There is growing scientific and public interest in the use of genetic information in healthcare, reflected in the availability of genetic tests marketed directly to the consumer, although these have primarily focused on physical disease to date. Gartner and colleagues conclude that, for the hypothetical case of a genetic test for susceptibility to cigarette smoking, family history data outperform genetic tests [1]. This is perhaps not surprising, for the simple reason that family history is an aggregate marker of both genetic and environmental influences on smoking behaviour, while genetic tests are necessarily only a marker of the former, and likely to be restricted to only some of the variants contributing to genetic influence.

The primary difficulty with genetic association studies has been the lack of convincing evidence for any particular association, and the endemic problem of non-replication, driven in large part by the small effect sizes typical of single gene effects [2]. We have learned that early studies often show great promise, but that this all too often turns to disappointment [3]. It is generally accepted that very large sample sizes will be required to detect genetic effects robustly, with odds ratios of around 1.2 for single gene effects likely to be the most we can hope for, and recent promising results support this [4].

The authors suggest that genetic testing may have clinical value in certain settings, for example in ‘matching smokers to the most effective cessation treatments’. This is potentially true, but even here there is reason to be cautious. While a (modest) number of studies now exist suggesting that specific genetic variants may modify treatment response to pharmacotherapies such as nicotine replacement therapy and bupropion (no published pharmacogenetic studies of varenicline yet exist), many of the same conclusions reached by Gartner and colleagues are likely to hold here also.

The primary reason for this, and the main difficulty with any clinical application of genetic tests, is that one must either have a test for a large genetic effect, or for a variant that is common in the population (or, ideally, both). In practice, it is becoming increasingly clear that common alleles confer small effects, while large effects are generally the preserve of rare alleles. Combining multiple tests within a single panel will not help, because the likelihood of possessing the exact combination of risk alleles which confers a substantial increased risk will become vanishingly small extremely quickly.

Therefore, most individuals will carry some risk alleles and some protective alleles, so that any net effect will be modest. This is illustrated nicely by the simulations of Gartner and colleagues, where the single most important factor determining the utility of a test is the magnitude of the genetic effects at the loci which it tests. From what little we currently know about the moderating role of genotype in response to pharmacotherapy for smoking cessation, effect sizes here are not likely to be substantially greater than for other smoking behaviour phenotypes [5,6], so that the same conclusions are likely to hold for pharmacogenetics applications.

One possible exception to this is in the identification of individuals at risk of serious adverse events to particular pharmacotherapies. However, few pharmacogenetic studies of smoking cessation have investigated adverse events, focusing instead on cessation outcomes [5,6]. The side effect profiles of most first-line treatments for smoking cessation are relatively benign, although there have been concerns about both bupropion and varenicline. Here pharmacogenetics may have some value in identifying individuals at particular risk of experiencing adverse events. Unfortunately, sufficient relevant data do not currently exist, and it remains unclear whether such an approach would ever be cost-effective, given the general efficacy of existing treatments [7] and the low frequency of adverse events.

None of this is to suggest that the genetic dissection of tobacco dependence and smoking cessation phenotypes is without value. Improved understanding of the neurobiological basis for these behaviours may provide novel treatment targets and facilitate drug discovery and development. Nevertheless, in the context of clinical practice, family history and phenotype data are likely to outperform genetic tests for the foreseeable future, and do not raise the same logistic, educational and ethical issues that genetic tests might [5,6]. Given that (moderately) effective interventions for smoking cessation exist, resources may be better used by encouraging more tobacco users to make use of existing treatments, and in the discovery of novel treatment targets.

Declaration of interest