The extremely different administration schedules that are used in testing recombinant interleukin-2 (rIL-2) therapies for cancer may account for the extreme variation in efficacy reported in various studies in animal models. A major point may be the variation of the time interval between tumor transplantation and rIL-2 therapy. We hypothesized that administration of rIL-2 before the immune system has mounted a specific cellular reaction against the tumor (associated antigens) might result in lesser efficacies than later rIL-2 administration. This hypothesis was tested in DBA/2 mice bearing a syngeneic SL2 lymphoma. When 7000 IV/day rIL-2 was administered to tumor-bearing mice for 5 consecutive days starting on day 1, 3, 4, 5, or 6 after tumor inoculation, the survival curve of the mice did not significantly differ from that of diluent-treated mice. In contrast, a significant difference was observed when treatment was begun on day 7, 8, 9, 10, or 12 (p < or = 0.004). rIL-2 therapies begun on day 9 or 10 were most effective, curing up to 80% of mice treated, despite there being an enormous burden of disseminated tumor present at that time (1-4% of the total body weight). When rIL-2 was administered for fewer than 5 consecutive days, beginning on day 10, the efficacy of the therapy dropped radically (p < or = 0.055). Involvement of a specific anti-tumor reaction was also tested. All mice that were cured of the tumor as a result of rIL-2 therapy proved to be specifically immune to the SL2 tumor. Furthermore, day 10-14 administration of rIL-2 was completely ineffective in CD4(+)-cell depleted mice (p = 0.0116 vs. rIL-2 therapy in non-depleted mice). Together, this implies that this form of rIL-2 therapy is mediated by tumor-specific T-cells. As a whole, these results indicate that T-cell mediated rIL-2 therapy of cancer in animal models is sensitive to the time when the rIL-2 is administered and to the length of time for which the rIL-2 is given. This should be taken into account when planning new therapy protocols and when analyzing published data.