Commentaries on Mathews et al. (2012)
Version of Record online: 20 NOV 2012
© 2012 The Authors. Addiction © 2012 Society for the Study of Addiction
Volume 107, Issue 12, pages 2077–2079, December 2012
How to Cite
Ware, J. and Munafò, M. R. (2012), Testing times. Addiction, 107: 2077–2079. doi: 10.1111/j.1360-0443.2012.03957.x
- Issue online: 20 NOV 2012
- Version of Record online: 20 NOV 2012
- British Heart Foundation
- Cancer Research UK
- Economic and Social Research Council
- Medical Research Council, and the National Institute for Health Research
- Genetic testing;
- personalized medicine;
After almost two decades of molecular genetic research, one clear lesson has been learned: the effects of individual genetic loci on complex behavioral phenotypes, such as addictive behaviors, are extremely small, typically accounting for (often much) less than 1% of phenotypic variance . One consequence of this is that robust associations have proved difficult to identify and have only begun to emerge through large, consortium-based studies that combine data from several thousands of participants. Even so, only a handful of genes have been shown to be unequivocally associated with addiction phenotypes and their individual contribution is extremely modest. Mathews and colleagues  argue that the commercialization of direct-to-consumer genetic tests for addiction susceptibility is premature. Here, we expand on this argument and question whether such tests will ever be informative at an individual level.
One proposed solution to the problem of the low proportion of variance explained by individual genetic variants is to develop polygene risk scores which combine a large number of variants in order to predict a more substantial proportion of phenotypic variance . This approach may eventually prove informative at the group level, but this does not mean that it will necessarily be informative at the individual level. A major difficulty with epidemiology (including genetic epidemiology) is that individual outcomes are likely to be influenced, to a large extent, by essentially random factors operating at several levels, from the cellular to the biographical . These stochastic events are averaged out at the group level, but can dominate at the individual level, meaning that it is essentially impossible to predict with any accuracy who will and will not develop disease. Even personal genome sequencing may only yield limited predictive capacity of uncertain clinical value .
Notwithstanding questions over whether genetic tests will ever be truly informative with respect to individual risks (with the exception of rare Mendelian diseases), there are also reasons to be concerned about the quality of currently-available tests. A striking example of this was provided by Ng and colleagues , who submitted their own DNA to two commercial genetic testing companies and received strikingly disparate results regarding their personal risk of a variety of disease outcomes. For seven diseases, the predictions made by the two companies’ tests across five individuals agreed in less than 50% of cases. This, in part, reflects genuine uncertainty and heterogeneity in the field regarding which associations may be considered to be proven, leading to the inclusion of different ‘risk’ variants in different direct-to-consumer tests. This problem is compounded by a lack of transparency on the part of the companies regarding the variants which are included within their tests.
There is, perhaps, more cause for optimism for the role of genetic testing in the context of pharmacogenetics. Here, genetic information is used to identify those who are likely to respond particularly well or particularly poorly to a particular pharmacotherapy. Some clear successes have been identified in other fields  and there is a growing literature on the association of genetic variants with response to treatments for addictive disorders . This is sometimes described as personalized medicine, but is more accurately characterized as stratified medicine (in that it does not apply to individual patients, but to sub-groups of patients). However, this approach is most likely to be successful in identifying those who may suffer an adverse reaction to their medication, for example due to reduced metabolism of the drug. In cases where adverse drug reactions are rare, it is not clear that genetic tailoring will more cost-effective than the current ‘one-size-fits-all’ approach, even if variants reliably associated with treatment response are identified .
Even if we accumulate the evidence base to develop stratified approaches to addiction treatments on the basis of genotype, and can be confident that these approaches will be cost-effective, more work will need to be done to establish the potential impact of these tests before translation into clinical practice. There are practical and ethical implications relating to the use of genetic information in routine clinical practice, which have been described in detail elsewhere . Whilst the effects of providing treatment tailored according to genotype on treatment compliance and willingness to re-engage with treatment in the future are beginning to be investigated [11, 12], this research is at a very early stage. Given the poor predictive power of genetic tests at the individual level, the apparent inconsistency in results provided by companies providing direct-to-consumer tests and the lack of evidence regarding the wider effects of providing genetic information to individuals regarding disease risk, the commercialization of such tests must be considered premature.
JW and MRM are members of the UK Centre for Tobacco Control Studies, a UKCRC Public Health Research: Centre of Excellence. Funding from British Heart Foundation, Cancer Research UK, Economic and Social Research Council, Medical Research Council, and the National Institute for Health Research, under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged.
Declaration of interests
MRM has received grant funding from Pfizer Ltd.