Special Editors and A Blog: Editorial
Fallin, M. Daniele
As a special editor for genetics articles in Epidemiology, I hope to promote work that illustrates the importance of epidemiologic rigor in genetic epidemiology applications. I encourage articles that demonstrate advances in the scientific approach to genetic epidemiology—through new methodology, innovative applications, or substantive results that employ rigorous epidemiologic methods and can inform the broad epidemiology audience and guide subsequent studies. What follows are some specific topic areas that I would welcome.
The Use of Causal Inference Methods in Genetic Applications
The discussions of causal inference currently in progress among epidemiologists and biostatisticians have received only limited attention in the field of genetic epidemiology. Genetic epidemiologic studies could be greatly enhanced by embracing this field of thought. In addition to confounding by ancestry, there are issues of understanding and accounting for selection bias, teasing out direct and indirect effects of genes and other factors, implications of the nongenetic context of genetic effects, and the use of genes as instrumental variables. These are important and useful applications of causal-inference methods in genetic epidemiology.
The Emphasis on Hypothesis-testing in Genetic Studies
While there has been extensive and heated discussion in epidemiology and biostatistics on the utility and limitations of a testing framework for making inferences, this debate has not occurred to the same depth in genetic epidemiology. Some of the issues in genetic epidemiology are new, but many have already been addressed. Salient approaches to this issue that can help to move the field forward are very welcome. A related topic is the more general discussion of Bayesian versus frequentist approaches. This again has been discussed in detail in the epidemiology world, but is less familiar to most of genetic epidemiology. The depth of discussion and application of alternative approaches could be greatly improved.
Estimation and Utility of Gene-environment Interactions
A lively debate is developing over whether gene-environment interactions are estimable with current study designs and sample sizes, and whether those that can be detected actually add substantive value beyond the marginal genetic or environmental findings. Proponents of gene-environment investigations highlight at least 2 advantages. One, synergistic effects will be missed by marginal analyses that assume no interaction. Two, gene or environmental risk-factor identification or mechanism characterization can be enhanced when considering (or restricting) the values of the other factor. Skeptics counter that strong synergistic effects can be detected in marginal analyses, and that across the spectrum of possible gene-environment interaction scenarios, only a small fraction would not be detectible by analyses that did not account for interactions. Furthermore, power to detect this small fraction of scenarios would require samples sizes and precision likely beyond the scope of most current designs. Methodological work as well as substantive gene-environment findings that add to this debate would be welcome. For substantive submissions, simply reporting both main and joint effects in a manner that would allow the reader to assess interaction on either the additive or multiplicative scale would be an important advance for the field.
Translation of Genetic Epidemiologic Findings
Much of the suggested “translation” of genetic epidemiology involves gene-based risk prediction, confirmation of diagnosis, or prediction of treatment response or adverse event. As genetic epidemiology moves from initial gene discovery to risk characterization, methods for development of screening, prediction modeling, and pharmacogenetics will be paramount. Clear discussion and applications of the design and analytic approaches necessary for these goals (in contrast to gene discovery) is encouraged.
The Use of Missing-data Methods for Genetic Studies
Imputation of missing genetic data is now commonplace, yet many design and analytic tools for dealing with missing data in epidemiology have not been extensively considered for genetic data. Articles that address particular methodological concerns or improve upon current imputation methods are encouraged.
Reviews and Meta-analyses
Finally, reviews and meta-analyses regarding a particular gene or disease, such as those resulting from the HuGE effort (www.cdc.gov/genomics/hugenet/reviews/index.htm), are very useful. Reviews that go beyond listing results to providing insight about weaknesses or strengths in the evidence are extremely important and can guide future directions.
I take on this new role with great pleasure and look forward to exciting genetics articles in our pages in the coming months and years.
© 2010 Lippincott Williams & Wilkins, Inc.