0.05, Mann-Whitney). The mean plasma warfarin concentration in warfarin treated rats was 1.63 M. These results suggest that warfarin treatment of the host animal can inhibit tumour metastasis without having any direct or indirect effect on the growth rate of the tumour cells.Current theories (Hart, 1980;Poste & Fidler, 1980) view metastasis as a multistep process in which successive obstacles are overcome by a small subpopulation of tumour cells capable of doing so. Each step requires different properties, and each influences subsequent steps. This complex process cannot be studied as a single unit, but requires subdivision: one way in which this can be achieved is to study influences that increase or decrease metastasis, and to attempt to define the point at which they have their effect. The coumarin group of anticoagulant drugs, including warfarin, represent an example of such an influence. Coumarins inhibit metastasis in several animal models (Ryan et al., 1969;Brown, 1973;Hilgard et al., 1977;Williamson et al., 1980). There is an extensive literature documenting the existing of coagulation disturbances in human cancer (Davis et al., 1969;Sun et al., 1979;Rickles & Edwards, 1983; Mannucci et al., 1985), and suggestive evidence of a role for coagulation in the spread and growth of tumours (Dvorak et al., 1979; Goeting et al., 1985;McCulloch & George, 1987 (Colucci et al., 1983). The possibility that coumarins have cytotoxic properties has been investigated in several different models (Boulos, 1971;Higashi & Heidelberger, 1971;Brown, 1973;Chang & Hall, 1973;Dolfini et al., 1979;McNeil et al., 1984) with diverse results. The variety of models employed and the frequent use of in vitro measures of cytotoxicity without reference to the effects of the drugs in vivo make interpretation of these studies particularly difficult. Several studies have noted a possible suppressive effect of coumarin treatment on primary tumour growth (Ryan et al., 1968(Ryan et al., , 1969Hilgard et al., 1977), but have used only crude methods which are prone to random error. On...