While our data on their own are inconclusive with respect to any association between variation in the CHRNA5-A3-B4
gene cluster and short-term smoking cessation in treatment-seeking smokers, it is notable that the magnitude of the effect we observed is consistent with recent evidence from a cohort study of pregnant women (Freathy et al., 2009
), a community study of pregnant women (Thorgeirsson & Stefansson, 2010
), and a clinical trial of treatment-seeking men and women (Baker et al., 2009
). Taken together, these data are suggestive of a weak effect of genetic variation in this region on smoking cessation. Given the lack of a genotype × treatment interaction in the Patch II sample, which employed a placebo-controlled design, it is unlikely that any effect operates only among those receiving NRT. Moreover, adjustment for cigarette consumption and tobacco dependence did not alter these results substantially, suggesting that any effect on cessation may not be mediated through dependence or heaviness of smoking. The observed per allele effect is equivalent to ~0.5% of phenotypic variance in quit success, which is clearly modest (although consistent with expectation for common variants and complex behavioral phenotypes). However, any effect on cessation may be cumulative over the multiple attempts to stop smoking any one individual might make.
Previous studies have observed that this variant and other nearby variants on chromosome 15 are associated with heaviness of smoking among current smokers but not smoking status (i.e., ever-smoker vs. never-smoker or current smoker vs. ex-smoker), indicating no association with either smoking initiation or cessation. However, large-scale case–control studies typically rely on self-report point prevalence smoking status and typically lack biochemical verification of current smoking status. This will reduce the ability to detect these associations. The data in the present study are derived from the prospective assessment of treatment-seeking smokers, where smoking status was verified with biochemical assessment, which may allow for a more precise measurement of smoking cessation.
It is also worth noting that the participants in both samples were selected to be heavy smokers actively seeking treatment to stop smoking. There may be important differences between this population and the wider population of smokers who attempt to stop smoking, most of whom do so spontaneously and without behavioral or pharmacological support. However, together with previous evidence (Baker et al., 2009
; Freathy et al., 2009
; Thorgeirsson & Stefansson, 2010
), there is now evidence from five independent samples indicating that the rs1051730 variant may be weakly associated with the short-term ability to stop smoking. This may be independent of both heaviness of smoking and tobacco dependence and whether or not the individual is using NRT. While we provide statistical evidence of association with smoking cessation, the clinical significance of this association (if genuine) is modest, given the small effect size and the lack of a specific effect among individuals receiving NRT. Nevertheless, it is larger than would be predicted from an effect on heaviness of smoking equivalent to 1 cigarette/day, which suggests it may operate via some unmeasured attribute related to cigarette smoking.
The mechanism by which genetic variation in the CHRNA5-A3-B4
gene cluster influences heaviness of smoking remains unclear, and it is therefore difficult to speculate on how it influences smoking cessation. One possibility is that individuals with one or more copies of the T allele may be more resistant to the aversive side effects of nicotine, as shown in recent studies of nAChR α5 knockout mice (Jackson et al., 2010
), and may therefore be able to tolerate higher doses of NRT. Conceivably, nicotine derived from combinations of smoking and NRT, or different combinations of NRTs, might thus provide different dose–response relationships for individuals with nAChRs encoded by haplotypes with these variants at this locus. However, NRT is generally well tolerated at high doses, and in the studies reported here, there was no variation in dose offered. Moreover, this explanation is also not consistent with our failure to observe a differential effect in the placebo and active arms of the Patch II study and the effect of rs1051730 reported in other studies (Baker et al., 2009
; Freathy et al., 2009
; Thorgeirsson & Stefansson, 2010
). A second possibility is that it might be associated with cessation because it influences heaviness of smoking or tobacco dependence. In particular, it may in fact be the case that dependence mediates any association between this variant and likelihood of smoking cessation success, but error variance in the measurement of dependence or heaviness of smoking may obscure this. In addition to measurement error, there is considerable interindividual variability in number of puffs taken per cigarette, depth of inhalation, and other aspects of smoking topography, meaning that cigarettes per day is a relatively poor indicator of typical nicotine consumption. This would be supported if adjusting for precessation cotinine levels (which more accurately assay nicotine consumption) attenuates the association of genotype with cessation success. Both of these possibilities should be investigated in future laboratory and clinical studies.
There are several limitations to this study, which should be considered when interpreting these results. First, the sample size was small by current standards of genetic epidemiology. However, this is perhaps offset to some degree by the greater precision afforded by the prospective assessment of biochemically verified smoking status (Phillips & Smith, 1993
). Second, the observed association only achieved statistical significance in one sample and would not survive Bonferonni correction for multiple tests across three outcome measures. However, the effect size estimates were comparable in both samples, and the overall effect on short-term abstinence was significant in the meta-analysis, with no evidence of between-study heterogeneity. Third, we did not genotype the rs16966968 variant, which has been proposed as the putative functional variant within this region, or other nearby variants identified in recent genome-wide analyses. However, the strong prior evidence for association with heaviness of smoking suggests that the rs1051730 is an acceptable marker at this locus. Fourth, the imprecision in self-report assessments of heaviness of smoking and tobacco dependence may have reduced our ability to determine whether the rs1051730 variant is simply acting as a marker for these measures compared with more direct measures of use and exposure, such as precessation levels of cotinine. For example, while we observed the predicted association between rs1051730 and cigarettes per day, we only observed an association with tobacco dependence in the Patch in Practice sample and not the Patch II sample, most likely due to differences in the measure of dependence used (Fagerström Test for Nicotine Dependence vs. Horn–Russell Tobacco Scale, respectively). This possibility can be investigated in future laboratory and clinical studies.
In conclusion, our data, in conjunction with previous studies, provide some support for a small effect of the rs1051730 on short-term smoking cessation, independently of heaviness of smoking, tobacco dependence, or the use of NRT. Future studies attempting to replicate this finding should attempt to do so in large samples of prospectively assessed smokers whose smoking status can be confirmed biochemically.