When three variables require simultaneous adjustment for treatment optimization, the experimental determination of the optimum treatment becomes more complicated than generally appreciated, especially when one variable is the interval between drug administrations. Molecular pharmacological studies at this institution suggest that teniposide, in doses that are achievable in vivo, blocks methotrexate efflux from cells, enhancing formation of methotrexate polyglutamates which are active retentive forms of the drug. Hence, the combination might show a synergistic effect if teniposide is given at an appropriate time in relation to administration of methotrexate. This paper considers the problem of estimating the optimal dose levels and timing of administration of these drugs in B6D2F1 mice bearing L1210 leukemia in vivo as a model for the analysis of multidrug regimens when time and dose are variables. Because of the complexity of these experiments, an adaptive approach was applied. Three cycles of treatment were given using methotrexate at 0-400 mg/kg, teniposide at 0-60 mg/kg, and an interval of 0-54 hours between the two drugs. The single drugs prolonged median survival up to 24 days under the conditions of these experiments. The combination given simultaneously resulted in median survival of up to 33.5 days, while use of an appropriate interval between drugs prolonged median survival to greater than 50 days. An analysis of the underlying dose-time response gives a much better appreciation of the relationships among the variables. The concept that optimal doses included less methotrexate and more teniposide as the interval between drugs is increased was developed only through modeling. This finding is critical in demonstrating the importance of including all nonnegligible variables in the experimental design if the results are to be considered valid. Confidence regions about the optimum dose and about the response at the optimum provide a sound basis for a claim of therapeutic synergism as demonstrated by use of a scheduling variable in the experiment design. A nonproportional hazards analysis permits the conclusion of nonproportionality and emphasizes the contribution of methotrexate to optimal short-term survival (15-24 days) and teniposide to long-term survival (24-39 days).