Toward the Last Cohort

Open Access
Publisher Website
Google Scholar
Abstract
Two essential elements contribute to the risk of all diseases, especially cancer: genetic variation and environmental exposures. The completion of a solid draft of the human genome will allow researchers to characterize genotypes to ever-finer degrees of detail, and to continue to identify genetic traits associated with disease in individuals and families. However, much of the genetic variation that is associated with cancer seems to modify risk only in the presence of environmental variability, both for exposures that increase, and those that decrease, risk. There are 10- to 200-fold differences in rates of disease across different geographic locations around the world and, over 50 years, up to 10-fold differences when comparing the same geographic location over time. These cannot be explained by differences in genes, but only by differences in exposures, modified by the interaction between genetic variation and such exposures.A broad array of exposures influences the presence or absence of disease in humans. The accurate characterization and measurement of many of the environmental exposures is difficult but there is extensive experience in the epidemiologic community. The human species has adapted to cultures, diets (both marginal and excessive), microorganisms and parasites, toxic exposures, and bad habits, in environments from the equator to the poles. We are an outbred species that, nonetheless, passed through a genetic bottleneck perhaps 150,000 years ago; the degree of fit between phenotype and environment is still undergoing rapid change. Thus, there is a wide variety of genetic susceptibilities to, and protection against, these exposures. It follows that an evaluation, both of variation in environmental exposures and of genetic variation, is essential to establishing the full picture of the causes of disease.Unlike the increasing ease with which genotypes (considered at small scale or at genomic level) can be generated, characterizing disease phenotypes remains problematic. There are myriad classification schemes that range from precise molecular characterization, such as that of bcr-abl leukemia, to vague, very heterogeneous, syndromes, such as schizophrenia. This imprecision and inconsistency is important for three reasons. First, specific molecularly defined subsets of diseases have been shown to be associated with particular exposures. For example, smoking is associated with hyperplastic polyps, in both the presence and absence of adenomas but not with a “pure adenoma” phenotype (1) and with MSI+ but not MSI− colon cancer (2). Second, variation in specific molecularly defined subsets of cancers explains, in part at least, variation by time and place. For example, the international variation in breast cancer is explained extensively by variation in ER+/PR+ breast cancer, with risk of other receptor-defined subsets being less varied (3). Third, there exists extensive evidence that molecularly defined subsets of disease also carry very different responses to therapy and prognoses (4, 5).For each of these reasons, it is essential to work toward a better—molecular—classification of disease, particularly of cancer, rather than continue to rely extensively on histopathology. The key to this problem is the collection of fresh tissue at the time of diagnosis or treatment, ensuring the ability to characterize tumors, especially, using protein profiles or mRNA expression—and in a setting that allows these findings to be related to exposure and genetics on the one hand, and response to therapy and outcome on the other.The greater the degree of precision of molecular phenotypic classification—and, thus, the greater the homogeneity of disease subsets—the higher the likelihood of being able to reduce susceptibility or increase resistance (prevention), to detect early disease, and, thus, treat at the earliest opportunity. As more outcomes accumulate with time, in a well-characterized and well-followed cohort and as molecular classification schemes improve, there will be an increasing capacity to define and refine homogeneous disease subsets.Early detection of disease, again especially cancer, is important because, by and large, a disease caught early is treated more readily, more simply, with fewer complications, and with better survival. Markers of an early disease need a variety of characteristics (6). Among the most important is specificity, which in this context, has two aspects: first, distinguishing disease from non-disease, and second, distinguishing disease that will present clinically from disease that remains indolent beyond the life span of the individual. It is clear that some current early detection methods work well in achieving the first kind of specificity but may be doing more poorly than we thought at the second; a recent article of Zahl et al. (7) provides the best evidence yet that methods for early detection need the capacity not only to detect disease but also to distinguish between disease that will progress and disease that will remain indolent.Among the best opportunities to develop early detection profiles that achieve this aim is to follow a large number of individuals over a prolonged period, sampling blood and other body fluids at multiple time points to establish, say, protein profiles with predictive power derived from both a cross-sectional picture and change over time (8).To attempt to establish the complete pattern of human disease susceptibility and resistance, to establish longitudinal profiles as early-detection markers, and to identify more precise phenotypes, what is needed is a study of a very large number of ethnically diverse individuals who are well characterized genetically, whose exposures are diverse and well mapped, and whose illness pattern and mortality can be monitored. This cohort, if we do it correctly, could be “The Last Cohort”; it would examine the impact of exposures on...