Circumvention of multidrug resistance in vitro by resistance modulators is well documented but their clinical use may be limited by effects on normal tissues. We have compared four resistance modifiers, both in terms of modulation of doxorubicin sensitivity in vitro and toxicity in vivo, in order to determine whether it is possible to select agents with clinical potential. Verapamil, D-verapamil and quinidine are all maximally active in the multidrug resistant cell line at about 7 microM and are not cytotoxic at this concentration. The tiapamil analogue Ro11-2933 is a highly potent resistance modulator such that at only 2 microM sensitization is greater than is seen with the other modulators at 7 microM. Since the ID50 concentration for Ro11-2933 is 17.7 microM (5-12-fold less than the other modifiers) we have used isobologram analysis to demonstrate that the interaction with doxorubicin is supra-additive and cannot be explained by additive toxicity. This method of analysis also revealed that when resistance modulation is related to the cytotoxicity of the modulator itself, all four modulators show comparable activity. On the other hand, measurement of the acute toxicity in mice of the modulators did reveal differences. The LD10 for verapamil (51 mg/kg) was about one third of that for quinidine (185 mg/kg) and this is consistent with the known maximum tolerated plasma levels in patients. Furthermore, whilst epirubicin alone was unable to reduce the growth rate of a multidrug resistant human tumour xenograft, the addition of quinidine, but not verapamil, at the maximum tolerated dose did do so. D-Verapamil was only about half as toxic as racemic verapamil and this too is consistent with clinical observations. The LD10 for Ro11-2933 (152 mg/kg) was comparable with that for quinidine. In the human tumour xenograft model maximal growth inhibition was observed with the combination of epirubicin and Ro11-2933 (45 mg/kg) and this degree of growth inhibition was comparable to that obtained with epirubicin alone in the drug sensitive xerografts. Ro11-2933 had no measurable effects on the plasma or tumour pharmacokinetics of epirubicin. These results suggest that it is possible to predict the clinical potential of a resistance modulator. Furthermore, Ro11-2933 is a promising agent for use in the clinic since maximal resistance modulation in vivo is observed at about one third of the LD10 dose.