Epidemiology and OMICS: Populations or Individuals?

Epidemiology has been used to identify factors which increase or decrease risk of specific health conditions, mortality, trends, survival, and health care service effectiveness. For as much as epidemiology has come into the mainstream, especially in news stories, policy debates, legal deliberations, and regulatory decisions, it is a science of focused drawing inference from populations. Classically populations in epidemiologic studies may be defined broadly, as in the entire population of a geographic region, or more narrowly, such as a group of individuals who have share certain characteristics, a specific disease, or specific exposures. These populations are measured at a specified point or period in time, and some estimate of risk of a particular outcome associated with a specified exposure is usually calculated. These measures are ideally accompanied by a statistical parameter that suggests the likelihood that the observed association, or one more extreme, is due to chance alone.

The next phase in the development of epidemiologic studies included biomarkers such as estrogen fraction measurements in urine, cotinine levels among smokers or those exposed to smoke, antibody titers, T-cell ratios, etc. These helped refine hypotheses, and identify groups of individuals at higher risk of a specific disease, or helped determine when to begin treatment.

Epidemiologic science moved into the OMICS era a decade agogenomics starting the process, followed by proteinomics, epigenomics, metabolomics, etc. Perhaps unfortunately, genetic technology has evolved more rapidly than epidemiologic methods to utilize these tools and data. This conundrum arose as a result of rapidly decreasing analytic costs combined with increasing information technology made these techniques more readily available in epidemiologic studies, albeit sometimes without the requisite epidemiologic analytic forethought. Like all other studies, epidemiologic research using these techniques still needs to be assessed for study design, sample size, analytic approach and control for confounding and bias. So, in a sense, epidemiology has moved as a discipline from general associations between exposure and disease, to highly specific ones. The latter molecular based studies not only allow us to refine hypotheses and identify susceptible populations or individuals, mechanisms, and pathways, but also are leading to highly targeted therapies, as developed countries move towards “personalized medicine.” Epidemiology is thereby playing a crucial role in emerging translating genomic science to improved patient care.

The move to more exact epidemiologic science may have other implications, however. Epidemiologic studies have been increasingly used as part of a body of evidence to address issues of causality, and may be used to inform a policy recommendation, (as might occur in an Institute Of Medicine report requested by a regulatory body), or to recommend an acceptable level of exposure (e.g., in IARC classifications of carcinogens). While there is considerable debate and discussion about how epidemiology informs the leap from association to causation, emerging molecular tools are closing this gap. However, increasingly epidemiologists and epidemiologic studies are used in an attempt to predict what will happen to a particular individual. This is especially true in legal cases, where results of population studies are applied to a single individual and asked to predict or verify whether a specific potential exposure is linked with a single outcome. Examples include exposure to pharmaceuticals resulting in an adverse affect, exposures to occupational hazards like asbestos, diesel, or exposures to disasters such as 9/11 first responders.

Tort cases often require that there is the general probability of causation between an exposure and a disease, or more specifically for criminal cases, that the evidence for causality is “beyond a reasonable doubt”. In vivo and in vitro studies may provide additional support for causality, but are less likely to be relevant than human studies. Randomized trials are often unethical or impractical. Therefore, the question becomes “do findings from epidemiologic studies, which were conducted among diverse populations, have applicability to this particular individual?”

With the advent of OMICS frequently utilized in epidemiologic studies, it should be possible to better answer this question. There will still be usual issues to debate about the differences in study populations, times, places, determinations of exposure, data analyses, confounding and biases, but if the genetic characteristics of the individual closely match those in the epidemiologic/omics study, the likelihood of inferring a population association to causation may increase. As epidemiologists, we may not be at that stage yet, but the application of epidemiologic/omics studies to individuals should provide fascinating new opportunities for translation.