Ll etal itc 4 pos warnings and quitting-addiction_published


Published on

Published in: Health & Medicine
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Ll etal itc 4 pos warnings and quitting-addiction_published

  1. 1. RESEARCH REPORT doi:10.1111/j.1360-0443.2011.03668.xThe association between exposure to point-of-saleanti-smoking warnings and smokers’ interest inquitting and quit attempts: findings from theInternational Tobacco Control Four Country Survey add_3668 425..433Lin Li1, Ron Borland1, Hua-Hie Yong1, Sara C. Hitchman2, Melanie A. Wakefield3,Karin A. Kasza4 & Geoffrey T. Fong2,5VicHealth Centre for Tobacco Control, Cancer Council Victoria, Melbourne, Victoria, Australia,1 Department of Psychology, University of Waterloo, Waterloo,Ontario, Canada,2 Centre for Behavioural Research in Cancer, Cancer CouncilVictoria, Melbourne,Victoria,Australia,3 Division of Cancer Prevention and PopulationSciences, Roswell Park Cancer Institute, Buffalo, NY, USA4 and Ontario Institute for Cancer Research, Toronto, Ontario, Canada5ABSTRACTAims This study aimed to examine the associations between reported exposure to anti-smoking warnings at thepoint-of-sale (POS) and smokers’ interest in quitting and their subsequent quit attempts by comparing reactions inAustralia where warnings are prominent to smokers in other countries. Design A prospective multi-country cohortdesign was employed. Setting Australia, Canada, the United Kingdom and the United States. Participants A total of21 613 adult smokers who completed at least one of the seven waves (2002–08) of the International Tobacco ControlFour Country Survey were included in the analysis. Measurements Reported exposure to POS anti-smoking warningsand smokers’ interest in quitting at the same wave and quit attempts over the following year. Findings Compared tosmokers in Canada, the United Kingdom and the United States, Australian smokers reported higher levels of awarenessof POS anti-smoking warnings, and this difference was consistent over the study period. Over waves in Australia(but not in the other three countries) there was a significantly positive association between reported exposure toPOS anti-smoking warnings and interest in quitting [adjusted odds ratio = 1.139, 95% confidence interval (CI)1.039–1.249, P < 0.01] and prospective quit attempts (adjusted odds ratio = 1.216, 95% CI 1.114–1.327, P < 0.001)when controlling for demographics, smoking characteristics, overall salience of anti-smoking information and aware-ness of anti-smoking material from channels other than POS. Conclusions Point-of-sale health warnings abouttobacco are more prominent in Australia than the United Kingdom, the United States or Canada and appear to actas a prompt to quitting.Keywords Australia, Canada, health warnings, longitudinal research, point-of-sale, smoking cessation, survey,United Kingdom, United States.Correspondence to: Lin Li, VicHealth Centre for Tobacco Control, the Cancer Council Victoria, 100 Drummond Street, Carlton, Victoria 3053, Australia.E-mail: lin.li@cancervic.org.auSubmitted 10 April 2011; initial review completed 30 June 2011; final version accepted 22 September 2011INTRODUCTION tobacco industry in the United States has spent the vast majority of their promotional budget in the retail envi-Increasing numbers of countries have acted to restrict/ ronment over the past decade [9]. This ranges fromban tobacco advertising and promotion in traditional promotional material to payments for shelf space andmedia such as television, radio, print and on billboards incentive schemes for retailers.[1–5]. The tobacco industry has responded by expanding A number of jurisdictions have banned POS tobaccotheir marketing in areas where it is still allowed [6,7]. advertising, and some are now moving to prohibit thePoint-of-sale (POS) tobacco promotion has become an display of tobacco products—putting tobacco productsimportant marketing channel for the tobacco industry, as under the counter. Regardless of these restrictions, thereother marketing opportunities have been banned [8]. The are good reasons for having prominent POS anti-smoking© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433
  2. 2. 426 Lin Li et al. Figure 1 An example of Australian point- of-sale anti-smoking warnings (notes: this picture was taken in a supermarket in New South Wales during the study period; source: Simon Chapman of University of Sydney and Anne Jones of Action on Smoking and Health Australia)warnings. This is the time when smokers purchase ciga- Warnings at the POS are less prominent in the otherrettes, so it should be a critical time to have them thinking countries. Some provinces of Canada started to mandateabout the harms. However, there is little empirical evi- POS warnings from around 2004/05 [15,16], but mostdence to support such initiatives. do not have size requirements. Pictorial warnings at the During the study period (2002–08), Australia had POS were only introduced in some provinces of Canadanotably stronger POS anti-smoking warnings than did (e.g. British Columbia) from early 2008. In the Unitedthe other three countries in the study (Canada, the United Kingdom, the Tobacco Advertising and Promotion (PointKingdom and the United States). Australia has had of Sale) Regulations 2004 allowed only one A5-sizedprominent POS warnings for some time, but with some (14.8 ¥ 21 cm) poster advertising tobacco in store withvariation across states. For example, the most populous 30% of that area taken up with a health warning (e.g.state of New South Wales requires a health warning ‘smoking seriously harms you and others around you’,message (e.g. ‘smoking is addictive’) and a Quitline in force from December 2004) [17]. The United Statesnumber (131 848) that features: (i) black writing on a has no systematic requirements for POS anti-smokingwhite background; and (ii) between 50 and 100 cm wide information.and have an area not less than 2000 cm2 [10] (See We expected an overall higher level of awareness ofFig. 1). In the second most populous state of Victoria, POS warnings in Australia, and given the improvementssimilar warnings are mandated near the POS or at the in regulations regarding POS warnings over the studyshop entrance. However, they are smaller in Victoria [A4 period, increases in awareness in Australia, Canada andin size (21 ¥ 29.7 cm)] [11]. In Queensland, under the the United Kingdom over-time. We were also interested inTobacco and Other Smoking Products Amendment Act determining if reported awareness of POS warnings were2004, from December 2005 all the retail outlets are associated with higher quitting interest and prospectively,required to display mandatory POS anti-smoking signs with higher levels of quitting activity, and whether thisthat encourage quitting. Two other states (South Austra- varied by country (in this case a proxy for warninglia and Western Australia), that began the 2002–08 strength).study period with weaker signs, strengthened them to besimilar to those of New South Wales and Victoria duringthe study period. METHODS In Australia, POS anti-smoking warnings have been Data source and participantsintegrated with other health communication strategies,such as mandating health warnings on cigarette packets The data come from the first seven waves of the Interna-and ongoing mass media campaigns that all have empha- tional Tobacco Control Four Country Survey (the ITC-4sized the health harms of smoking and promoted the Survey), which has been running annually since 2002Quitline number, sometimes using the same base mate- in Australia, Canada, the United Kingdom and therials [12–14]. United States. A detailed description of the conceptual© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433
  3. 3. Association between POS warnings and quitting interest and attempts 427framework and methods of the ITC-4 Survey has been Measuresreported by Fong et al. and Thompson et al. [18,19], and The ITC-4 Survey included questions about smokingmore detail is available at http://www.itcproject.org. behaviour, cigarette brand information, quit history,Briefly, the ITC-4 Survey employs a prospective multi- exposure to anti-tobacco advertising, psychosocial mea-country cohort design and involves annual telephone sures and demographics.surveys of representative cohorts of adult smokers ineach of the four countries. The sampling is conducted viarandom-digit dialling. The sample size per country is Outcome measuresapproximately 2000. At each survey wave, the approxi-mately 30% lost to follow-up are replenished from the The key outcomes were interest in quitting and quitsame sampling frame. attempts. Interest in quitting was assessed by: ‘Are Participants were aged 18+ years, had smoked at least you planning to quit smoking?’. Response options were100 cigarettes life-time and had smoked at least once in ‘within the next month’, ‘within the next 6 months’,the past 30 days at the time of recruitment. A total of ‘some time in the future, beyond 6 months’, ‘not planning21 613 respondents who provided complete information to quit’ and ‘don’t know’. The first three categories werefor at least one of the seven waves (between 2002 and recoded to ‘having some interest in quitting’ and the2008) were included in the analysis. The sample size for remaining as ‘not having an interest’. Quit attempts wereeach country, the numbers of recruits in each wave/year, assessed at the next wave based on the answer to: ‘Sinceand their characteristics are summarized in Table 1. we last talked to you [in last survey date], have you madeTable 1 Sample characteristics, by country. Australia Canada United Kingdom United States Total n = 4806 n = 5265 n = 5251 n = 6291 n = 21 613Gender (% female) 53.5 53.7 55.8 55.1 54.6Identified minority group (%) 12.8 11.3 4.9 20.3 12.7Age at recruitment (years) (%) 18–24 14.2 12.8 8.5 11.7 11.7 25–39 35.6 30.3 31.3 26.3 30.6 40–54 34.0 36.5 33.8 36.6 35.3 55+ 16.1 20.4 26.4 25.5 22.4Education at recruitment (%)a Low 64.1 48.6 60.6 45.8 54.1 Moderate 21.9 36.4 25.1 38.3 30.9 High 13.9 14.8 13.5 15.9 14.6Income at recruitment (%)a Low 26.9 28.3 31.1 36.9 31.2 Moderate 32.7 34.4 31.5 33.3 33.0 High 34.0 28.4 27.6 22.7 27.8 No information 6.5 8.9 9.8 7.1 8.1Cigarettes per day at recruitment (%) 1–10 29.8 31.4 29.8 31.1 30.6 11–20 40.1 42.7 53.4 45.8 45.6 21–30 22.8 21.1 11.7 13.1 16.9 31+ 7.1 4.4 4.8 9.4 6.6No. of new recruits in each wave (n) Wave 1 (in 2002) 2305 2214 2401 2138 9058 Wave 2 (2003) 258 517 255 684 1714 Wave 3 (2004) 532 545 586 889 2552 Wave 4 (2005) 362 519 503 742 2126 Wave 5 (2006) 686 594 613 745 2638 Wave 6 (2007) 539 556 523 711 2329 Wave 7 (2008) 142 320 370 382 1196Percentages are based on unweighted data. For some variables the number of cases were fewer than the total, due to some ‘don’t know’ and ‘missing’cases. aFor the definition of each category please see the Measures section.© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433
  4. 4. 428 Lin Li et al.any attempts to stop smoking?’, or if a respondent stated Cigarettes per day (CPD) was asked at each wave andthat s/he had quit smoking since the previous survey recoded to: ‘already quit’, ‘1–10 CPD’, ‘11–20 CPD’ anddate. ‘21+ CPD’. Respondents were also asked at each wave about their endorsement for regulation of tobacco pro- ducts via the following statement: ‘Tobacco productsMeasures of exposure to POS health warnings should be more tightly regulated’. Response optionsRespondents were asked at each wave about their notic- were ‘strongly agree’, ‘agree’, ‘neither agree nor dis-ing of anti-smoking cues in a range of specific locations agree’, ‘disagree’ or ‘strongly disagree’.or media, such as on television, radio, posters or bill- Data analysisboards. In this context the following question regardingPOS anti-smoking warnings was asked: ‘In the last 6 All analyses were conducted using STATA version 10.1.months, have you noticed advertising or information that Both bivariate and multivariate analyses were conductedtalks about the dangers of smoking, or encourages quit- to examine (i) the association between reported exposureting, on store windows or inside stores where tobacco is to POS anti-smoking warnings and smokers’ interest insold?’. Those who answered ‘yes’ to this question were quitting in the same wave (i.e. cross-sectional associa-regarded as having been exposed to POS warnings. tion) and (ii) the association between POS exposure and subsequent quit attempts (i.e. longitudinal association, in which quit attempts were measured one wave after theCovariates exposure to POS warnings). We first used c2 tests to lookA compound measure of exposure to anti-smoking warn- at the unadjusted association between POS exposure andings in places other than at the POS was created from quit interest/attempts for each wave, and then used mul-the following eight items: exposure on television, radio, tivariate analyses to look at the fully adjusted associationmovies, posters/billboards, newspapers/magazines, ciga- while considering data from all waves.rette packs, leaflets and the internet (scores from 0–8). In Taking into consideration the correlated nature of theaddition, a measure of overall salience of anti-smoking data within people and across survey waves, we used theinformation was used based on the following question: generalized estimating equations (GEE) approach to‘In the last 6 months how often, if at all, have you noticed compute parameter estimates. All GEE models included aanti-smoking information?’ (1, never; 2, rarely; 3, some- specification for an unstructured within-subject correla-times; 4, often; 5, very often). tion structure, and parameter estimates were computed Demographics measured included sex (male, female), using robust variance. Our large sample size allowed usage at recruitment (18–24, 25–39, 40–54, 55 years and to assume an unstructured correlation structure in GEEolder) and identified majority/minority group, which was which helped us to estimate all possible correlationsbased on the primary means of identifying minorities in between within-subject responses and included them ineach country (i.e. racial/ethnic group in the United the estimation of the variances. Both main outcomesKingdom, Canada and the United States; and English were dichotomous, so the GEE models included a specifi-language spoken at home in Australia). Due to the differ- cation for the binomial distribution of the dependentences in economic development and educational systems variables. For the outcome of quit attempts a ‘forward’across countries, only relative levels of income and edu- specification was used in the analysis so longitudinalcation were used. ‘Low’ level of education referred to association was examined. In GEE modelling we firstthose who completed high school or less in Canada, the explored the country effect and its interactions forUnited States and Australia, or secondary/vocational or both outcome variables. Because country effect and itsless in the United Kingdom; ‘moderate’ meant commu- interaction were significant, we constructed separatenity college/trade/technical school/some university GEE models for each country. In each of the models we(no degree) in Canada and the United States, college/ included the following covariates: age, sex, ethnicity,university (no degree) in the United Kingdom or baseline education, baseline income and also the follow-technical/trade/some university (no degree) in Australia; ing time-varying covariates reported at each wave:and ‘high’ referred to those who completed university or endorsement of regulation for tobacco products, overallpostgraduate studies in all countries. Household income salience of anti-smoking information, reported exposurewas also grouped into ‘low’ [less than $30 000 (country- to anti-smoking warnings in eight places other than POSspecific dollars) (£30 000 in the United Kingdom) per and cigarettes per day.year], ‘moderate’ [$30 000–59 999 (£30 000–44 999 Ethics approvalin the United Kingdom)], and ‘high’ categories [equalto or greater than $60 000 (£45 000 in the United The study protocol was approved by the institutionalKingdom)]. review boards or research ethics boards of the University© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433
  5. 5. Association between POS warnings and quitting interest and attempts 429of Waterloo (Canada), Roswell Park Cancer Institute 1.249, P < 0.01], in that those who were exposed to POS(United States), University of Strathclyde (United warnings were 13.9% more likely at the same wave toKingdom), University of Stirling (United Kingdom), The report having quitting interest than those who wereOpen University (United Kingdom) and The Cancer not exposed (Table 3). Similarly, in Australia those whoCouncil Victoria (Australia). were exposed to POS warnings at one wave were 21.6% more likely to report having made quit attempts at the next wave than those who were not exposed (adjustedRESULTS odds ratio = 1.216, 95% CI 1.114–1.327, P < 0.001,Reported exposure to POS anti-smoking warnings by Table 4). No significant association between exposurewave and country to POS warnings and outcomes was found in the other three countries.Figure 2 shows that smokers in Australia reported higherlevels of noticing POS anti-smoking warnings comparedto their counterparts in the other three countries, and DISCUSSIONthis was consistent over time. It is also notable thatthe reported exposure to POS warnings in Australia POS anti-smoking warnings are designed to warn of theincreased over waves. There was also a smaller increase in dangers of smoking and encourage smokers to quit. Thisthe United Kingdom, but no clear evidence of an increase study shows clearly that the stronger, larger Australianin either Canada or the United States. POS warnings are noticed and are associated with increased interest in quitting and prospectively with quitInterest in quitting and quit attempts by wave attempts for smokers in Australia. However, the weaker or non-existent warnings in the other countries appear toTable 2 provides the reported quitting interest and have no measurable effects. In Australia, the positiveattempts over waves in each of the four countries. It can association between exposure to POS warnings and out-be seen that the proportions of reported interest in quit- comes remained even when exposure to anti-smokingting were generally higher than those of quit attempts in warnings in other domains and overall salience of anti-all the countries. Those who were exposed to POS anti- smoking information were controlled for.smoking warnings were more likely to report having quit- The findings raise two important questions. First, doesting interest in Australia, but this was not the case in the only finding a relation with quitting in Australia suggestother three countries. Similar results were found for quit some threshold effect, and secondly, how likely is it thatattempts with a positive effect in Australia and no clear the relation we found for quitting are causal? The Austra-pattern in the other three countries, although the pattern lian warnings are required to be placed in prime POSof significant results by year in Australia was less clear positions—they are not incidental to signs promoting orthan for quitting interest (Table 2). announcing tobacco sales, and they clearly promote the Quitline number. They are typically at least A4 in size andAssociations between exposure to POS warnings and black-on-white text, although there is variation betweenquitting interest and attempts: GEE modelling results states [12]. The small state of Tasmania now mandatesGEE modelling results show that over the waves in coloured graphic warnings using the same images as areAustralia (but not in the other three countries) there was on the cigarette pack warnings [20]), but the sample sizea significantly positive association between exposure is too small and the introduction too recent to be able toto POS warnings and quitting interest [adjusted odds test for an independent effect of these warnings. The dataratio = 1.139, 95% confidence interval (CI) 1.039– are clearly consistent with some threshold of exposure 100.0% Percent(yes)Figure 2 Noticing point-of-sale (POS) 50.0%anti-smoking warnings in shops by wave Australiain International Tobacco Control Four 0.0%Country Survey (ITC-4) countries (notes: W1- W2- W3- W4- W5- W6- W7- Canada‘W1’ means ‘wave 1’ of the survey; this 2002 2003 2004 2005 2006 2007 2008 UKapplies to the other waves. Proportions Australia 39.3% 41.2% 42.4% 44.9% 52.4% 51.1% 57.8% USreported here were based on weighted Canada 26.0% 21.1% 24.4% 24.9% 22.2% 21.2% 27.9%data. A significant linear trend over years/ UK 14.1% 14.0% 16.6% 15.5% 17.6% 19.4% 21.2%waves was found in Australia, but not inthe other three countries) US 18.6% 20.6% 21.8% 22.7% 22.0% 19.0% 23.5%© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433
  6. 6. 430 Lin Li et al. Percentage of participants who reported having some intention to quit. Exposure and interest in quitting variables were from the same wave (i.e. cross-sectional association). bPercentage of participants who reported having made that is related to warning message design and position. at least one quit attempt between two consecutive surveys. Quit attempts occurred in the next wave/s after exposure (i.e. longitudinal association). ‘–’ stands for ‘not applicable’, because the wave 8 attempt data were not available Not exposed However, the failure to find any evidence of effects of the 64.9 72.9 57.9 68.5 newly introduced POS anti-smoking warnings in some Wave 7 (2008) – – – – Canadian provinces weakens the argument for a thresh- old effect, and is more consistent with between-country Exposed 71.7** 62.6* 74.8 70.9 differences. It is also consistent with their warnings not – – – – evoking the broader anti-smoking efforts as do Australian warnings. In other studies of the ITC-4 cohort, we have Not exposed not found between-country effects for the impact of 68.6 49.1 74.5 46.9 56.7 40.8 69.6 46.1 health warnings on cigarette packs [14], even though in Wave 6 (2007) the United States they are far smaller and less prominent than in the other countries. One possibility for this Exposed 77.6*** 49.6 78.6 45.5 61.5 40.0 71.8 48.1 pattern of findings is that pack warnings are available to smokers all the time, and smokers are prone to notice at the time of writing. *Differences were significant at P < 0.05, Pearson’s c2 test (exposed versus not exposed; *is placed next to the higher figure); **P < 0.01; ***P < 0.001. them from time to time, even when they are tucked away Not exposedTable 2 Reported interest in quitting and quit attempts over time, by country and exposure status to point-of-sale (POS) anti-smoking warnings. on the side of the pack. By contrast, little time is spent in 69.8 46.7 75.1 42.6 60.1 41.3 71.2 44.7 stores, so they need to be prominent enough to be noticed Wave 5 (2006) with some regularity to have any real effect. This suggests that POS warnings should be designed and placed in such Exposed 76.7** 79.9** 46.1 74.4 41.9 40.5 71.2 41.6 a way that they are large and prominent enough to be noticed. One unique and relevant aspect of Australia’s Not exposed tobacco control programme is its comprehensiveness. 69.9* 70.9 47.5 76.2 42.8 63.9 41.1 40.3 It has integrated communications from mass-media Wave 4 (2005) campaigns, pack warnings and POS warnings. Recent research into the different measures designed to reduce Exposed 77.8** 50.6 75.8 43.4 66.9 39.0 64.2 43.4 smoking rates in Australia found that increases in tobacco taxes and greater exposure to televised mass media campaigns contributed to the decreased smoking Not exposed prevalence among Australian adults [21], and this is 72.2 41.9 77.2 42.9 63.0 44.9 69.2 41.1 consistent with international evidence showing that Wave 3 (2004) strong disease-related messages are potent motivators of making quit attempts [8]. The overall broad and multi- Exposed 77.5** 50.4** 74.9 43.0 58.0 44.1 69.0 47.2 location strategy of using the same prominent warnings on the cigarette packs and at POS, as well as via mass media would probably provide additional opportunities Not exposed for smokers to remember the warnings and this, in turn, 61.3* 72.3 41.9 80.7 43.9 38.3 72.0 38.9 might lead to an increase in the likelihood that the warn- Wave 2 (2003) ings would lead to action. Our data indicate that overall salience of anti-smoking information was associated Exposed 47.4* 75.6 80.3 46.5 54.3 34.3 68.4 44.7 with interest in quitting in all countries but Canada (for reasons that are unclear) and quit attempts in Australia and the United Kingdom. A recent study by Brennan Not exposed et al. [13] found that graphic health warnings on ciga- 65.5* 75.3 34.1 81.1 43.2 31.8 75.6 36.3 rette packets and on television operated in a complemen- Wave 1 (2002) tary manner to positively influence Australian smokers’ awareness of the health risks of smoking and motivation Exposed 76.9 34.6 82.3 45.8 59.0 30.2 71.4 37.3 to quit [13], suggesting that such a multi-location strat- egy may enhance effects. It is worth reflecting on the fact that in most jurisdictions where POS is critical for United Kingdom %Attemptedb %Attempted %Attempted %Attempted % Intendeda % Intended % Intended % Intended United States tobacco marketing, the POS pack displays have tradition- ally served as a prompt for smokers to purchase tobacco Australia Canada rather than avoid it, and to pose a risk for relapse for those who have quit [22,23]. a© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433
  7. 7. Association between POS warnings and quitting interest and attempts 431Table 3 Association between exposure to point-of-sale (POS) anti-smoking warnings and interest in quitting—generalized estimatingequation (GEE) modelling results. Adjusted odds ratioa 95% CI P-valueAustralia POS warnings 1.139 1.039–1.249 <0.01 Seen in other health warning domains (number) 1.065 1.034–1.097 <0.001 Overall salience of anti-smoking information 1.090 1.052–1.129 <0.001Canada POS warnings 0.919 0.826–1.022 0.122 Seen in other health warning domains (number) 1.096 1.064–1.129 <0.001 Overall salience of anti-smoking information 1.035 0.997–1.075 0.071United Kingdom POS warnings 1.014 0.908–1.132 0.806 Seen in other health warning domains (number) 1.072 1.044–1.101 <0.001 Overall salience of anti-smoking information 1.063 1.031–1.097 <0.001United States POS warnings 0.911 0.825–1.007 0.069 Seen in other health warning domains (number) 1.065 1.037–1.094 <0.001 Overall salience of anti-smoking information 1.083 1.045–1.121 <0.001Exposure and intention variables were from same wave (i.e. cross-sectional association). Exposure to POS health warnings was coded as ‘1’ for ‘yesexposed’, and ‘0’ for ‘not exposed’ (the referent). Main effect (not shown in table) for country and its interaction (exposure*country) was significant atthe P < 0.001 level, therefore separate models were constructed for each country. aAdjusted for gender, age, ethnicity, baseline education, baselineincome, endorsement for regulation of tobacco products at each wave, a compound measure of exposure to anti-smoking warnings in places other thanPOS, overall salience of anti-smoking information, cigarettes per day at each wave, wave and cohort. CI: confidence interval.Table 4 Longitudinal association between exposure to point-of-sale (POS) anti-smoking warnings and subsequent quit attempts—generalized estimating equation (GEE) modelling results. Adjusted odds ratioa 95% CI P-valueAustralia POS warnings 1.216 1.114–1.327 <0.001 Seen in other health warning domains (number) 0.995 0.968–1.023 0.718 Overall salience of anti-smoking information 1.042 1.006–1.079 0.023Canada POS warnings 0.909 0.820–1.007 0.067 Seen in other health warning domains (number) 1.011 0.983–1.039 0.450 Overall salience of anti-smoking information 0.997 0.961–1.034 0.854United Kingdom POS warnings 0.949 0.835–1.079 0.425 Seen in other health warning domains (number) 1.005 0.976–1.035 0.738 Overall salience of anti-smoking information 1.062 1.025–1.100 0.001United States POS warnings 0.987 0.900–1.082 0.774 Seen in other health warning domains (number) 1.022 0.997–1.048 0.081 Overall salience of anti-smoking information 0.959 0.928–0.992 0.016Exposure to POS health warnings was coded as ‘1’ for ‘yes exposed’, and ‘0’ for ‘not exposed’ (the referent). ‘Having made quit attempts’ were measuredone wave AFTER the exposure to POS warning, and a ‘forward’ specification was used in the analysis. Main effect (not shown in table) for country andits interaction (exposure*country) was significant at the P < 0.001 level, therefore separate models were constructed for each country. aAdjusted forgender, age, ethnicity, baseline education, baseline income, endorsement for regulations of tobacco products at each wave, a compound measure ofexposure to anti-smoking warnings in places other than POS, overall salience of anti-smoking information, cigarettes per day at each wave, wave andcohort. When interest in quitting was also adjusted for the effect remained essentially the same for each country, including the significant effect inAustralia. CI: confidence interval.© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433