|Home | About | Journals | Submit | Contact Us | Français|
P. Vineis. Imperial College London
There are several aspects that are still unresolved in the study of gene–environment interactions (GEI), particularly in the occupational setting.
Errors of measurementMeasurements of genes (genotyping) and of environmental agents are usually prone to error to different degrees. Genotyping is in general more accurate than the majority of methods used to measure environmental exposures. This implies a lower degree of classification errors that in turn means an easier identification of associations with disease. Unless the size of the studies is calculated to achieve an adequate power of detection for associations of disease with the variables most affected by classification errors, that is, the environmental ones (diet, job titles, pollutants, etc), it can be safely predicted that such studies will come up with a number of genetic associations but very few credible environmental associations with disease. In addition, as the majority of genetic polymorphisms are believed to act through biological interaction with environmental agents, it may also become difficult to make sense of genetic observations if the environmental component is substantially misclassified.
Testing of multiple hypothesesWhile accurate genotyping measurement makes easier the detection of gene–disease associations, it also contributes to enhance the chance of false positive results if, as is more and more the case, associations between each of hundred or thousands of genetic variants and disease phenotype are tested. In the light of reasonable assumptions on the a priori probability of an association being real and of the power of the study, values of p of the order of 10−7 have been worked out as a significance threshold to separate real from false associations (however associations with less extreme p values may still be worth considering for replication studies, particularly when supported by information on biological mechanism).
InteractionsFurther problems arise when the joint effects of genes and environment are considered, namely gene–environment interactions. This label – GEI – has become current for any study involving genotyping as well as some measures of environmental factors. The term interaction lends itself to confusion as it has several meanings: biological, statistical and public health.
Mendelian randomisationNot only exposures to environmental agents are imperfectly measured, but the relation of environmental exposures to disease is prone to confounding. Mendelian randomisation (MR) is a way to exploit genetic testing in epidemiology to overcome some of the limitations of observational epidemiology. MR has been suggested as a way to overcome confounding by exploiting the random allocation of alleles from parents to offspring. The association between a gene variant and a disease is not subject to the confounding by behavioural or socioeconomic factors that has clearly led to misleading findings in conventional observational epidemiological studies. Nor do reverse causation or other biases inherent in observational research apply to studies based on Mendelian randomisation. The problems arise, however, in case of a lack of association between the gene and the disease. The lack of an association between gene and disease could be interpreted in several different ways: (a) the relationship between exposure and disease was in fact confounded or biased; (b) the phenotype is only partially related to the studied allele, and other haplotypes/gene variants are also involved; (c) penetrance/expression of the gene depends on circumstances, including other exposures; (d) the power of the study could be too limited to show an association. Therefore, it is likely that there is an asymmetry between positive and negative findings from MR.
I wish to thank Rodolfo Saracci for thoughtful discussions
F. Kauffmann. Inserm U780 ‐ Epidemiology and Biostatistics and Université Paris‐Sud
Asthma is a complex disease with environmental and genetic determinants, with a varying age of onset. Occupational exposures cause around 15% of adult asthma. Around 250 specific occupational exposures are concerned. The precise nature of exposures inducing asthma in high risk occupations, such as nurses and cleaners, is currently being studied. Asthma runs in families and various strategies are used to assess the role of potential candidate genes and to search for new genes. The role of at least eight candidate genes has been replicated in at least five studies, including IL13, IL4 and HLA‐DRB1. Genome‐wide screens have identified six asthma and/or atopy susceptibility genes by positional cloning (ADAM33, PHF11, DPP10, GPRA, HLA‐G, CYFIP2). Few linkage studies have considered interaction with environmental factors in asthma and none has looked at occupational hazards. Whereas linkage studies need families, genome‐wide association studies, that is, the search for association of a phenotype, such as asthma, with 100000, 300000 or even 1 million SNPs (single nucleotide polymorphisms) spread out over the genome, can be implemented by a classical case‐control design. Recently a genome‐wide association study (Moffatt et al. Nature 2007;448(7152):470–3; Gabriel program) has evidenced a potential gene (ORMDL3) of unclear function for childhood asthma. The development of genome‐wide approaches for associations will provide new challenges and offer new opportunities.
The evidence regarding gene–environment interactions regarding occupational hazards comes from classical association studies and it remains to be established whether wider approaches, that is, not based on candidate interactions, may be efficient. Candidate interactions may be defined as the explicitation of a hypothesis regarding a particular environmental factor in relation to the known function (or potentially locus) of a specific gene (eg, endotoxin for CD14), or regarding a particular gene in relation to the known role of a specific environmental factor (such as smoking or coal dust). The first type of study could be labelled gene by environment interaction (the most usual term), whereas the second type could be labelled environment by gene interaction to emphasise that the driving partner of the research was the environmental risk factor.
Few gene by environment interactions for asthma are demonstrated (see Castro‐Giner et al. Occup Environ Med 2006;63(11):776–86, 761. for review). Interactions of HLA II with isocyanates is established. Pathways involving both immunologic and non‐immunologic mechanisms are candidate. Altered detoxification of reactive oxygen species concerns occupational hazards and the role of GST and NAT genes have been shown for isocyanates. Interaction with HLA II has also been observed/suggested for acid anhydrides, red cedar, laboratory animal allergens.
Challenges raised by two attempts to search for gene by occupational hazards for asthma interaction in the context of candidate interactions will be illustrated based on data from the epidemiological study on the Genetics and Environment of Asthma (EGEA). The first approach will consider around 1300 SNPs in the HLA region. The second will consider the potential interaction with a gene relevant for irritant‐induced asthma.
Large samples and replications are essential for adequate power and to overcome the multiple comparison issue. International collaboration, which has already started for establishing genetic determinants of occupational hazards through the European Gabriel program should be developed.
Supported by ANR‐SEST ANR 05‐SEST‐020‐02/05‐9‐97, Ga2len (Global Allergy and Asthma European network, FOOD‐CT‐2004‐506378, NOE FP6) and Gabriel grant 01896, IP, FP6 LSH‐2004‐1.2.5‐1.