By Kevin E. Noonan --
Early last week, the Journal of the American Medical Association published a study ("Interaction Between the Serotonin Transporter Gene (5-HTTLPR), Stressful Life Events, and Risk of Depression") by Neil Risch at the University of California, San Francisco and colleagues at the National Institutes of Mental Health, Johns Hopkins University, Yale University School of Medicine, the Rockefeller University, the University of Pittsburgh School of Medicine, and Virginia Commonwealth University that showed no statistically-significant link between depression and the short HTT allele. The study provided a statistical meta-analysis of data accumulated for 14,250 participants from 14 separate studies from the medical literature and a second meta-analysis of 10,943 participants from 10 of these studies. These results were contrary to the results published by Avshalom Caspi and collaborators (including Terrie E. Moffitt (at right), Knut Schmidt Nielsen Professor, Departments of Psychology and Neuroscience, Psychiatry and Behavioral Sciences, and Institute for Genome Sciences and Policy, Duke University) that individuals bearing at least one copy of a specific genetic polymorphism in a serotonin transporter gene were more susceptible to depression as a consequence of environmental (life) stressors (see Caspi et al., 2003, "Influence of Life Stress on Depression: Moderation by a Polymorphism in the 5-HTT Gene," Science 301: 386-89).
Dr. Moffitt has graciously provided us with three points in rebuttal to the conclusions of the Risch study, set forth in her own words:
1. The JAMA article ignores the wider body of scientific evidence. In the past 6 years, extensive research in experimental neuroscience using both animals and humans has validated the original report by showing that 5-HTTLPR short allele-carriers are excessively vulnerable to stress. Experimental studies that expose human participants to stressors in the laboratory show that individuals having the 5-HTTLPR short genotype have greater stress responses on measures of cognitive reactivity (e.g., Beevers et al., 2007, J. Abnorm. Psychol. 116: 208-12; Fox et al., 2009, Proc. Biol. Sci. 276: 1747-51), hormonal reactivity (e.g., Chen et al., 2009, Psychoneuroendocrinology 34: 681-86; Gotlib et al., 2008, Biol. Psychiatry. 63: 847-51), physiological reactivity (e.g., Lonsdorf et al., 2009, Psychol. Sci. 20: 198-206) and reactivity in the brain's emotion-circuitry (e.g., Hariri et al., 2006, Arch. Gen. Psych. 62: 146-52; Canli et al. 2006, Proc. Natl. Acad. Sci. U.S.A. 103: 16033-38). This vulnerability of 5-HTTLPR S carriers has also been confirmed in animal research (e.g., Carola et al., 2008, Biol. Psychiatry. 63: 840-46; Barr et al., 2004, Biol. Psych. 55: 733-38; Kalin et al., 2008, Mol. Psychiatry. 13: 1021-27, online at DOI: 10.1038/mp.2008.37; Watson et al., 2009, PLoS One, published online January 14, 2009). Further validation comes from studies which have shown that 5-HTTLPR Short allele carriers are vulnerable not only to depression, but also to other mental-health problems caused by stress, including PTSD and anxiety (e.g., Gunthert et al., 2007, Psychosom. Med. 69: 762-68; Kilpatrick et al., 2007, Am. J. Psychiatry. 164: 1693-99).
2. The selection of studies for meta-analysis clearly fails to represent the pool of papers in the literature. 4 out of 17 positive replications were included (24%) but 6 out of 9 published negative
studies (67%), a bias that violates a basic requirement of the meta-analysis method to examine a representative sample of the existing literature. Further, the article opted not to analyze several well-designed studies of individuals exposed to stress. For example, studies of children who are victims of abuse (e.g., Cicchetti et al., 2007, Dev. Psychopathol. 19: 1161-80; Kaufman et al., 2004, Proc. Natl. Acad. Sci. U.S.A. 101: 17316-21) and patients suffering hip fractures, strokes and coronary heart disease (Kohen et al., 2008, Arch. Gen. Psychiatry. 65: 1296-302; Lenze et al., 2008, Amer. J. Geriatric Psych. 13: 428-32; Otte et al., 2007, Am. J. Psychiatry. 164: 1379-84) have reported positive findings for this "GxE" hypothesis. Several other positive replications from studies with very strong research designs were omitted from the meta-analysis as well (Kendler et al., 2005, Arch. Gen. Psych. 62: 529-35; Drachmann-Bukh et al., 2009, J. Affect. Disord. 2009 Mar. 30 (e-publ. ahead of print at doi:10.1016/j.jad.2009.02.023); Lazary et al., 2008, Biol. Psychiatry. 64: 498-504).
3. Within the smaller subset of studies in the meta-analysis, there is important heterogeneity. The article says this heterogeneity was not statistically significant. However, we have learned that the test fell just short of statistical significance. One obvious characteristic that varies widely across studies in the meta-analysis is measurement quality. As one example, meta-analysis gives more mathematical weight to studies with the largest samples. But in this case the big studies had to collect their data through telephone calls or self-complete postal questionnaires, which are known to be weak methods of measuring life events and depression. Not surprisingly, these big studies with weak measures tended not to find positive results, tilting the meta-analysis toward a null finding. In the existing literature of tests of this GxE hypothesis, 13 studies have replicated the finding, using measures collected in face-to-face clinical assessments, and 10 studies failed to replicate the finding, using measures collected via phone or postal questionnaires. This clear methodological pattern was ignored in the meta-analysis. This is one example of heterogeneity could have been followed up in a meta-analysis, there may be others.
Dr. Moffitt concluded her response to Dr. Risch's work by noting that "[w]hat is needed is not less research into gene-environment interaction, as Risch et al. recommend, but more research of better quality, and a more thorough and thoughtful evaluation of it."
Dr. Moffitt has extensive experience in studying the biological basis of complex behaviors, and in addition to her Duke University appointment is also Professor of Social Behaviour and Development, Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, London and Associate Director, Dunedin Multidisciplinary Health and Development Research Unit, Dunedin School of Medicine, Dunedin, New Zealand.
This controversy illustrates one of the many pitfalls frequently faced by academic inventors motivated by the dual (and occasionally conflicting) needs of disseminating the results of their studies as rapidly as possible and protecting the intellectual property aspects of their work to maximize the likelihood that it can be commercialized (and not merely expropriated by others based on its disclosure in the academic literature). As a result, often the research is not as developed as it might be if it were developed by a commercial concern initially. This situation presents particular challenges to university technology transfer offices and their clients, university faculty and university administrations. It is an imperfect system, which motivates some to say there is a flaw in the basic premise of promoting university patenting under the Bayh-Dole regime. However, the alternative (not protecting university-generated inventions) is indeed worse, since then the public stands no chance of seeing the fruits of these labors helping to finance the next generation of technologies.