This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation

The Inherent Complexities of Gene–Environment Interactions

Child Study Center and PACE Center, Yale University, New Haven, Connecticut; and Department of Psychology, Moscow State University, Russia.

Address correspondence to Dr. Elena L. Grigorenko, Yale University, PACE Center, 340 Edwards St., PO Box 208358, New Haven, CT 06520-8358. E-mail: elena.grigorenko{at}yale.edu

	Abstract

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 
The article outlines the complexities of gene–environment interactions in the determination of human disease, especially as they relate to aging, and stresses the importance of continuing such studies, in spite of their inherent difficulties. First, a capsule review of the literature pertaining to studies of gene–environment interactions is presented, and designs and methodologies used to detect these interactions are briefly discussed. Second, research questions and problems that can be addressed as outcomes of gene–environment interaction studies are exemplified. Third, a number of illustrative examples of gene–environment interactions are presented. Fourth, various types of gene–environment interactions are briefly discussed. Fifth, concluding remarks are offered, and possibilities of studying gene–environment interaction within social and biological research on aging are outlined.

"The majority of potters do not die of bronchitis. It is quite possible that if we really understood the causation of this disease, we should find that only a fraction of potters are of a constitution that rendered them liable to it."

(Haldane, 1938, p. 102)

WITH this thought, Haldane (1938) captured the essence of what is now referred to as "gene–environment interaction." When stating that only a percentage of all potters exposed to occupational dust develop bronchitis, Haldane referred to instances in which the combined effect of genetic and environmental risks or protective factors is substantially greater than that anticipated from a simple sum of their independent effects.

Today, there are several distinct branches of medical science that address the issue of gene–environment interaction (Shostak, 2003). Specifically, this interaction is central to the fields of ecological genetics (Puga, Micka, Chang, Liang, & Nebert, 1996), genetic epidemiology (Khoury, Beaty, & Cohen, 1993), molecular epidemiology (Perera & Weinstein, 1999), and toxicogenomics (Iannaccone, 2001). Although each of these fields of study is characterized by unique research paradigms and methodologies, collectively they define a set of assumptions central to current conceptualizations of gene–environment interaction. Specifically, researchers within these fields (a) assume an important role of both genetic and environmental factors in the causality of various human diseases but (b) stress the fundamental role of the interaction between genes and environment in the diseases' manifestations and (c) stress the value of understanding the types of adaptation to a variety of environments that result in disease or disease-free living (for a review, see Shostak, 2003). Thus, collectively, these fields attempt to enhance our understanding of public health risks posed both by individual and by population genetic susceptibility to disease in the context of harmful and protective environments.

The emergence of these fields is at least in part attributable to the popular premise in the scientific community that most complex human behaviors, whether disorders or not, evolve through the interaction of genetic and environmental factors (Khoury, Burke, & Thomson, 2000). These fields call jointly for a careful consideration of history of exposure to adverse and protective environments while investigating genetic susceptibilities.

The overarching objective of these efforts is to create disease prevention technologies (Shostak, 2003, p. 2330), where combined genetic screening and environmental (e.g., dietary) intervention programs avert or minimize adverse health outcomes. For example, it has been estimated that by preventing smoking, approximately 50% of bladder cancer in males and 25% in females could be avoided (International Agency for Research on Cancer, 1986). Similarly, there is extensive evidence generated by epidemiological studies that implicates solar ultraviolet radiation as the major environmental risk factor for melanoma, and there are explicit public health warnings about how to avoid harmful exposure (e.g., Armstrong, Kricker, & English, 1997; Scarlett, 2003).

Correspondingly, this strong public health focus on gene–environment interactions and their impacts and consequences is intended to differentiate levels of risk within a given population by stratifying this population and, ultimately, to better understand the causal mechanisms underlying human diseases (Kelada, Eaton, Wang, Rothman, & Khoury, 2003). The extreme interpretation of gene–environment interaction holds that "all human disease is the result of interactions between genetic variation and the environment (broadly defined to include dietary, infection, chemical, physical, and social factors)" (Khoury et al., 1993, p. 7). This thought can be taken even further by substituting the phrase "all human disease" with "all human complex behavioral traits and conditions" and by stating that an understanding of the causal networks underlying these complex traits is conditioned upon the understanding of the contribution of gene–environment interactions to the expression of these traits. Clearly, aging exemplifies one such "complex human condition," and, when broadly defined, it can be viewed as an outcome of corresponding aging and longevity genetic mechanisms interacting with various environments in which people age. This idea is captured by the phrase "So-and-So aged well, whereas So-and-So did not"; there is an assumption of the inevitability of aging (i.e., the realization of genetic programming) and also of its modifiability (i.e., its susceptibility to environmental impacts).

However, in making such a broad statement regarding the causal role of gene–environment interaction, one can be over-inclusive in interpreting all effects as interactive effects and thus overgeneralize. Here, the only purpose of such a broad statement is in a theoretical exploration of the utility of accounting for gene–environment interaction to improve the research's potential to detect and map genetic effects while studying human diseases and complex traits. It is of note that the underlying assumption of this discussion is that gene–environment interaction is central to understanding the causes (etiology) of complex human diseases and traits. This discussion does not extend to the debate over the issue of rarity versus commonality of gene–environment interactions; the framing postulation here is that gene–environment interaction is important and common.

However, even with this assumption in mind, there is still a question of the utilitarian value of taking gene–environment interaction into account while studying the etiology of human conditions. To arrive at a statement regarding the serviceable value of including gene–environment interaction term(s) in statistical models with an aim of improving the likelihood of understanding how diseases develop and what causes and can prevent them, a number of relevant issues should be discussed. First, main interpretations of gene–environment interaction are presented in a sketch, and designs and methodologies used to detect the interaction are briefly discussed. Second, problems that can be addressed as outcomes of gene–environment interaction studies are described. Third, a number of illustrative examples of gene–environment interactions are presented. Fourth, various types of gene–environment interactions are discussed. Fifth, and finally, concluding remarks are offered, and possibilities of studying gene–environment interaction within social and biological research on aging are briefly discussed. Overall, the main objective of this review is to stress the importance of, despite its inherent difficulty, studying gene–environment interactions in the context of understanding human diseases, especially those manifested in older age at higher prevalence, because these diseases typically occur as a result of exposure of predisposing genotypes to detrimental environments.

	DEFINITIONS, ASSUMPTIONS, AND METHODS: A CAPSULE REVIEW OF LITERATURE ON GENE–ENVIRONMENT INTERACTION

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 
The tradition of distinguishing genes and environments (with further subdivision of the latter into shared and nonshared [Jinks & Fulker, 1970]) has been instrumental in psychology for over three decades now. This distinction has provided a number of useful insights regarding not only main, but also interactive, effects of these components (Eaves, Last, Martin, & Jinks, 1977). Although never conceived as such, gene–environment interaction is sometimes interpreted with an element of simplification, so that genes are equated to the causal factors operating within the body and environments to causal factors operating outside the body. To avoid such simplification, key issues regarding the concept of gene–environment interaction are discussed here briefly.

The term "gene–environment interaction" has at least four meanings depending on the context in which it is used. In the context of studying evolution, the presence of gene–environment interaction attests to variance in fitness. Evolutionarily, the presence of significant gene–environment interactions indicates the presence of some "uncertainty," an inability to accurately predict the behavior (i.e., fitness) of a particular combination of genes (referred to as a genotype) in a given environment from the main effects of genotype and environment. Multiple types of gene–environment interactions have been investigated in this context. As most of this work is done with plants, worms, flies, and animals, the genotypes are typically known, and the central idea is to rank-order the genotypes with respect to the trait of interest in a variety of different environments. Clearly, this ordering of genotypes and environments can produce a number of different outcomes, but the following situations are of greatest interest (Remold & Lenski, 2001): (a) A single genotype is advantageous across all environments and thus demonstrates the greatest fitness; (b) no single genotype is advantageous across all environments; (c) genotypes can be rank-ordered and their ranks preserved across various environments, although the extent of variation in phenotypes associated with each genotype is much greater in one environment than another; and (d) both rank orders of genotypes' fitness and the extent of phenotypic variation vary across environments. A great deal of work has been carried out in hopes of understanding the architecture of gene–environment interaction with regard to variance in fitness (e.g., Li et al., 2003; Yadav et al., 2003), and a review of this substantial literature is outside the scope of this article. However, of special interest here is one of many remarkable observations generated in the literature, specifically that a strong gene–environment interaction can arise randomly. For example, in Escherichia coli (a common bacterium intensively studied by geneticists because of its limited number of genes and ease of growth and maintenance in laboratory settings), a single new random mutation (i.e., a mutation that had not experienced any pressure of biological selection forces in any of the environments tested in the laboratory) is sufficient to produce a strong gene–environment interaction. Moreover, contributions of new random mutations to the presence and magnitude of gene–environment interaction can vary for different environmental factors (Remold & Lenski, 2001). These observations have a number of implications, one of which is that gene–environment interaction is an important mechanism underlying phenotypic plasticity, an ability to adapt to new environmental conditions, and that "phenotypic plasticity must often depend on many different genes rather than a handful of ‘plasticity’ genes" (Remold & Lenski, 2001, p. 11392).

In the context of studying patterns of similarities between relatives, gene–environment interaction typically refers to "the genetic control of sensitivity to the environment" (Eaves & Erkanli, 2003, p. 287) or reactions to specific environments as a function of the underlying genetic make-up. It is important to note here that only some types of gene–environment interactions can be described this way (Falconer & Mackay, 1996). Acknowledging the variety of ways in which gene–environment interaction has been interpreted in the literature, Plomin and colleagues (Plomin, DeFries, McClearn, & Rutter, 1997) make reference to a number of adoption studies illustrating the role of gene–environment interactions in the development of maladaptive behaviors (e.g., conduct problems) in adoptees who had biological parents with criminal histories, antisocial personality diagnoses, and drug abuse who were adopted into families with legal or mental health problems (Bohman, 1996; Brennan, Mednick, & Jacobsen, 1996; Cadoret, Yates, Troughton, Wodworth, & Stewart, 1995).

A number of examples below feature this type of gene–environment interaction, defined as genetic sensitivity (also referred to as "reaction norm") or susceptibility. Multiple types of research designs are available to detect and assess the magnitude of this type of gene–environment interaction. Here, in addition to the adoption design mentioned above, three more approaches are presented briefly. The first and most user-friendly design is that of a case–control study: Both participants with and without the phenotype (disease or a complex condition or trait) of interest are recruited. For this design, it is very important that the control participants are selected from the same source population as the target participants to avoid both environmental and genetic biases attributable, for example, to different frequencies of certain environments (e.g., diets) and genetic polymorphisms (indicators of genetic variation between people) in different populations. This type of study typically calls for a large number of cases and controls to obtain sufficient power for the detection of the effect of the interaction. To minimize this need for large sample sizes, a case-only design was developed. Multiple versions of the case-only design exist. In the framework of the singleton case-only design, cases with a high-risk genotype are compared with cases without this genotype with respect to prevalence of environmental exposure. In the framework of the family-based case-only design, alleles observed in both affected individuals (cases) and their relatives (parents or siblings, including twins) are investigated. It is assumed that high-risk alleles are observed more often in cases than would be expected by chance based on the parental profile of alleles. The presence of gene–environment effects is acknowledged if, when stratified by exposure status, the relative risk is substantially increased for cases with high-risk alleles.

The context in which gene–environment interaction is still primarily investigated is in laboratory animal studies. In such studies, variation in genotype and environment can be both controlled and manipulated. These studies have provided a number of novel insights into the dynamics of in vivo gene–environment interactions. Specifically, the insights ranged from detecting powerful gene–environment effects in the experiment by Cooper and Zubeck (1958) on the effects of quality of rearing environment (enriched versus impoverished) on two distinct strains of rats bred for their learning ability to deal with mazes ("bright" versus "dull") to realizations of difficulties related to replicating genetic effects due to subtle environment differences across different laboratories (Crabbe, Wahlsten, & Dudek, 1999). Commenting on caveats of studying gene–environment interactions in animals, Wahlsten and colleagues (2003) wrote: "Animal models retain an advantage over human populations in the study of genotype by environment interaction because genotype can be replicated and held constant while manipulating environment. Research on humans is severely limited in its ability to accomplish this .... Fortunately, as the identity of specific genes affecting specific behavioral characteristics or disease traits in human becomes known, this situation should improve" (p. 309). And it has, at least to a certain degree, as illustrated below.

Finally, in its narrowest statistical interpretation, gene–environment interaction indicates "any nonadditive effect of an organism's genotype and its environment on the expression of a trait of interest" (Remold & Lenski, 2001, p. 11388). In other words, when gene–environment interaction is present, the link between the environment and a phenotype is moderated by the genotype, and genotype–phenotype relationships are moderated by the environment (Carey, 2002). Although narrowest in its absolute sense, this interpretation of gene–environment interaction is present in all statistical methodologies aimed at detecting the interaction. For example, in studies of animals, the preferred approach is the use of multifactorial analysis of variance-based strategies (e.g., Wahlsten et al., 2003). In studies of humans, the preferred method is to subdivide a sample of unrelated or related individuals into subgroups based on the value of environmental exposure so that a test of homo- or heterogeneity among genetic parameters within a subgroup can be performed. Typically, these analyses are based on the assumptions that environmental measures are fixed (not random) and independent of genetic effects on the outcome of interest.

In the framework of the case–control study of gene–environment interaction, two methods of the interaction evaluation, multiplicative and additive, are typically used (Mucci, Wedren, Tamimi, Trichopoulos, & Adami, 2001). Under the multiplicative model, the presence of gene–environment interaction is recognized when the observed relative risk for a subgroup of a target sample, characterized by both the high-risk genetic variant and the high-risk environment variable, is greater than the product of the two individual relative risks. Under the additive model, the observed relative risk is expected to be greater than the sum of the relative risks (different modifications of this general approach have been proposed [e.g., Semenza, Ziogas, Largent, Peel, & Anton-Culver, 2001]). In the framework of the case-only design applied to a study of unrelated individuals, only the multiplicative method is used, and the observed relative risk is interpreted as a direct measure of gene–environment interaction; this design does not permit estimation of individual effects of high-risk environments or genotypes (Mucci et al., 2001). In the family framework of the case-only design, multiple approaches have been proposed, among which are the transmission-based approach (e.g., Selinger-Leneman, Genin, Norris, & Khlat, 2003), the log-linear modeling approach (e.g., Selinger-Leneman et al., 2003), logistic regression (e.g., Eaves & Sullivan, 2001; Liang & Beaty, 2000; Marcus et al., 2000), various applications of the variance component framework (e.g., Neale & Cardon, 1992; Purcell & Sham, 2002), generalized estimating equation modeling (Liang & Zeger, 1993), the extended Mantel–Haenszel method (Liang, 1987), and Markov Chain Monte Carlo (e.g., Corander & Sillanpää, 2002; Eaves & Erkanli, 2003).

Given the developed definitions, research methods, and statistical techniques, what kinds of research issues are typically addressed when gene–environment interaction is considered?

	GENE–ENVIRONMENT INTERACTION: TYPES OF RESEARCH QUESTIONS

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 
The importance of accounting for gene–environment interactions is becoming more evident along with the progress in understanding the genetic mechanisms underlying complex human disorders (Gunzerath & Goldman, 2003). The role of these interactions is particularly important in studying health issues in aging, where genetic predispositions for various disorders might be amplified by long-term exposure to detrimental environments (e.g., occupational hazards) or, on the contrary, be diminished by exposures to preventive environments (e.g., diet).

Typically, in studies of gene–environment interactions, either the genes or the environments are held constant, whereas, correspondingly, environments or genes are varied or manipulated. This is easy to accomplish in a laboratory setting, and a number of animal studies have generated informative data regarding the role of gene–environment interaction in aging (e.g., Reich et al., 2001; Speakman et al., 2004). However, the application of this framework to studies of humans is more difficult because the exertion of meaningful experimental control over genes or environments in humans is impossible.

Yet, a number of pseudo-experimental designs developed within this general paradigm have been useful, including case–control studies, various forms of family studies, studies of population isolates, and migrant studies (Gunzerath & Goldman, 2003). The more the field learns about gene–environment interaction, the easier it is to formulate specific questions that can be used to enhance current knowledge.

Historically, much of the momentum for studies of gene–environment interaction has arisen from pharmacogenetics, a field of study aimed at the discovery of genetic variants that affect drug efficacy with a goal of devising clinical treatments that are both safe and effective. Ultimately, the idea is to individualize clinical treatment by finding the most beneficial drug with the least number of side effects, given the genetic background of the patient. A substantial number of studies linking specific genetic polymorphisms and responses to specific drugs are present in the literature (e.g., Murphy, Kremer, Rodrigues, & Schatzberg, 2003). Given the topic's remote connection to the main body of this article, however, it will not be discussed in detail.

The recent progress in molecular genetics has led to the identification of a number of relevant genes for a variety of complex disorders. This progress has led to the incorporation of genetic polymorphisms, now known as risk factors, into epidemiologic research to detect and quantify gene–environment interactions for known types of risk factors. These investigations are typically aimed at understanding both risk-enhancing and risk-diminishing interactions. To illustrate, consider the following example. Approximately 5,000 people in the United States are diagnosed with the inherited genetic disorder ₁-antitrypsin (AAT) deficiency. AAT is a protein, produced mainly in the liver, that functions to balance out in the lung a potentially detrimental impact of the enzyme neutrophil elastase (an enzyme involved in breaking down complex proteins into their simple compounds and in the elimination of damaged or aging cells and microbes) (Sharp, de Serres, Newman, Sandhaus, Walsh, & Hood, 2003). AAT deficiency and a corresponding disruption of the AAT–neutrophil elastase balance result in the increase of digestive function of the elastase and lead to hyperresponsiveness of the lung, which results in tissue damage and weakening of lung function (Malerba, Radaeli, Ceriani, Tantucci, & Grassi, 2003). Correspondingly, AAT deficiency is associated with such conditions as emphysema, asthma, and chronic obstructive pulmonary disease; in addition, many patients experience liver problems, including hepatitis, cirrhosis, and liver failure (Sharp et al., 2003). The AAT gene is located on chromosome 14p and is characterized by, among many silent variants (i.e., genetic polymorphisms that do not result in the change of the protein function), two common functional variants (referred to as S and Z alleles), which result in severe deficiency of plasma levels of the AAT protein. Typically, the carriers of these genetic variants experience pulmonary diseases. However, some individuals carrying these variants and diagnosed with AAT protein deficiency do not demonstrate health problems and are asymptomatic. In other words, some other factors, either environmental, genetic, or interactive, protect these individuals from the development of detrimental diseases even though they are genetically predisposed toward them. Thus, although the clinical pathology, genetic mechanisms, and biochemical pathways of AAT deficiency are well characterized, the environmental factors and various gene–environment and gene–gene interactions impacting the onset, manifestation, and course of the disease are not understood. The challenge here is to decipher the significance of general environmental exposure associated with protective (i.e., for asymptomatic patients with AAT deficiency) as well as detrimental (i.e., for those patients who develop secondary clinical symptoms such as liver problems) effects of gene–environment interactions.

Another area where a design inclusive of gene–environment interaction might be useful is in investigating genetic variants that only predispose for a disorder rather than control its manifestation but that are abundant in the general population. In this context, one oft-cited example is that of the APOE gene, which increases risk of Alzheimer's disease by only about two times; however, its frequency in the population and its assumed co-action with various environmental agents bring attention to this gene as a population risk factor for Alzheimer's disease.

Similarly, insulin resistance/type 2 diabetes is a serious public health problem that is expected to affect approximately 250 million people worldwide by 2020. At this point, there is convincing evidence that body composition, habitual physical activity, and diet are substantial factors in this disorder. It is also now widely accepted that genetic variation affects the risk for developing insulin resistance and diabetes. Correspondingly, an intensive effort to identify insulin resistance genes is underway. However, no single gene controlling the disorder has been identified. It is much more likely that insulin resistance and diabetes are associated with co-action of multiple genes, and a number of candidate genes have been proposed. One popular candidate is the small intestine fatty acid binding protein (FABP2) gene, a gene belonging to a family of more than 20 genes responsible for the production of tissue-specific proteins (Bernlohr, Simpson, Hertzel, & Banaszak, 1997). The FABP2 gene is considered to be a candidate for insulin resistance because of its involvement in fatty acid absorption, a deficiency that could affect insulin resistance/type 2 diabetes (Weiss, Brown, Shuldiner, & Hagberg, 2002). The FABP2 gene has a number of variants, one of which (referred to as Ala54Thr, meaning that amino acid 54 has been changed from an alanine to a threonine) has been associated with insulin resistance in some but not all relevant published studies. The demarcation line between supportive and unsupportive studies is of interest to the current discussion because, as stated by Weiss and colleagues (2002) in their review of the FABP2-based association studies, the studies in which the samples were stratified by such factors as habitual physical activity and diet, have revealed the association between Ala54Thr FABP2 genotype and insulin resistance, whereas the studies that did not take these relevant factors into account did not produce such an association. To illustrate, the heavily marine-based diet of the Canadian Inuit was found to change the rank order of the genotype homozygous (i.e., having two identical copies) for the risk allele so that instead of being the anticipated risk factor, it was shown to be the protective factor, at least for the percentage of screened individuals (Hegele, Young, & Connelly, 1997). Thus, the understanding of types and magnitudes of gene–environment interactions (here, the interaction between the genotype and the diet) for such complex diseases as insulin-dependent/type 2 diabetes appears to be extremely important.

Finally, researchers have recently begun an investigation into gene–environment interaction using a reversed direction paradigm, that is, investigating mutations in a gene and trying to understand which environmental agents could have caused such mutations. An example of this methodology's application is the differentiation of exogenous (i.e., derived externally) and endogenous (i.e., derived internally) mutations in the p53 tumor suppressor gene, the mutations of which are associated with a number of cancers (Hussain & Harris, 1998). This work is driven by the hypothesis that "chemicals and physical carcinogens leave footprints of their activities because of the base changes they induce. ... knowledge of the pattern mutations found in genes commonly mutated in human cancer, such as the p53 tumor suppressor gene, allows for predictions to be made on the likelihood of an exogenous DNA-damaging agent being involved" (Jones, Buckley, Henderson, Ross, & Pike, 1991, as cited in Shostak, 2003, p. 2333).

Having briefly reviewed definitions, research designs, methodologies, and types of questions relevant to studies of gene–environment interactions, I now present a brief and selective set of illustrations of the successes in identifying and quantifying these interactions. Of note here is that, given the relatively small literature on established manifestations of gene–environment interactions, its subset investigating the role of these interactions in aging is even smaller. Thus, this review points out illustrations of such interactions in a broadly defined field of human disease across the life span whose prevalence increases with age (e.g., cancer, jeopardized immune responses, and lack of physical fitness).

	GENE–ENVIRONMENT INTERACTION: ILLUSTRATIONS FROM THE LITERATURE

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 
A series of relevant illustrations comes from studies of cancer: Such examples are now so numerous that it is accepted that the overwhelming majority of cancers result from the interactions of specific environments and genetic susceptibilities (Mulvihill & Tulinius, 1987). Specifically, the accumulation of evidence clearly indicates that etiologically, cancer cannot be viewed as a phenomenon controlled by either genetic or environmental factors alone; researchers consider the interaction between genes and environment as a unified causal mechanism (Hoover, 2000). This statement is supported by observations that only a small portion of cancers is characterized by specific disruption of single genes transmitted in a simply Mendelian fashion and that rates of cancer vary in association with change in exposure to different environmental (technologic, dietary, cultural) factors. Correspondingly, cancer researchers continue to seek to understand the action of such risk factors as oncogenes (genes that, under normal circumstances, promote cell growth and duplication but may undergo mutational changes, causing cells to grow too quickly and form tumors), tumor suppressor genes (genes that, under normal circumstances, slow down cell division or cause cells to die at the appropriate time but may undergo mutational alterations, leading to too much cell growth and cancer), DNA repair genes (genes that are involved in DNA repair pathways but, when mutationally altered, permit accumulation of damage throughout the DNA), genes encoding proteins involved in the metabolic processes of established risk factors, and genes that orchestrate the response of the organism to a potentially harmful agent in the context of co-participation of these genetic factors and cancer-triggering environmental factors in the same causal pathways.

To illustrate, consider an example of breast cancer. BRCA1 is a tumor suppressor gene. Most women are born with two copies of the gene and have no problem. Those with the inherited BRCA1 mutation have only a single copy. A single copy is sufficient for tumor suppression until, for some reason, it is "knocked out," and the woman is left without the gene. Then she becomes extremely vulnerable to breast or ovarian cancer. The lack of the BRCA1 allele is seen as the cause of breast cancer. Yet, could we not ask what causes the second copy of the gene to get knocked out? Could not environmental risk factors like a woman's past exposure to radiation be critical in understanding how the mutation leads to cancer (Conrad, 1999, as cited in Shostak, 2003, p. 2335)? Thus, the effects of genetic cancer risk factors are better understood when relevant environments are taken into account.

Of note is that the mutations in genes highly relevant to cancer, such as BRCA1 and BRCA2, present a 40–80% lifetime risk of developing disease (Mucci et al., 2001). However, these mutations are uncommon and account for <5–10% of the disease (Firozi et al., 2002). It is becoming more evident that common genetic alterations in genes involved in metabolizing potentially harmful chemical agents (which themselves might not substantially influence cancer risk) should be considered as important players in pathways toward cancer development when co-acting with environmental exposures.

Consider an example. There is an excess of epidemiologic data causally linking lung cancer and granulomatous lung disease (manifested as hypersensitivity pneumonitis or allergic alveolitis) to exposure to beryllium dust or fumes; these data led to classification of beryllium and its compounds as human carcinogens group 1 (International Agency for Research on Cancer, 1993). Exposure typically occurs among workers in beryllium-manufacturing plants where different technologic processes can produce respirable beryllium particles, but exposure can also occur among secondary users of beryllium products and in facilities where beryllium is present (Cullen, Cherniack, & Kominsky, 1986; Williams, 1994). Exposure to beryllium sets off a cell-mediated delayed hypersensitivity reaction, which in turn leads to the proliferation of beryllium-specific T lymphocytes. Such proliferation can result in granuloma formation and eventual respiratory failure characterized by high mortality rates (Newman, Lloyd, & Daniloff, 1996; Williams, 1996). Following an initial report implicating the major histocompatibility complex (MHC) class II cells in a beryllium-specific immune response (Richeldi, Sorrentino, & Saltini, 1993), a number of association studies have been done to investigate a strong association between the inheritance of a gene belonging to the MHC family of immune-response genes, human leukocyte antigen (HLA) DPB1^Glu69 alleles, and disease risk in workers exposed to beryllium (for a review, see McCanlies, Kreiss, Andrew, & Weston, 2003). The essence of this work is that although environmental exposure to beryllium is a prerequisite for developing granulomatous lung disease and beryllium-specific lung cancer, the inheritance of at least one HLA-DPB1^Glu69 allele "modifies an individual's risk" (McCanlies et al., 2003, p. 395) for the manifestation of these diseases by substantially increasing the odds ratios for granuloma formation in this individual.

Yet another relevant illustration is one of the many enzymes involved in cell metabolism of the carcinogen N-acetyltransferase (NAT). There are two NAT enzymes, NAT1 and NAT2; both enzymes act by attaching an acetyl compound to a carcinogenic compound so that a resulting metabolite becomes more reactive and manageable by the cell than the products of the original exposure (Evans, 1989). These enzymes are involved in (a) deactivation of aromatic amines, a component of tobacco smoke and other environmental exposures (e.g., cooking fumes, industrial carcinogens), but (b) activation of heterocyclic amines, formed in barbequed, broiled, or fried meats (Mucci et al., 2001). One of these enzymes, NAT2, is known to have several polymorphisms (Grant et al., 1997), which are typically subdivided into two groups as compared with the action of the normal allele: (a) rapid acetylator variants, increasing the rate of metabolism, and (b) slow acetylator variants, diminishing metabolic activity (Hein, Ferguson, Doll, Rustan, & Gray, 1994). Of interest here is the "double-edge sword" action of these variants. As the action of the NAT enzyme is typically defined as a detoxifying mechanism, the rapid acetylator variants should expedite the deactivation of carcinogenic aromatic amines, which would result in decreased exposure of cells to mutagens and the formation of DNA adducts. However, as the NAT enzyme activates and enhances carcinogenicity of specific amines (e.g., from proteins present in specific cooked foods, alcohol, and tobacco), rapid acetylator variants might increase the rate of formation of reactive compounds, which would result in increased exposure of cells to mutagens (Mucci et al., 2001). Yet, according to the literature, there are no clear patterns of direct associations between NAT-specific genotypes with respect to cancer susceptibility.

However, the literature is replete with different accounts of specific gene–environment interactions. For example, it was reported that dose–response relationships are important for increased risk of breast cancer for (a) female carriers of rapid acetylator variants who smoke 20 or more cigarettes per day (Agundez, Ladero, Olivera, Abildua, Roman, & Benitez, 1995) and (b) female carriers of slow acetylator variants who consume a great deal of well-done meat (Ambrosone et al., 1996). In addition, it is important to note that in the context of various environmental carcinogenic factors, (a) NAT polymorphisms are reportedly involved in manifestations of multiple forms of cancer (colorectal [e.g., Katoh et al., 2000], lung [e.g., Nyberg, Hou, Hemminki, Lambert, & Pershagen, 1998], kidney [Semenza et al., 2001], and bladder [Green, Banks, Berrington, Darby, Deo, & Newton, 2000]); (b) NAT genes appear to interact with other genes coding for detoxifying enzymes, resulting in further increased risk (Firozi et al., 2002); and (c) these relationships vary by population (Marcus, Vineis, & Rothman, 2000; Taylor et al., 1998) and gender (Semenza et al., 2001). Thus, the same or different genetic polymorphisms (e.g., NAT1 and NAT2 polymorphism) might co-act with the same or different environments (e.g., smoking or cooking fumes) and impact the same or different phenotypes (e.g., various forms of cancer). Clearly, this example underscores the complexities related to detection and quantification of gene–environment interactions.

Although cancer studies provide the most examples of gene–environment interaction, examination of gene–environment interaction is not limited to cancer research. Consider the following examples. One of the newest challenges of public medicine is the increased prevalence of allergic diseases (Bjorksten et al., 1998). The major target in battling this rising prevalence has been environment; however, it is well known that both asthma (Holberg et al., 1996) and total serum immunoglobulin (Ig) E, one of the many proteins involved in the immune response, levels are genetically controlled (Marsh et al., 1994). In their literature review, Baldini and colleagues (Baldini, Vercelli, & Martinez, 2002) investigated a hypothesis of linking improved environments to worsened public health with respect to these particular types of diseases. Their reasoning went like this: Enforcement of public health policies resulted in dramatically improved hygiene conditions. One of the side effects of this generic improvement was in the associated change in "the type and level of stimulation from the microbial environment which may have indirectly influenced the postnatal development of immune system, leading to an increased predisposition to allergic diseases during childhood" (Baldini et al., 2002, pp. 188–189). Generally speaking, the patterns of immune response registered in newborns and in older children and adults are very different. Specifically, whereas immature immune responses are characterized by the dominance of the so-called T-helper (Th) 2-type cytokines (e.g., specific types of the protein interleukin [IL]-4, 5, 6, and 10), the mature responses are associated with the dominance of Th1-type cytokines. Thus, the maturation of the immune system is associated with a transition from the low- to high-Th1 profile. This maturation is a complex process, but some elements of this process have been successfully characterized.

Specifically, two factors, one environmental and one genetic, are of interest for this discussion. A key factor in the initiation of immune response to bacterial infection is an endotoxin bacterial lipopolysaccharide (LPS). The presentation of LPS simulates a cascade of cellular effects such as stimulation of IL-12 production, which is hypothesized to be a requisite signal for the differentiation of naive T cells into Th1-like cells. It is of note that the absence of IL-12 has been linked to the induction of IL-4 and interferon (both are Th2-type cytokines) production in naive T cells, preventing or slowing down their transformation into Th1 cells. This hypothesis of the importance of early life exposure to LPS was supported by evidence revealing substantially lower prevalence of allergic diseases (Von Ehrenstein, Von Mituis, Illi, Baumann, Bohm, & Von Kries, 2000; Von Mutius et al., 2000) and lower allergic sensitization (Hesselmar, Aberg, Aberg, Eriksson, & Bjorksten, 1999) in young children living in the same geographic areas as control children, but who differed in the factor of consistent exposure to animals and, correspondingly, LPS. However, it has been shown that in individuals with diagnosed asthma, the exposure to LPS increases the severity of the disease (Kline et al., 1999). Thus, exposure to LPS appears to be an environmental factor modulating the development of the immune system, the effect of which depends on the timing of exposure.

The genetic factor of relevance is the CD14 gene located at 5q31.33 (Goyert, Ferrero, Rettig, Yenamandra, Obata, & Le Beau, 1988). The gene encodes the receptor for LPS; the LPS-driven engagement of the CD14 gene dramatically increases the production of IL-12. Consequently, functional variability in the CD14 gene resulting in lower levels of CD14 has been linked (for a review, see Baldini et al., 2002) to developmental deficiency in Th1 function. Moreover, it has been demonstrated that blood cells from farmers' children expressed significantly higher amounts of CD14 than those from nonfarmers' children, indicating gene–environment modulation of the development of allergic disease (Lauener et al., 2002). Thus, Baldini and colleagues (2002) hypothesized that the effects of the presence of the CD14-based genetic risk factors might be magnified by the absence of early developmental exposure to LPS, preventing innately defective Th1 function from being strengthened by environmental challenge absent in environments with improved sanitary and nutritional conditions. As allergens stimulate adaptive immune responses that are characterized by Th2 profile and enhanced IgE production, the presence of risk alleles in CD14 might lead to down-regulation of the production of Th1-type cytokines, resulting in the predominance of Th2 response profile. The decrease in opportunities for young children to experience LPS attacks at crucial times in the development of their immune systems might result in magnification, rather than suppression, of the effects of genetic risk factors. Thus, this example provides yet another dimension of the complexity of detecting gene–environment interaction effects: It is possible that many of these effects are time constrained and detectable only at certain developmental phases.

The final example in this section arises from a different area of research, that of exceptional performance. Angiotensin-converting enzyme (ACE) is responsible for metabolism of a number of peptides (i.e., simple proteins including a limited number of amino acids) in various tissues, including skeletal muscles. Physiologically, ACE is a central element in the functioning of the metabolic renin–angiotensin and the kallikrein–kinin systems. It is thought to influence circulatory homeostasis through the degradation of vasodilator kinins and the derivation of the vasopressor octapeptide angiotensin II (Woods, Humphries, & Montgomery, 2000). At a population level, there is a large amount of interindividual variation in plasma ACE levels; this variation, however, is clustered within families and is controlled, at least partially, by the ACE gene located on chromosome 17 (17q23.3) and also known as DCP1 (Woods et al., 2000). There are two polymorphisms in the DCP1 gene: so-called D and I alleles. The D allele is associated with increased levels of serum, tissue ACE, and vasopressor angiotensin II, and decreased half-life of the vasodilator bradykinin. Individuals homozygous for the D allele were reported to have a worse prognosis for a number of cardiac and renal conditions (Woods et al., 2000). In addition, it has been reported that DD individuals who smoke appeared to be at higher risk for arteriosclerosis (Sayed-Tabatabaei et al., 2004). The I allele is associated with lower ACE activity in both serum and tissues. Studies indicate: (a) higher frequencies of the I allele and the II genotype in individuals from specific high-performing groups of athletes such as elite distance runners, rowers, and mountaineers as compared with the general population (e.g., Gayagay et al., 1998; Montgomery et al., 1998; Myerson, Hemingway, Budget, Martin, Humphries, & Montgomery, 1999); (b) substantially higher training-related increase of loaded repetitive biceps flexion among individuals who are homozygous for the I allele as compared with those homozygous for the D allele (Montgomery et al., 1998); and (c) an I allele–associated performance advantage correlating with an increasing element of endurance in sports.

It is notable, however, that these results were reported when specific groups of athletes engaged in particular sports (e.g., biking and mountain climbing) were compared with control groups from the general population. However, when variability linked to different sports, degrees of endurance within each sport, and degrees of performance among the athletes was introduced so as to dilute the homogeneity of the target group, the associations were not replicated (Karjalainen et al., 1999; Rankinen et al., 2000; Taylor, Mamotte, Fallon, & van Bockxmeer, 1999). Attempting to reconcile this pattern of diverse results, Woods and colleagues (2000) reviewed relevant publications on DCP1-based association studies and mechanisms of the ACE action and suggested the following interpretation: If the physiologic impact of the I allele is via a local muscle effect rather than a central cardiorespiratory mechanism and in enhanced vasodilation and substrate delivery to the working muscles, then the engagement in physical training is a necessary condition for benefiting from the presence of the I allele (and the II genotype). In other words, the improved muscle efficiency associated with the low ACE levels, which in turn is associated with the I allele, can occur only during a structured training program assuming high levels of endurance. Thus, the performance advantage reflecting the benefit of the gene (the I allele) can be recovered only in certain environments (prolonged periods of highly demanding and specific training); the gene–environment interaction is a necessary condition for the manifestation of the effect.

In an attempt to summarize the gene–environment interaction data in the field, Kelada and colleagues (2003) compiled an illustrative table of links between specific diseases and gene–environment interactions. A few important observations have been made based on this summary. First, some genes appear to be associated with a number of different environmental exposures, thus forming a set of gene–environment interactions that might be linked to the same or different phenotypes. Second, the literature contains a number of contradictory findings both in terms of the presence or absence of effects and in terms of an effect's direction and magnitude with regard to the same phenotypes. Third, although complex, the current examples of gene–environment interactions in the literature are rather simple, relatively speaking (first order), and typically involve only two factors. As Kelada and colleagues (2003) pointed out, risks for common complex diseases are likely to be derived as functions of multiple genes interacting with each other and multiple environments to produce higher-order interactions.

Thus, even in the fields of study where gene–environment interactions have been accepted as major "players," a great deal of contradictory data, complexities in interpretation, concerns about lack of power to detect these interactions, and other caveats are still encountered. Clearly, the large-scale, well-designed studies more and more common in the field will resolve at least some, if not all, of these contradictions. Below, in anticipation of such studies, a few comments of relevance are offered.

	THE STRUCTURE OF INTERACTIONS: APPRECIATING THE COMPLEXITY

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 
At this point, the field's knowledge of anticipated sample sizes needed to reliably detect effects of interest (i.e., power considerations), including gene–environment interaction, is not extensive. As per a general consideration regarding the detection and estimation of the magnitude of interactive effects (Hallqvist, Ahlbom, Didrechsen, & Reuterwall, 1996), the sample sizes required to detect gene–environment interactions and appreciate their effects are large. Correspondingly, "current sample sizes are significantly underpowered to evaluate either the interactions between specific genes or the G x E effects" (Gunzerath & Goldman, 2003, p. 550). Therefore, it is no surprise that studies of gene–environment interactions are susceptible to both type I and type II statistical error (Sayed-Tabatabaei et al., 2004).

Although "underpowering" is typical, some designs require smaller sample sizes than others (e.g., case-only relative design is more efficient than case–control design [Gauderman, Witter, & Thomas, 1999; Witter, Gauderman, & Thomas, 1999]), but typically the advantages are stratified (e.g., the case–sibling design is preferred for a dominant mode of transition, whereas the case–parent design is advantageous for a recessive mode [Gauderman, 2002]). However, although these stimulation studies are extremely helpful for obtaining ideas about the designs and sample sizes needed for detection and quantification of gene–environment interaction, they are also quite sobering, given that very few, if any, complex human traits will follow clear-cut dominant/recessive modes of transmission.

In addition, there is another characteristic of gene–environment interaction studies that has barely been noted. This issue brings out for discussion the fact mentioned by Kelada and colleagues in their review (2003): We most likely grossly oversimplify interactions when we model them. What has been investigated or modeled so far are mostly first-order interactions, where either identified or hypothesized high-risk genetic factors are assumed to co-act with high-risk environments.

A handful of examples of such first-order interactions can be found in the applications of gene–environment interaction models in alcoholism. For example, as reviewed by Heath and Nelson (2002), a number of gene–environment interactions appear to be relevant to the development of alcohol dependency and alcoholism. Specifically, in determining relevant risk, what must be taken into account are not only genetic vulnerability (Heath, Slutske, & Madden, 1997) and alcohol-associated environmental factors such as peer pressure to have a drink (Higuchi, Matsushita, Imazeki, Kinoshita, Takagi, & Kono, 1994), growing up in rural and urban settings (Dick, Rose, & Viken, 2001), marital status (Heath, Jardine, & Martin, 1989), religious affiliation (Boomsma, de Geus, van Baal, & Koopmans, 1999), and history of child abuse (Nelson, Little, Heath, & Kessler, 1996), but also multiple possible interactions between susceptibility genes and risky environments (Gunzerath & Goldman, 2003).

However, even hypothesizing a possible structure for such interactions is difficult. One way of addressing this difficulty is in stating that the solution lies in "disentangling them [risk factors] or even in discerning which are primarily environmental or genetic" (Gunzerath & Goldman, 2003, p. 551). Another possible way of addressing this difficulty is in attempting to subtype these interactions, if possible, into higher-order and more complex interactions. There are multiple examples of such subtyping in plant genetics (e.g., Yadav et al., 2003), but very few in animal and human genetics.

For example, smoking is typically considered to be an environmental risk factor for a number of cancers and is especially detrimental for the carriers of cancer-related genetic risk factors (see above). However, smoking per se is a heritable trait (Li, 2003; Tyndale, 2003), the manifestation of which is moderated by various environmental factors (Afifi Soweid, Khawaja, & Salem, 2004) and gene–environment interactions (e.g., Silberg, Rutter, D'Onofrio, & Eaves, 2003). Thus, the "environmental" factors of smoking itself can be represented as being influenced by genes, environments, and gene–environment interactions.

Similarly, diseases unfolding during the life span and viewed as pathologic consequences of aging in fact manifest themselves within individual particularities of aging. In other words, when aging is viewed as a process occurring independently of unfortunate "side effects" of aging such as disease (e.g., Johnson, 2002), the course of disease is nested within individual differences in aging. Thus, genetic mechanisms affecting aging themselves create environmental context for the manifestation of many diseases that occur late in life.

There are many similar examples, especially in the general medical literature. For example, the predominant way of acquiring schistosomiasis (a class reference to diseases of liver, gastrointestinal tract, and bladder caused by schistosomes, trematode worms that parasitize people) is to be infected from infested water. Thus, it is a purely environmental event. However, it has been convincingly demonstrated that both the intensity of infection (Marquet et al., 1996; Zinn-Justin, Marquet, Hillaire, Dessein, & Abel, 2001) and the susceptibility for reinfection are genetically controlled (e.g., Ross et al., 1998). However, the duration and intensity of infection, by themselves, are environmental risk factors for severe hepatic fibrosis, a lethal disease developing in 2–10% of individuals infected by a type of schistosomes (Schistosoma mansoni); yet, the occurrence of this disease is genetically controlled (Chevillard et al., 2003; Dessein et al., 1999). This work, collectively, provides a wonderful example of nested gene–environment interactions, where effects that are perceived as environmental (e.g., duration and intensity of infection with regard to the development of hepatic fibrosis) are themselves outcomes of genes, environments, and their interaction.

This work also has another dimension: All genes identified so far are immune genes, involved in common transduction pathways. This observation raises a question of whether such stratification of gene–environment interactions might be helpful in enhancing power for gene mapping, as the genes controlling the development of complex human traits are likely to be characterized by pleiotropic effects (the influence of a single gene over multiple phenotypes), if not individually, then at least in terms of the pathways in which these genes are involved.

At this point, there is very little evidence in the literature to support this hypothesis. However, consideration of gene–environment interaction in genetic studies has been stated to enhance the power of gene-mapping efforts in recent theoretical (e.g., Purcell & Sham, 2002), simulation (Selinger-Leneman et al., 2003), and empirical (Martin et al., 2002) studies. Thus, there is at least a firm foundation for further exploration of the potential benefits of modeling gene–environment interactions as nested effects.

	IN LIEU OF CONCLUSION

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 
At the onset of this article, I suggested that Khoury and colleagues' (1993) statement that all human diseases are an outcome of gene–environment interactions could be extended by stating that understanding of all complex human traits is conditioned upon understanding of the contribution of gene–environment interactions to the expression of these traits. This, of course, is not a new thought for psychology. It can be traced to writings of a number of researchers, for example, Turkheimer:

"Individual genes ... and their environments (which include other genes) interact to initiate a complex developmental process that determines adult personality. Most characteristic of this process is its interactivity: Subsequent environments to which the organism is exposed depend on earlier states, and each new environment changes the developmental trajectory, which affects future expression of genes and so forth. Everything is interactive, in the sense that [nothing] proceeds from cause to effect; any individual gene or environment event produces an effect only by interacting with other genes and environment." (p. 161)

The task at hand, however, is to break this seemingly all-inclusive circle and start untangling this web of endless gene and environment co-action. Yes, it will require large samples, appreciation of cultural, cohort, and developmental effects, exploration of new designs and methodologies, revisions of phenotypes of interest, improving the precision of measurement, and many other factors that are not necessarily readily available to researchers. However, understanding these interactions might be our only pathway to "phenotypic prevention" (Shostak, 2003, p. 2330): a science of preventing unwanted phenotypes from arising not by altering an individual, but by letting an individual develop without ever manifesting detrimental phenotypes.

The development of such phenotypic preventions appears to be especially relevant to research on aging. There is a growing awareness that detrimental environmental exposures occurring in the presence of genetic predisposition might result in chronic conditions. A relevant example is that of obesity: Genes presumed to be evolutionarily protective in the context of famine are hypothesized to interact negatively with the modern environments of the developed world, where calories are present in excess; they also have protective functions in environments of the developing world, where food can be scarce (Collins, 2004). Because research on aging allows studying extended or even lifelong exposure to certain environments, studies of aging populations promise new and important realizations of the role of gene–environment interaction throughout the life span.

Based on the evidence available in the literature, the most effective studies of gene–environment interactions are those of large case–control studies of prospective and retrospective cohorts. Although potentially extremely valuable, these studies require the consideration of a number of factors at the design level, which will maximize their potential. As pointed out by Collins (2004), these considerations might include

• large numbers of participants to ensure adequate sample sizes (up to several hundred thousand)
  
• overrepresentation of minority groups to allow for determination of ethnic and cultural effects
  
• depending on the phenotype of interest, broad sampling from groups representing at-risk genotypes (e.g., APOE4 versus other genotypes relevant to Alzheimer's disease) and types of environmental exposure
  
• family recruitment to include multiple generations to examine cohort effects, developmental trajectories, and genetic mechanisms of vertical transmission
  
• access to databases of clinical and laboratory information regarding the disease of interest
  
• records and assessments of dietary and lifestyle characteristics
  
• collection and storage of biological specimens
  
• creation of databanks accessible to multiple scientists from different disciplines
  
• dynamic modification of hypotheses generated at the onset of the study and reflecting the development of the field
  
• engagement of work and life communities in which the participants are engaged, and
  
• explanation of the premise and potential outcomes of the work securing informed and confidential consent procedures and fulfillment of multiple ethical considerations.
  

Clearly, such studies can be conducted only with remarkable amounts of scientific, financial, and political support. A number of cohort studies of impressive magnitudes that will provide a great deal of information on human diseases, in general, and the interaction between diseases and aging, in particular, are already in progress in the United Kingdom, Iceland, Estonia, Germany, Canada, and Japan, and we in the field hope to learn a tremendous amount about the role of gene–environment interactions from these studies.

	Acknowledgments

 
Preparation of this essay was supported by grant REC-9979843 from the National Science Foundation and by a grant under the Javits Act Program (grant no. R206R00001) as administered by the Institute for Educational Sciences, U.S. Department of Education. Grantees undertaking such projects are encouraged to express freely their professional judgment. This article, therefore, does not necessarily represent the position or policies of the National Science Foundation, the Institute for Educational Sciences, or the U.S. Department of Education, and no official endorsement should be inferred. I express my gratitude to Ms. Robyn Rissman for her editorial assistance.

	References

TOP Abstract Definitions, Assumptions, and... Gene-Environment Interaction:... Gene-Environment Interaction:... The Structure of Interactions:... In Lieu of Conclusion References

 

• Afifi Soweid, R. A., Khawaja, M., Salem, M. T. (2004). Religious identity and smoking behavior among adolescents: Evidence from entering students at the American University of Beirut. Health Communication, 16,47-62.[Medline]
• Agundez, J. A., Ladero, J. M., Olivera, M., Abildua, R., Roman, J. M., Benitez, J. (1995). Genetic analysis of the arylamine N-acetyltransferase polymorphism in breast cancer patients. Oncology, 52,7-11.[Medline]
• Ambrosone, C. B., Freudenheim, J. L., Graham, S., Marshall, J. R., Vena, J. E., Brasure, J. R., et al. (1996). Cigarette smoking, N-acetyltransferase 2 genetic polymorphisms, and breast cancer risk. JAMA, 276,1494-1501.[Abstract/Free Full Text]
• Armstrong, B. K., Kricker, A., English, D. R. (1997). Sun exposure and skin cancer. Australasian Journal of Dermatology, 38,(suppl. 1), S1-S6.
• Baldini, M., Vercelli, D., Martinez, F. D. (2002). CD14: an example of gene by environment interaction in allergic disease. Allergy, 57,188-192.[Medline]
• Bernlohr, D. A., Simpson, M. A., Hertzel, A. V., Banaszak, L. J. (1997). Intracellular lipid-binding proteins and their genes. Annual Review of Nutrition, 17,277-303.[Medline]
• Bjorksten, B., Dumitrascu, D., Foucard, T., Khetsuriani, N., Khaitov, R., Leja, M., et al. (1998). Prevalence of childhood asthma, rhinitis and eczema in Scandinavia and Eastern Europe. European Respiratory Journal, 12,432-437.[Abstract]
• Bohman, M. (1996). Predisposition to criminality: Swedish adoption studies in retrospect. In G. R. Bock & J. A. Goode (Eds.), Genetics of criminal and antisocial behavior (pp. 99–114). Chichester, UK: Wiley.
• Boomsma, D. L., de Geus, E. J., van Baal, G. C., Koopmans, J. R. (1999). A religious upbringing reduces the influence of genetic factors on disinhibition: Evidence for interaction between genotype and environment on personality. Twin Research, 2,115-125.
• Brennan, P. A., Mednick, S. A., Jacobsen, B. (1996). Assessing the role of genetics in crime using adoption cohorts. In G. R. Bock & J. A. Goode (Eds.), Genetics of criminal and antisocial behavior (pp. 115–128). Chichester, UK: Wiley.
• Cadoret, R. J., Yates, W. R., Troughton, E., Wodworth, G., Stewart, M. A. (1995). Gene–environment interaction in the genesis of aggressivity and conduct disorders. Archives of General Psychiatry, 52,916-924.[Abstract/Free Full Text]
• Carey, G. (2002). Human genetics for the social sciences. Thousand Oaks, CA: Sage.
• Chevillard, C., Moukoko, C. E., Elwali, N. E., Bream, J. H., Kouriba, B., Argiro, L., et al. (2003). IFN-gamma polymorphisms (IFN-gamma +2109 and IFN-gamma +3810) are associated with severe hepatic fibrosis in human hepatic schistosomiasis (Schistosoma mansoni). Journal of Immunology, 171,5596-5601.[Abstract/Free Full Text]
• Collins, F. S. (2004). The case of a US prospective cohort study of genes and environment. Nature, 249,475-477.
• Conrad, P. (1999). A mirage of genes. Sociology of Health and Illness, 21,228-241.
• Cooper, R. M., Zubek, J. P. (1958). Effects of enriched and restricted early environment on the learning ability of bright and dull rats. Canadian Journal of Psychology, 12,159-164.[Medline]
• Corander, J., Sillanpää, M. J. (2002). A unified approach to joint modeling of multiple quantitative and qualitative traits in gene mapping. Journal of Theoretical Biology, 218,435-446.[Medline]
• Crabbe, J. C., Wahlsten, D., Dudek, B. C. (1999). Genetics of mouse behavior: Interactions with laboratory environment. Science, 284,1670-1672.[Abstract/Free Full Text]
• Cullen, M. R., Cherniack, M. G., Kominsky, J. R. (1986). Chronic beryllium disease in the United States. Seminars in Respiratory Medicine, 7,203-209.
• Dessein, A. J., Hillaire, D., Elwali, N. E., Marquet, S., Mohamed-Ali, Q., Mirghani, A., et al. (1999). Severe hepatic fibrosis in Schistosoma mansoni infection is controlled by a major locus that is closely linked to the interferon-gamma receptor gene. American Journal of Human Genetics, 65,709-721.[Medline]
• Dick, D. M., Rose, R. J., Viken, R. J. (2001). Exploring gene–environment interactions: Socioregional moderation of alcohol use. Journal of Abnormal Psychology, 110,625-632.[Medline]
• Eaves, L. J., Erkanli, A. (2003). Markov Chain Monte Carlo approaches to analysis of genetic and environmental components of human developmental change and G x E interaction. Behavior Genetics, 33,279-299.[Medline]
• Eaves, L. J., Last, K., Martin, N. G., Jinks, J. L. (1977). A progressive approach to non-additivity and genotype–environment covariance in the analysis of human differences. British Journal of Mathematic and Statistical Psychology, 30,1-42.
• Eaves, L. J., Sullivan, P. (2001). Genotype–environment interaction in transmission disequilibrium tests. Advances in Genetics, 42,223-240.[Medline]
• Evans, D. A. N. (1989). N-Acetyltransferase. Pharmacological Therapy, 42,157-243.
• Falconer, D. S., Mackay, T. F. C. (1996). Introduction to quantitative genetics (4th ed.). Essex, UK: Longman.
• Firozi, P. F., Bondy, M. L., Sahin, A. A., Chang, P., Lukmanji, F., Singetary, E. S., et al. (2002). Aromatic DNA adducts and polymorphisms of CYP1A1, NAT2, and GSTM1 in breast cancer. Carcinogenesis, 23,301-306.[Abstract/Free Full Text]
• Gauderman, W. (2002). Sample size requirements for matched case–control studies of gene–environment interaction. Statistics in Medicine, 21,35-50.[Medline]
• Gauderman, W., Witter, J., Thomas, D. (1999). Family-based association studies. Journal of the National Cancer Institute Monographs, 26,31-39.
• Gayagay, G., Yu, B., Hambly, B., Boston, T., Hahn, A., Celermajer, D. S., et al. (1998). Elite endurance athletes and the ACE I allele—The role of genes in athletic performance. Human Genetics, 103,48-50.[Medline]
• Goyert, S. M., Ferrero, E., Rettig, W. J., Yenamandra, A. K., Obata, F., Le Beau, M. M. (1988). The CD14 monocyte differentiation antigen maps to a region encoding growth factors and receptors. Science, 239,497-500.[Abstract/Free Full Text]
• Grant, D. M., Hughes, N. C., Janezic, S. A., Goodfellow, G. H., Chen, H. J., Gaedigk, A., et al. (1997). Human acetyltransferase polymorphisms. Mutation Research, 376,61-70.[Medline]
• Green, J., Banks, E., Berrington, A., Darby, S., Deo, H., Newton, R. (2000). N-Acetyltransferase 2 and bladder cancer: An overview and consideration of the evidence for gene–environment interaction. British Journal of Cancer, 83,412-417.[Medline]
• Gunzerath, L., Goldman, D. (2003). G x E: A NIAAA Workshop on Gene–Environment Interactions. Alcoholism: Clinical and Experimental Research, 27,540-562.[Medline]
• Haldane, J. B. S. (1938). Heredity and politics. London: George Allen and Unwin.
• Hallqvist, J., Ahlbom, A., Didrechsen, F., Reuterwall, C. (1996). How to evaluate interaction between causes: A review of practices in cardiovascular epidemiology. Journal of Internal Medicine, 239,377-382.[Medline]
• Heath, A. C., Jardine, R., Martin, N. G. (1989). Interactive effects of genotype and social environment on alcohol consumption in female twins. Journal of Studies on Alcohol, 50,38-48.[Medline]
• Heath, A. C., Nelson, E. C. (2002). Effects of the interaction between genotype and environment. Alcohol Research and Health, 26,193-201.[Medline]
• Heath, A. C., Slutske, W. S., Madden, P. A. F. (1997). Gender differences in the genetic contribution to alcoholism risk and to alcohol consumption patterns. In R. W. Wilsnack & S. C. Wilsnack (Eds.), Gender and alcohol (pp. 114–149). Rutgers: Rutgers University Press.
• Hegele, R. A., Young, T. K., Connelly, P. W. (1997). Are Canadian Inuit at increased genetic risk for coronary heart disease? Journal of Molecular Medicine, 75,364-370.[Medline]
• Hein, D. W., Ferguson, R. J., Doll, M. A., Rustan, T. D., Gray, K. (1994). Molecular genetics of human polymorphic N-acetyltransferase: Enzymatic analysis of 15 recombinant wild-type, mutant, and chimeric NAT2 allozyme. Human Molecular Genetics, 3,729-734.[Abstract/Free Full Text]
• Hesselmar, B., Aberg, N., Aberg, B., Eriksson, B., Bjorksten, B. (1999). Does early exposure to cat or dog protect against later allergy development? Clinical and Experimental Allergy, 29,611-617.[Medline]
• Higuchi, S., Matsushita, S., Imazeki, H., Kinoshita, T., Takagi, S., Kono, H. (1994). Aldehyde dehydrogenase genotypes in Japanese alcoholics. Lancet, 343,30-37.
• Holberg, C., Elston, R., Halonen, M., Wright, A., Taussig, L., Morgan, W., et al. (1996). Segregation analysis of physician-diagnosed asthma in Hispanic and non-Hispanic white families. A recessive component? American Journal of Respiratory and Critical Care Medicine, 154,144-150.[Abstract]
• Hoover, R. N. (2000). Cancer—Nature, nurture, or both? New England Journal of Medicine, 343,135-136.[Free Full Text]
• Hussain, S. P., Harris, C. C. (1998). Molecular epidemiology of human cancer: Contribution of mutation spectra studies of tumor suppressor genes. Cancer Research, 58,4023-4037.[Free Full Text]
• Iannaccone, P. M. (2001). Toxicogenomics: "The call of the wild chip.". Environmental Health Perspectives, 109,A8-A11.[Medline]
• International Agency for Research on Cancer. (1986). Tobacco smoking. Lyon, France: International Agency for Research on Cancer, World Health Organization.
• International Agency for Research on Cancer. (1993). Some metals and metallic compounds: Beryllium and beryllium compounds. Lyon, France: International Agency for Research on Cancer, World Health Organization.
• Jinks, J. L., Fulker, D. W. (1970). Comparison of the biometrical, genetic, MAVA and classical approaches to the analysis of human behavior. Psychological Bulletin, 73,311-349.[Medline]
• Johnson, T. E. (2002). Subfield history: Caenorhabditis elegans as a system for analysis of the genetics of aging. Science of Aging Knowledge Environment, 2002, RE4..
• Jones, P. A., Buckley, J. D., Henderson, B. E., Ross, R. K., Pike, M. C. (1991). From gene to carcinogen: A rapidly evolving field in molecular epidemiology. Cancer Research, 1,3617-3620.
• Karjalainen, J., Kujala, U. M., Stolt, A., Mantysaari, M., Viitasalo, M., Kainulainen, K., et al. (1999). Angiotensinogen gene M235T polymorphism predicts left ventricular hypertrophy in endurance athletes. Journal of American College of Cardiology, 34,494-499.[Abstract/Free Full Text]
• Katoh, T., Boissy, R., Nagata, N., Kitagawa, K., Kuroda, Y., Itoh, H., et al. (2000). Inherited polymorphism in the N-acetyltransferase 1 (NAT1) and 2 (NAT2) genes and susceptibility to gastric and colorectal adenocarcinoma. International Journal of Cancer, 85,46-49.
• Kelada, S. N., Eaton, D. L., Wang, S. S., Rothman, N. R., Khoury, M. J. (2003). The role of genetic polymorphisms in environmental health. Environmental Health Perspectives, 111,1055-1064.[Medline]
• Khoury, M. J., Beaty, T. H., Cohen, B. H. (1993). Fundamentals of genetic epidemiology. New York: Oxford University Press.
• Khoury, M. J., Burke, W., Thomson, E. J. (2000). Genetics and public health in the 21st century (vol. 40). Oxford, UK: Oxford University Press.
• Kline, J. N., Cowden, J. D., Hunninghake, G. W., Schutte, B. C., Watt, J. L., Wohlford-Lenane, C. L., et al. (1999). Variable airway responsiveness to inhaled lipopolysaccharide. American Journal of Respiratory and Critical Care Medicine, 160,297-303.[Abstract/Free Full Text]
• Lauener, R. P., Birchler, T., Adamski, J., Braun-Fahrlander, C., Bufe, A., Herz, U., et al. (2002). Expression of CD14 and Toll-like receptor 2 in farmers' and non-farmers' children. Lancet, 360,465-466.[Medline]
• Li, M. D. (2003). The genetics of smoking related behavior: A brief review. American Journal of Medical Sciences, 326,168-173.
• Li, Z. K., Yu, S. B., Lafitte, H. R., Huang, N., Courtois, B., Hittalmani, S., et al. (2003). QTL x environment interactions in rice. I. Heading date and plant height. Theoretical and Applied Genetics, 108,141-153.[Medline]
• Liang, K.-Y. (1987). Extended Mantel–Haenszel estimating procedures for multivariate logistic regression models. Biometrics, 43,289-299.[Medline]
• Liang, K.-Y., Beaty, T. H. (2000). Statistical designs for familial aggregation. Statistics in Medical Research, 9,543-562.
• Liang, K.-Y., Zeger, S. L. (1993). Regression analysis for correlated data. Annual Review of Public Health, 14,43-68.[Medline]
• Malerba, M., Radaeli, A., Ceriani, L., Tantucci, C., Grassi, V. (2003). Airway hyperresponsiveness in a large group of subjects with alpha1-antitrypsin deficiency: A cross-sectional controlled study. Journal of Internal Medicine, 253,351-358.[Medline]
• Marcus, P. M., Hayes, R. B., Vineis, P., Garcia-Closas, M., Caporaso, N. E., Autrup, H., et al. (2000). Cigarette smoking, N-acetyltransferase 2 acetylation status, and bladder cancer risk: A case-series meta-analysis of a gene–environment interaction. Cancer Epidemiology, Biomarkers, and Prevention, 9,461-467.[Abstract/Free Full Text]
• Marcus, P. M., Vineis, P., Rothman, N. (2000). NAT2 slow acetylation and bladder cancer risk: A meta-analysis of 22 case–control studies conducted in the general population. Pharmacogenetics, 10,115-122.[Medline]
• Marquet, S., Abel, L., Hillaire, D., Dessein, H., Kalil, J., Feingold, J., et al. (1996). Genetic localization of a locus controlling the intensity of infection by Schistosoma mansoni on chromosome 5q31–q33. Nature Genetics, 14,181-184.[Medline]
• Marsh, D. G., Neely, J. D., Breazeale, D. R., Ghosh, B., Freidhoff, L. R., Ehrlich-Kautzky, E., et al. (1994). Linkage analysis of IL4 and other chromosome 5q31.1 markers and total serum immunoglobulin E concentrations. Science, 264,1152-1156.[Abstract/Free Full Text]
• Martin, L. J., Mahaney, M. C., Almasy, L., Hixson, J. E., Cole, S. A., MacCluer, J. W., et al. (2002). A quantitative trait locus on chromosome 22 for serum leptin levels adjusted for serum testosterone. Obesity Research, 10,602-607.[Medline]
• McCanlies, E. C., Kreiss, K., Andrew, M., Weston, A. (2003). HLA-DPB1 and chronic beryllium disease: A HuGE review. American Journal of Epidemiology, 157,388-398.[Abstract/Free Full Text]
• Montgomery, H. E., Marshall, R., Hemingway, H., Myerson, S., Clarkson, P., Dollery, C., et al. (1998). Human gene for physical performance. Nature, 393,221-222.[Medline]
• Mucci, L. A., Wedren, S., Tamimi, R. M., Trichopoulos, D., Adami, H.-O. (2001). The role of gene–environment interaction in the aetiology of human cancer: Examples from cancers of the large bowel, lung, and breast. Journal of Internal Medicine, 249,477-493.[Medline]
• Mulvihill, J. J., Tulinius, H. (1987). Cancer ecogenetics: Studying genetic and environment interactions through epidemiology. International Journal of Epidemiology, 16,337-340.[Free Full Text]
• Murphy, G. M. J., Kremer, C., Rodrigues, H. E., Schatzberg, A. F. (2003). Pharmacogenetics of antidepressant medication intolerance. American Journal of Psychiatry, 160,1830-1835.[Abstract/Free Full Text]
• Myerson, S., Hemingway, H., Budget, R., Martin, J., Humphries, S., Montgomery, H. (1999). Human angiotensin I-converting enzyme gene and endurance performance. Journal of Applied Physiology, 87,1313-1316.[Abstract/Free Full Text]
• Neale, M. C., Cardon, L. R. (1992). Methodology for genetic studies of twins and families (vol. 67). Dordrecht, the Netherlands: Kluwer.
• Nelson, C. B., Little, R. J., Heath, A. C., Kessler, R. C. (1996). Patterns of DSM-III-R alcohol dependence symptom progression in a general population survey. Psychological Medicine, 26,449-460.[Medline]
• Newman, L. S., Lloyd, J., Daniloff, E. (1996). The natural history of beryllium sensitization and chronic beryllium disease. Environmental Health Perspectives, 104S,(suppl.), 937-943.
• Nyberg, F., Hou, S. M., Hemminki, K., Lambert, B., Pershagen, G. (1998). Glutathione S-transferase mu1 and N-acetyltransferase 2 genetic polymorphisms and exposure to tobacco smoke in nonsmoking and smoking lung cancer patients and population controls. Cancer Epidemiology, Biomarkers, and Prevention, 7,875-883.[Abstract]
• Perera, F., Weinstein, I. B. (1999). Molecular epidemiology: Recent advances and future directions. Carcinogenesis, 213,517-524.
• Plomin, R., DeFries, J. C., McClearn, G. E., Rutter, M. (1997). Behavioral genetics (3rd ed.). New York: W. H. Freeman.
• Puga, A., Micka, J., Chang, C. Y., Liang, H. C., Nebert, D. W. (1996). Role of molecular biology in risk assessment. Advances in Experimental Medicine & Biology, 387,395-404.
• Purcell, S., Sham, P. (2002). Variance components models for gene–environment interaction in quantitative trait locus linkage analysis. Twin Research, 5,572-576.
• Rankinen, T., Wolfarth, B., Simoneau, J. A., Maier-Lenz, D., Rauramaa, R., Rivera, M. A., et al. (2000). No association between the angiotensin-converting enzyme ID polymorphism and elite endurance athlete status. Journal of Applied Physiology, 88,1571-1575.[Abstract/Free Full Text]
• Reich, E. E., Montine, K. S., Gross, M. D., Roberts, L. J., II, Swift, L. L., Morrow, J. D., et al. (2001). Interactions between apolipoprotein E gene and dietary -tocopherol influence cerebral oxidative damage in aged mice. Journal of Neuroscience, 21,5993-5999.[Abstract/Free Full Text]
• Remold, S. K., Lenski, R. E. (2001). Contribution of individual random mutations to genotype-by-environment interactions in Escherichia coli. Proceedings of the National Academy of Sciences U S A, 98,11388-11393.[Abstract/Free Full Text]
• Richeldi, L., Sorrentino, R., Saltini, C. (1993). HLA-DPB1 glutamate 69: a genetic marker of beryllium disease. Science, 262,242-244.[Abstract/Free Full Text]
• Ross, A. G., Sleigh, A. C., Li, Y. S., Williams, G. M., Li, Y., Waine, G. J., et al. (1998). Epidemiological identification of Chinese individuals putatively susceptible or insusceptible to Schistosoma japonicum: A prelude to immunogenetic study of human resistance to Asian schistosomiasis. Annals of Tropical Medicine and Parasitology, 92,765-774.[Medline]
• Sayed-Tabatabaei, F. A., Schut, A. F. C., Hofman, A., Bertoli-Avella, A. M., Vergeer, J., Witteman, J. C. M., et al. (2004). A study of gene–environment interaction on the gene for angiotensin converting enzyme: a combined functional and population based approach. Journal of Medical Genetics, 41,99-103.[Abstract/Free Full Text]
• Scarlett, W. L. (2003). Ultraviolet radiation: sun exposure, tanning beds, and vitamin D levels. What you need to know and how to decrease the risk of skin cancer. Journal of the American Osteopathic Association, 103,371-375.
• Selinger-Leneman, H., Genin, E., Norris, J. M., Khlat, M. (2003). Does accounting for gene–environment (G x E) interaction increase the power to detect the effect of a gene in a multifactorial disease? Genetic Epidemiology, 24,200-207.[Medline]
• Semenza, J. C., Ziogas, A., Largent, J., Peel, D., Anton-Culver, H. (2001). Gene–environment interactions in renal cell carcinoma. American Journal of Epidemiology, 153,851-859.[Abstract/Free Full Text]
• Sharp, R. R., de Serres, F., Newman, L., Sandhaus, R. A., Walsh, J. W., Hood, E. (2003). Environmental, occupations, and genetic risk factors for.
• Shostak, S. (2003). Locating gene–environment interaction: At the intersections of genetics and public health. Social Science and Medicine, 56,2327-2342.
• Silberg, J., Rutter, M., D'Onofrio, B., Eaves, L. J. (2003). Genetic and environmental risk factors in adolescent substance use. Journal of Child Psychology and Psychiatry and Allied Disciplines, 44,664-676.
• Speakman, J. R., Talbot, D. A., Selman, C., Snart, S., McLaren, J. S., Redman, P., et al. (2004). Uncoupled and surviving: Individual mice with high metabolism have greater mitochondrial uncoupling and live longer. Aging Cell, 3,87-95.[Medline]
• Taylor, J. A., Umbach, D. M., Stephens, E., Castranio, T., Paulson, D., Robertson, C., et al. (1998). The role of N-acetylation polymorphisms in smoking-associated cancer: Evidence of a gene–gene–exposure interaction. Cancer Research, 58,3603-3610.[Abstract/Free Full Text]
• Taylor, R. R., Mamotte, C. D., Fallon, K., van Bockxmeer, F. M. (1999). Elite athletes and the gene for angiotensin-converting enzyme. Journal of Applied Physiology, 87,1035-1037.[Abstract/Free Full Text]
• Turkheimer, E. (2002). Three laws of behavior genetics and what they mean. Current Directions in Psychological Science, 9,160-164.
• Tyndale, R. F. (2003). Genetics of alcohol and tobacco use in humans. Annals of Medicine, 35,94-121.[Medline]
• Von Ehrenstein, O. S., Von Mituis, E., Illi, S., Baumann, L., Bohm, O., Von Kries, R. (2000). Reduced risk of hay fever and asthma among children of farmer. Clinical and Experimental Allergy, 30,187-193.[Medline]
• Von Mutius, E., Braun-Fahrlander, C., Schierl, R., Riedler, J., Ehlermann, S., Maisch, S., et al. (2000). Exposure to endotoxin or other bacterial components might protect against the development of atopy. Clinical and Experimental Allergy, 30,1230-1234.[Medline]
• Wahlsten, D., Metten, P., Phillips, T. J., Boehm, S. L., II, Burkhart-Kasch, S., Dorow, J., et al. (2003). Different data from different labs: Lessons from studies of gene–environment interaction. Journal of Neurobiology, 54,283-311.[Medline]
• Weiss, E. P., Brown, M. D., Shuldiner, A. R., Hagberg, J. M. (2002). Fatty acid binding protein-2 gene variants and insulin resistance: Gene and gene–environment interaction effects. Physiological Genomics, 10,145-157.[Abstract/Free Full Text]
• Williams, W. J. (1994). Beryllium disease. In W. R. Parkes (Ed.), Occupational lung disorders (pp. 571–592). Oxford, UK: Butterworth–Heinemann.
• Williams, W. J. (1996). United Kingdom beryllium registry: Mortality and autopsy study. Environmental Health Perspectives, 104S,(suppl.), 949-951.
• Witter, J., Gauderman, W., Thomas, D. (1999). Asymptotic bias and efficiency in case–control studies of candidate genes and gene–environment interactions: Basic family designs. American Journal of Epidemiology, 149,693-705.[Abstract/Free Full Text]
• Woods, D. R., Humphries, S. E., Montgomery, H. E. (2000). The ACE I/D polymorphism and human physical performance. TEM, 11,416-420.[Medline]
• Yadav, R. S., Bidinger, F. R., Hash, C. T., Yadav, Y. P., Yadav, O. P., Bhatnagar, S. K., et al. (2003). Mapping and characterization of QTL x E interactions for traits determining grain and stover yield in pearl millet. Theoretical and Applied Genetics, 106,512-520.
• Zinn-Justin, A., Marquet, S., Hillaire, D., Dessein, A., Abel, L. (2001). Genome search for additional human loci controlling infection levels by Schistosoma mansoni. American Journal of Tropical Medicine and Hygiene, 65,754-758.[Abstract]

This article has been cited by other articles:

				J. Miranda, T. G. McGuire, D. R. Williams, and P. Wang Mental Health in the Context of Health Disparities Am J Psychiatry, September 1, 2008; 165(9): 1102 - 1108. [Full Text] [PDF]

This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation