Bioethics Forum Essay
A Decade’s Worth of Gene-Environment Interaction Studies, in Hindsight
In the early 2000s, Avshalom Caspi, Terrie Moffitt, and their colleagues published two papers (here and here), which suggested that we could finally begin to tell rather simple but evidence-based stories about how genetic and environmental variables interact to influence the emergence of complex phenotypes. It’s hard to exaggerate the level of interest those papers generated. According to Google Scholar, they’ve now been cited more than 7,000 times. To put that in perspective, Watson and Crick’s paper on the structure of DNA has been cited only about 9,500 times.
In 2011, however, Laramie Duncan and Matthew Keller published a paper in the American Journal of Psychiatry, which told psychiatrists about the results of their meta-analysis of the 103 gene-environment interaction (GxE) studies that had been done in the first decade of this millennium. The news wasn’t good. According to Duncan and Keller, the results of their meta-analysis “were consistent with the existence of publication bias, low statistical power, and a high false discovery rate.”
This month, Laramie Duncan published a piece with Alisha Pollastri and Jordan Smoller in the American Psychologist that aims to inform psychologists about the results that Duncan and Matthew broke to psychiatrists in 2011. The new piece is called “Mind the Gap: Why Many Geneticists and Psychological Scientists Have Discrepant Views about Gene-Environment Interaction Research.” A franker, tad-longer title would have been, “Psychologists (and Everybody Else) Need to Get Up To Speed on What Psychiatric Geneticists Already Know: Gene-Environment Interaction Research Hasn’t Panned Out Yet.” Or as the new piece puts it, “the first decade of cGxE research has produced few, if any, reliable results.”
The “c” before the GxE in the previous sentence is intended to emphasize that the studies in the meta-analysis all purported to investigate how single “candidate” genes interacted with some environmental variable. The first of the two Caspi-Moffitt studies mentioned above was published in 2002, and it investigated the idea that one candidate gene (MAOA), involved in the metabolism of serotonin, interacted with childhood maltreatment to increase the likelihood of adult antisocial behavior. The second study, published in 2003, investigated the idea that another candidate gene (5-HTT), this one involved with the transport of serotonin, a neurotransmitter, interacted with stressful life events to significantly increase the likelihood of depression. As Duncan et al. point out in their new piece in the American Psychologist, at the time genes associated with neurotransmitter systems seemed like logical candidates for investigation. After all, the leading pharmacological treatments of the day targeted those systems. With the benefit of hindsight, it’s easier to be skeptical about the idea that, because psychiatric treatments of the day targeted those neurotransmitter systems, it made sense to investigate those same systems to understand the etiology of the disorders they were trying to treat.
Were there any other reasons to be skeptical, earlier, about the grounds for picking those candidate genes in particular? Duncan et al. do give another: when the studies investigating the interactions between candidate genes and environmental variables were getting off the ground in the early 2000s, it was already clear that 95 percent of the results regarding candidate genes were false positives. That is, Duncan et al. invite the reader to wonder: in the early 2000s, when those results regarding single candidate genes were failing to withstand closer scrutiny, wasn’t there already reason to be skeptical about the veracity of findings from studies that investigated the interaction between those genes and environmental variables?
Duncan et al. do not say that no true results emerged from the cGxE studies of the 2000s. Nor do they rule out the possibility that such candidate gene-based interaction studies will, in the future, when conducted with sufficiently large numbers of subjects and sufficiently stringent statistical safeguards, yield meaningful results. They do, however, seem rather skeptical about the probability of such results emerging. Rather, they seem to be betting on a couple of other strategies. One newer form of cGxE study begins with loci that already have been conclusively demonstrated by Genome-Wide Association Studies (GWAS) to have significant effects. The second strategy entails “Genome-Wide Interaction Studies”—i.e., GWAS with an environmental variable for which researchers test the gene-environment interaction. Such strategies may not yield stories as easy to tell as those produced by cGxE studies in the early 2000s, but they may help us get closer to the truth that everybody pursuing any of these studies is after.
Erik Parens is a senior research scholar at The Hastings Center and a core faculty member of the Center for Research on Ethical, Legal and Social Implications of Psychiatric, Neurologic and Behavioral Genetics at Columbia University Medical Center. A version of this commentary originally appeared on the center’s website.
Posted by Susan Gilbert at 04/23/2014 12:50:41 PM |