Use of response surface statistical designs to detect effects of biologic response modifiers such as IL-2

Recently many biologic response modifiers have been discovered or produced, but their effects are incompletely understood. The effects of these agents are complicated by their dependence not merely upon the administered dose, nor even the interval of time between administration and observation of effect, but also because of the cascade of interactions that can be invoked. Conventional experimental designs that test a specific combination looking for statistical significance, and that involve doing a series of studies varying one parameter at a time, are likely to require large numbers of animals or patients to answer the question of possible effect, and have a high likelihood of missing the effect altogether because of not selecting the correct combination of parameter levels. Response surface designs are intended for simultaneous multiparameter evaluation and so are less likely to miss an effect. However, if fixed sample designs are employed, the number of subjects required may still be quite large. Sequential designs based on response surface models have been used in other fields but are only now being applied in the fields of biology and medicine. We have explored the use of these designs to evaluate the potential usefulness of interleukin-2 (IL-2) to increase tumor uptake of subsequently administered antibody in nude mice. We found that these methods enabled us to efficiently determine optimal combinations of dose of IL-2 and interval of time between IL-2 and antibody administration for enhancement. Enhancement of tumor uptake of the antibody by IL-2 was substantial in amount and sufficient to stimulate studies in patients where we will use similar sequential methodology for trial design.