We found a small, statistically non-significant effect on 28-day adherence of communicating a tailoring of oral dose of NRT to genotype rather than phenotype. This represented 1.5 days, or 1.4 mg per day difference between groups, in mean NRT consumed over a 28 day period. No adverse effects on motivation in those failing to quit were detected. Neither trial hypothesis was supported.
Interpretation of Study Results
Taken together with the secondary outcomes, our results are consistent with the intervention having a small positive effect upon adherence amongst these participants. The evidence supporting this effect is provided by the 95% confidence interval, with the upper limit of 10.8% being consistent with there being a small positive effect of the intervention on the primary outcome. The confidence interval is also consistent with there being a zero effect of the intervention, but it is inconsistent with anything other than a minute negative effect on smoking cessation. Further evidence to support the intervention having a small effect upon adherence is provided by the secondary outcome of NRT adherence at 7 days, which is statistically significant, and a the biochemically validated prolonged abstinence at 6 months. The higher abstinence observed in the genotype group may in part have been due to the higher proportion of participants in that group who initiated at least some use of NRT as part of their quit attempt (see ) as well as the higher proportion who reported using NRT beyond the 28 day treatment period. These findings suggest that the intervention may have had an effect both on initiation of the use of NRT and treatment persistence. The factors that might help interpret these observations will be explored in a separate paper.
This first empirical test of the behavioural impact of pharmacogenomics reduces the uncertainty around the possible size of impact. The effect on adherence, if there is one, is likely to be smaller than the two day difference in adherence at one month on which we powered the study.
Overall, 11% of participants (68/633) were abstinent at six months. The difference in abstinence between groups at six months is larger than would be expected from the small observed difference in adherence to NRT at 28 days and the negligible difference in short-term abstinence. If the impact of the intervention on abstinence is not explained by NRT adherence over 28 days, there are a number of other explanations. First, this may be a chance finding. Second, this may be due to bias in the conduct of the study; however randomisation was effective, there was no unequal attrition across groups and any bias might be expected to be apparent during behavioural support and not after it. Third, this may be a valid finding, not mediated by adherence to NRT during the first 28 days of a quit attempt. Whilst we did not assess adherence after 28 days in detail, during the six month telephone follow-up we asked participants whether they had consumed NRT beyond the initial treatment period, with more in the genotype arm reporting such use than those in the phenotype arm. This finding should, however, be considered with caution. The measure used relies on retrospective long-term recall and the analysis was not specified prior to conducting the study. The observed effect on long-term abstinence may also reflect impact of the intervention on other smoking-related behaviours which we did not assess. Further analyses of these data, modelling causal effects of psychological and other variables collected during the trial, may shed some light on this.
We found no effect on continuing intention to quit smoking from informing participants that their oral NRT dose was tailored to genotype rather than phenotype. In particular people in the genotype arm who were not abstinent at six months were as likely to plan a further quit attempt as those in the phenotype arm. Despite the plausibility and frequently expressed concerns for such an effect, these findings are consistent with other, related evidence synthesised in a systematic review which provide no evidence to suggest that communicating personalised genetic risk information engenders feelings of fatalism 
. The results of the current study add to this by showing, we believe for the first time, that feelings of fatalism are also not engendered by failure to change behaviour following personalised genetic risk information.
Strengths and Limitations of the Study
The study has several strengths. First, it is novel, being the first to test the behavioural impact of pharmacogenomic tailoring of medication. Second, we used a robust design with evidence of success in balancing the two groups for key confounders including baseline smoking variables and genotype. Third, we powered for a plausible and clinically important difference. Fourth, we used objective measurement with biochemical verification of smoking status, which strengthens the interpretation of smoking abstinence.
There are also limitations. The primary endpoint was not smoking abstinence which would have required a larger trial. Adherence was chosen as the endpoint as the impact on smoking cessation of the intervention was predicted to work through adherence to NRT. The measure of adherence was based in part on self-report of medication use. Consumption was assessed using ‘pill’ counts as well as diary records and the practice nurse checked for discrepancies at each clinic visit and reconciled them with participants. We used more than one index of adherence to increase reliability. While the measures used lack the precision of some electronic devices, such devices cannot be used to measure length of exposure to patches or gums. Regarding the second hypothesis, we did not measure quitting behaviour beyond six months to assess whether motivation to make a quit attempt in those who were not abstinent reflected in actual behaviour. Motivation is, however, a reasonable indicator of subsequent behaviour 
The experimental design constrains the conclusions that can be reached. The design that was ultimately chosen does not clearly separate the relative influences of communication of the basis for prescribing and dose tailoring. The most attractive experimental design to investigate the effects of communication would have required us to use the same dosing algorithm in both arms but to randomise each arm to receive different information on the method used to tailor the dose. Clearly this would have necessitated deceiving participants and we regarded this as unacceptable. We considered seeking informed consent to this approach, but rejected it because of concerns that this might undermine trust in the trial as a whole. The design is limited in two further ways that may have produced a conservative estimate of the intervention effect. First, all participants received a prescription tailored to phenotype for transdermal NRT which may have served to dilute the impact of the intervention. The design was further limited by not including a group that received standard care only, or a group that received no behavioural treatment i.e. only tailored NRT prescriptions. Future studies might usefully assess the impact of communicating treatments tailored by genotype in different treatment contexts.
As expected, a minority of smokers (3.1%) who received an invitation to join a study of smoking cessation did so. In the UK around 5% of smokers a year use behavioural support and medication provided by NHS clinics. That our uptake is similar suggests that the results are likely to be generalisable to smokers trying to stop smoking with support and medication. It was our impression that people participated primarily to stop smoking rather than to participate in research.
Implications for Clinicians and Policy Makers
The results of the current study suggest that communicating to smokers that their NRT dose has been tailored by genotype is unlikely to cause harm. The effects on adherence were small at best, and the effects on abstinence and their mechanism, unclear. If the observed effects on smoking cessation at six months are replicated, however, genotype tailoring could contribute positively at a population level to smoking cessation interventions. Any possible contribution of genotype to tailored prescribing in smoking cessation should not detract from the increasing evidence for prescribing larger doses of NRT for more dependent smokers 
. Beyond smoking cessation, the current study is broadly consistent with effects of DNA-based risk communication observed across a range of health behaviours 
, suggesting that these effects are likely to be small or non-existent, and not have the aggregate behavioural impact that many anticipated.
Unanswered Questions and Future Research
Unanswered questions arising from these findings include: are the effects on smoking abstinence at six months real and how are they mediated, and, to what extent does the behavioural response to the communication of tailoring of medication to genotype vary by disease or service context?
This first empirical test of the behavioural impact of pharmacogenomics suggests that the impact on adherence to NRT may be small, at best. Further studies are warranted given the design constraints and the potential for behavioural impact with potential clinical significance. No adverse effect was detected on motivation to start another quit attempt amongst those who were not abstinent at six months. The unexpected higher rate of abstinence at six months in the genotype arm needs further investigation.