TOBACCO

doi: 10.1111/1753-6405.12380

A Commentary re Quitting activity and tobacco brand Switching: findings from the ITC-4 Country Survey Brent Caldwell Department of Medicine, University of Otago, New Zealand

C

owie and colleagues have reported the results of their analysis of quitting activity and brand switching in data collected from the International Tobacco Control Four Country Survey (ITC-4) to determine whether smokers’ interest in quitting, and their quitting behaviour, influences their decision to switch brands, or whether the reverse is true.1 The ideal method to test the causal direction of associations between switching brands, making an attempt to quit, and quit success would be to conduct a randomised controlled trial (RCT). RCTs ensure factors that potentially confound the relationship between the putative cause and the putative outcome are removed from the equation, since they are (on average) equally distributed between people who are exposed to the putative cause and those who are not. Furthermore, treatment allocation in RCTs is controlled by the researchers, thereby removing treatment selection bias. However, use of a randomised controlled trial to assess the role of switching brands is fraught with difficulties. Would it be ethical to randomise people to either switch brands or continue to smoke their usual brand? How could subjects’ compliance with brand switching be encouraged or monitored? How could trial participants or researchers be blinded to the allocation assignment? In situations where random allocation of trial participants to an intervention is not possible, researchers are limited to the examination of associations in data from observational studies, such as surveys. Sir Austin Bradford Hill proposed a checklist to assist in assessing the causal nature of observed associations.2 The ITC-4 was designed as a quasi-experimental study, with features that mimic aspects of RCTs to allow it to meet many of Bradford Hill’s criteria for judging causality. The ITC-4 aims to recruit

114

participants who are representative of the whole population of smokers and former smokers (aged 18 years or older who have smoked more than 100 cigarettes in their lifetimes and who have smoked at least once in the past 30 days). It does this by using random-digit dialling to recruit and screen people for eligibility, and by enrolling a large number of respondents (n=2,000 per country).3 Furthermore, the ITC-4 is a longitudinal survey, prospectively following the same cohort of smokers over time, enabling it meet Bradford Hill’s temporality and quasi-experimental criteria. Note that subjects who are lost to follow-up are replenished using the same sampling frame, therefore only a sub-sample of the data is truly longitudinal. Cowie and colleagues used Generalised Estimating Equations (GEE) to test for the temporal direction in the associations between brand switching, quit interest and quit behaviour. Because ITC-4 was conducted in four countries, Bradford Hill’s criterion of consistency can be tested by incorporating a country term in the GEE. A dose-response relationship could be established between quit interest, quit attempts and brand switching if ITC4 recorded these variables on a finely graded scale; for example, by measuring quit interest on a 7-point Likert scale, measuring the number of quit attempts per wave, and the duration of each period of abstinence. However, in the analyses of Cowie and colleagues, quit interest was a dichotomous variable composed of those who intended to quit in the next month or next six months compared to those who intended to quit beyond six months or had no intention to quit. Similarly, quit attempts was dichotomised by counting those who had made any quit attempt since the previous survey wave (or in the last year

for new recruits) compared to those who had not attempted to quit. The duration of quitting was similarly divided into those whose longest period of abstinence was less than one month compared to those who abstained for more than one month. There are plausible mechanisms by which switching brands might increase quitting and quit success, and why smokers who fail in a quit attempt might switch brands. If smokers switch to a different brand, it is likely that they will not enjoy it as much as their usual brand, and therefore smoking will be less rewarding and easier to give up than if they had continued to smoke their usual brand. Much of the reward from smoking comes from secondary conditioned reinforcement of the sensory characteristics of smoking. When smokers switch brands, they will initially only experience the primary reinforcement of nicotine, and will not gain reinforcement from the sensory effects (unless they are very similar to their usual brand), until they have smoked the new brand for some time and had the opportunity for its new sensory effects to be paired with the reward of nicotine. Therefore, smokers may find it easier to quit shortly after switching brands than if they continued to smoke their usual brand. On the other hand, delaying a quit attempt is known to reduce the chances of it being successful, so if switching brands is merely a delaying tactic, it is likely to lessen the probability of achieving abstinence. Changes in the relative price of cigarette brands may be an important confounder of the relationship between quit intentions, attempting to quit and quit success, because the relative price of cigarette brands influences all three of these variables. Cowie and colleagues argue that price did not confound their analyses because brand switching did not increase in response to the tax increase in 2010 in Australia. However, unless the tax increase resulted in an increase in the price disparity between brands, it is unlikely to have motivated brand switching, therefore the absence of brand switching after the tax increase is not evidence of the abstinence of a causal association between price and brand switching. It is certainly possible that there were greater changes in the price disparity between brands at timepoints during the survey period that were unrelated to the tax increase, compared to changes in price disparity between brands that occurred around the time of this tax increase, and this has not been captured or

Australian and New Zealand Journal of Public Health © 2015 Public Health Association of Australia

2015 vol. 39 no. 2

Tobacco

Quitting activity and tobacco brand switching

References

adjusted for in the analysis, possibly because changes in the price of survey respondents’ usual brand of cigarette were not recorded in ITC-4. Cowie and colleagues attempted to minimise the effects of selection bias and confounding by using GEE to adjust for three potential confounding factors (age, sex and income). Ideally, these models would have adjusted for additional factors such as cigarettes smoked per day and nicotine dependence. Using appropriate statistical techniques to adjust for a sufficiently large number of confounding factors in observational data can dramatically alter the results. This was illustrated in a trial that showed that use of electronic cigarettes was significantly associated with lower abstinence; however, the association was no longer significant when entropy balancing was employed to adjust for concurrent use of other cessation aids.4 Cowie and colleagues have attempted the very difficult task of using an observational study to test for causality. Their analyses appear to show that smokers who aim to quit in the near future are more likely to switch brands than those who are less interested in quitting in the short-term. However, smokers who plan to quit in the near future tend not to switch brands prior to quitting but rather do so only after failing to quit. Therefore, switching brands is a sign that smokers are interested in quitting, but need more help to quit. Smokers need to be informed that switching brands will not necessarily help them to quit, and that if they are having difficulty quitting it may be more effective for them to use nicotine replacement therapy than to change their brand of cigarettes.

2015 vol. 39 no. 2

1. Cowie A, Swift E, Partos T, Borland R. Quitting activity and tobacco brand Switching: findings from the ITC-4 Country Survey. Aust N Z J Public Health. 2015; 39(2): 109-13 2. Hill AB. The Environment and Disease: Association or Causation? Proceedings of the Royal Society of Medicine 1965;58(5):295-300. 3. Thompson ME, Fong GT, Hammond D, Boudreau C, Driezen P, Hyland A, Borland R, Cummings KM, Hastings GB, Siahpush M, Mackintosh AM, Laux FL. Methods of the International Tobacco Control (ITC) Four Country Survey. Tobacco Control 2006;15 Suppl 3:iii12-8. 4. Pearson JL, Stanton CA, Cha S, Niaura RS, Luta G, Graham AL. E-cigarettes and smoking cessation: Insights and cautions from a secondary analysis of data from a study of online treatment-seeking smokers. Nicotine & Tobacco Research 2014.

Correspondence to: Dr Brent Caldwell, Department of Medicine, University of Otago, PO Box 7343, Wellington 6021, New Zealand; e-mail: [email protected]

doi: 10.1111/1753-6405.12383

Unpicking causal possibilities: Authors’ response to Caldwell Genevieve A. Cowie,1 Elena Swift,2 Timea Partos,2 Ron Borland2 1. Department of Epidemiology & Preventive Medicine, School of Public Health & Preventive Medicine, Monash University, Victoria 2. Cancer Council Victoria

We agree with Caldwell, both that RCTs are the only fail-safe method for showing causal associations and that such an approach was not an option here. However, we add one other important reason. An RCT can only test possible causal factors that are under the researcher’s control. Brand switching is a

personal choice. Thus, even an RCT of forced brand switching would not provide direct evidence of the causal impact of a chosen switch: the two are very different. This is a problem that effectively precludes adequately controlled RCTs for many behavioural interventions: we can offer them to people, but can’t expect pre-commitment without first describing the different requirements of the various experimental conditions, thus creating different expectancies. In the more focussed comments, Caldwell argues that relative price might be a confounder for the asymmetrical relationships we found. However, price is an unlikely confounder; if switching to a cheaper brand was an alternative to quitting, we might expect switching to be associated with less subsequent quit interest. However we found the reverse asymmetry - the interest to switching association was stronger. Controlling for dependence made no difference to the findings. We believe our modest claims as to which causal direction was most likely are reasonable, as apparently does Caldwell, given his take out message is essentially the same as ours. Looking for asymmetrical relationships is a useful way of testing potential causal models where RCT evidence is not available, or in this case not achievable. Correspondence to: Dr Ron Borland, Cancer Council Victoria, 615 St Kilda Rd, Melbourne, VIC 3004; e-mail: [email protected]

Australian and New Zealand Journal of Public Health © 2015 Public Health Association of Australia

115

A commentary re.: Quitting activity and tobacco brand switching: findings from the ITC-4 Country Survey.

A commentary re.: Quitting activity and tobacco brand switching: findings from the ITC-4 Country Survey. - PDF Download Free
45KB Sizes 6 Downloads 8 Views