Defining and analyzing cohorts using molecular markers of cancer risk

Cancer is currently regarded to be the phenotypic expression of an accumulation of heritable alterations in the regulators of cell growth and differentiation. Though detailed knowledge of the sequence and in vivo mechanistic effects of these alterations is rudimentary for most, if not all, cancers, their identification does offer the potential for classifying groups of individuals who are heterogeneous with respect to their cancer risks, into more nearly homogeneous subgroups. In this paper, we illustrate the value of using markers, which we define as any manifestation of cellular molecular diversity, to increase subgroup homogeneity. In the context of time-to-event data, we demonstrate for both somatic mutations (acquired p53 abnormalities in gastric mucosal cells) and inherited polymorphisms (polymorphisms in the phase 1 and 2 detoxifying enzymes) how knowledge regarding the population frequency of the marker the effect of the marker on the risk of cancer development, and/or the effect of the marker on response to therapy, can be used to plan and analyze such trials. Using as paradigms demographic features of the recently begun Shandong precancerous gastric lesion intervention trial, and the recently completed alpha-tocopherol beta-carotene (ATBC) lung cancer prevention study, we review the information, assumptions, and mathematical structure required for planning cancer prevention trials. We graphically demonstrate how informative markers make available strategies for selection, stratification, and optimal weighing, which, when properly implemented, increase the power of tests of effective cancer prevention agents.