The purpose of this grant application is to further develop the continual reassessment method, a design scheme for phase I clinical trials in cancer. The focus concerns methodological improvements that enable the issue of patient heterogeneity to be addressed. The motivation stems from the fact that the right treatment level for certain groups of patients may not be the right treatment level for other groups. Currently all patients are included in such trials without regard to clear groupings such as gender differences. For gender specific diseases such as breast cancer or prostate cancer this is not a concern but more generally, for example in studies in lung cancer which may include women, even if in a minority, it is likely that ignoring gender differences may lead to unacceptable errors in recommended treatment level. In the presence of gender differences the final recommended level of a Phase I study will be some sort of average across the genders. This average will be naturally weighted according to the percentage of each group so, should women enter the study as minority, as is the case in most lung cancer studies, the final recommended dose is likely to be much more appropriate for males than for females. The same applies should the males be in a minority and, perhaps worse still in the case where there is a balance across the genders, we may end up with a recommended dose that is """"""""correct"""""""" on average but that is inappropriate for both males and females. This is a difficult issue to address in view of the typically small sample sizes, in particular should one group be very much smaller than the other. However our preliminary findings suggest that careful modeling can be of very considerable help. The information obtained in one group can be of great use in the other group although, as we anticipate, the correct levels for the groups are different. We have considered modifications of the continual reassessment method that can be applied to two groups of patients to determine appropriate target dose levels for each groups. As a first step the method takes the specification of a simple relationship between the dose toxicity curves for the two groups and runs the continual reassessment method on the bivariate model using maximum likelihood. We have demonstrated consistency of the method under fairly weak conditions and provide several simulations to give an idea how the method works in practice. We anticipate our working models to be incorrect in this area and we are planning to study the effect of severe model inadequacy on the procedure.
| O'Quigley, John (2002) Curve-free and model-based continual reassessment method designs. Biometrics 58:245-9 |
| O'Quigley, J; Shen, L Z; Gamst, A (1999) Two-sample continual reassessment method. J Biopharm Stat 9:17-44 |
