Ll etal itc 4 pos warnings and quitting-addiction_published

The association between exposure to point-of-saleanti-smoking warnings and smokers’ interest inquitting and quit attempts: findings from theInternational Tobacco Control Four Country Surveyadd_3668 425..433

Lin Li1, Ron Borland1, Hua-Hie Yong1, Sara C. Hitchman2, Melanie A. Wakefield3,Karin A. Kasza4 & Geoffrey T. Fong2,5

VicHealth 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, Canada5

ABSTRACT

Aims 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: [email protected] 10 April 2011; initial review completed 30 June 2011; final version accepted 22 September 2011

INTRODUCTION

Increasing numbers of countries have acted to restrict/ban tobacco advertising and promotion in traditionalmedia such as television, radio, print and on billboards[1–5]. The tobacco industry has responded by expandingtheir marketing in areas where it is still allowed [6,7].Point-of-sale (POS) tobacco promotion has become animportant marketing channel for the tobacco industry, asother marketing opportunities have been banned [8]. The

tobacco industry in the United States has spent the vastmajority of their promotional budget in the retail envi-ronment over the past decade [9]. This ranges frompromotional material to payments for shelf space andincentive schemes for retailers.

A number of jurisdictions have banned POS tobaccoadvertising, and some are now moving to prohibit thedisplay of tobacco products—putting tobacco productsunder the counter. Regardless of these restrictions, thereare good reasons for having prominent POS anti-smoking

RESEARCH REPORT doi:10.1111/j.1360-0443.2011.03668.x

© 2012 The Authors, Addiction © 2012 Society for the Study of Addiction Addiction, 107, 425–433

warnings. This is the time when smokers purchase ciga-rettes, so it should be a critical time to have them thinkingabout the harms. However, there is little empirical evi-dence to support such initiatives.

During the study period (2002–08), Australia hadnotably stronger POS anti-smoking warnings than didthe other three countries in the study (Canada, the UnitedKingdom and the United States). Australia has hadprominent POS warnings for some time, but with somevariation across states. For example, the most populousstate of New South Wales requires a health warningmessage (e.g. ‘smoking is addictive’) and a Quitlinenumber (131 848) that features: (i) black writing on awhite background; and (ii) between 50 and 100 cm wideand have an area not less than 2000 cm2 [10] (SeeFig. 1). In the second most populous state of Victoria,similar warnings are mandated near the POS or at theshop entrance. However, they are smaller in Victoria [A4in size (21 ¥ 29.7 cm)] [11]. In Queensland, under theTobacco and Other Smoking Products Amendment Act2004, from December 2005 all the retail outlets arerequired to display mandatory POS anti-smoking signsthat encourage quitting. Two other states (South Austra-lia and Western Australia), that began the 2002–08study period with weaker signs, strengthened them to besimilar to those of New South Wales and Victoria duringthe study period.

In Australia, POS anti-smoking warnings have beenintegrated with other health communication strategies,such as mandating health warnings on cigarette packetsand ongoing mass media campaigns that all have empha-sized the health harms of smoking and promoted theQuitline number, sometimes using the same base mate-rials [12–14].

Warnings at the POS are less prominent in the othercountries. Some provinces of Canada started to mandatePOS warnings from around 2004/05 [15,16], but mostdo not have size requirements. Pictorial warnings at thePOS were only introduced in some provinces of Canada(e.g. British Columbia) from early 2008. In the UnitedKingdom, the Tobacco Advertising and Promotion (Pointof Sale) Regulations 2004 allowed only one A5-sized(14.8 ¥ 21 cm) poster advertising tobacco in store with30% of that area taken up with a health warning (e.g.‘smoking seriously harms you and others around you’,in force from December 2004) [17]. The United Stateshas no systematic requirements for POS anti-smokinginformation.

We expected an overall higher level of awareness ofPOS warnings in Australia, and given the improvementsin regulations regarding POS warnings over the studyperiod, increases in awareness in Australia, Canada andthe United Kingdom over-time. We were also interested indetermining if reported awareness of POS warnings wereassociated with higher quitting interest and prospectively,with higher levels of quitting activity, and whether thisvaried by country (in this case a proxy for warningstrength).

METHODS

Data source and participants

The data come from the first seven waves of the Interna-tional Tobacco Control Four Country Survey (the ITC-4Survey), which has been running annually since 2002in Australia, Canada, the United Kingdom and theUnited States. A detailed description of the conceptual

Figure 1 An example of Australian point-of-sale anti-smoking warnings (notes: thispicture was taken in a supermarket in NewSouth Wales during the study period;source: Simon Chapman of University ofSydney and Anne Jones of Action onSmoking and Health Australia)

426 Lin Li et al.


framework and methods of the ITC-4 Survey has beenreported by Fong et al. and Thompson et al. [18,19], andmore detail is available at http://www.itcproject.org.Briefly, the ITC-4 Survey employs a prospective multi-country cohort design and involves annual telephonesurveys of representative cohorts of adult smokers ineach of the four countries. The sampling is conducted viarandom-digit dialling. The sample size per country isapproximately 2000. At each survey wave, the approxi-mately 30% lost to follow-up are replenished from thesame sampling frame.

Participants were aged 18+ years, had smoked at least100 cigarettes life-time and had smoked at least once inthe past 30 days at the time of recruitment. A total of21 613 respondents who provided complete informationfor at least one of the seven waves (between 2002 and2008) were included in the analysis. The sample size foreach country, the numbers of recruits in each wave/year,and their characteristics are summarized in Table 1.

Measures

The ITC-4 Survey included questions about smokingbehaviour, cigarette brand information, quit history,exposure to anti-tobacco advertising, psychosocial mea-sures and demographics.

Outcome measures

The key outcomes were interest in quitting and quitattempts. Interest in quitting was assessed by: ‘Areyou planning to quit smoking?’. Response options were‘within the next month’, ‘within the next 6 months’,‘some time in the future, beyond 6 months’, ‘not planningto quit’ and ‘don’t know’. The first three categories wererecoded to ‘having some interest in quitting’ and theremaining as ‘not having an interest’. Quit attempts wereassessed at the next wave based on the answer to: ‘Sincewe last talked to you [in last survey date], have you made

Table 1 Sample characteristics, by country.

Australia Canada United Kingdom United States Total

n = 4806 n = 5265 n = 5251 n = 6291 n = 21 613

Gender (% 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 11.725–39 35.6 30.3 31.3 26.3 30.640–54 34.0 36.5 33.8 36.6 35.355+ 16.1 20.4 26.4 25.5 22.4

Education at recruitment (%)a

Low 64.1 48.6 60.6 45.8 54.1Moderate 21.9 36.4 25.1 38.3 30.9High 13.9 14.8 13.5 15.9 14.6

Income at recruitment (%)a

Low 26.9 28.3 31.1 36.9 31.2Moderate 32.7 34.4 31.5 33.3 33.0High 34.0 28.4 27.6 22.7 27.8No information 6.5 8.9 9.8 7.1 8.1

Cigarettes per day at recruitment (%)1–10 29.8 31.4 29.8 31.1 30.611–20 40.1 42.7 53.4 45.8 45.621–30 22.8 21.1 11.7 13.1 16.931+ 7.1 4.4 4.8 9.4 6.6

No. of new recruits in each wave (n)Wave 1 (in 2002) 2305 2214 2401 2138 9058Wave 2 (2003) 258 517 255 684 1714Wave 3 (2004) 532 545 586 889 2552Wave 4 (2005) 362 519 503 742 2126Wave 5 (2006) 686 594 613 745 2638Wave 6 (2007) 539 556 523 711 2329Wave 7 (2008) 142 320 370 382 1196

Percentages 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.

Association between POS warnings and quitting interest and attempts 427


any attempts to stop smoking?’, or if a respondent statedthat s/he had quit smoking since the previous surveydate.

Measures of exposure to POS health warnings

Respondents were asked at each wave about their notic-ing of anti-smoking cues in a range of specific locationsor media, such as on television, radio, posters or bill-boards. In this context the following question regardingPOS anti-smoking warnings was asked: ‘In the last 6months, have you noticed advertising or information thattalks about the dangers of smoking, or encourages quit-ting, on store windows or inside stores where tobacco issold?’. Those who answered ‘yes’ to this question wereregarded as having been exposed to POS warnings.

Covariates

A compound measure of exposure to anti-smoking warn-ings in places other than at the POS was created fromthe following eight items: exposure on television, radio,movies, posters/billboards, newspapers/magazines, ciga-rette packs, leaflets and the internet (scores from 0–8). Inaddition, a measure of overall salience of anti-smokinginformation was used based on the following question:‘In the last 6 months how often, if at all, have you noticedanti-smoking information?’ (1, never; 2, rarely; 3, some-times; 4, often; 5, very often).

Demographics measured included sex (male, female),age at recruitment (18–24, 25–39, 40–54, 55 years andolder) and identified majority/minority group, which wasbased on the primary means of identifying minorities ineach country (i.e. racial/ethnic group in the UnitedKingdom, Canada and the United States; and Englishlanguage spoken at home in Australia). Due to the differ-ences in economic development and educational systemsacross countries, only relative levels of income and edu-cation were used. ‘Low’ level of education referred tothose who completed high school or less in Canada, theUnited States and Australia, or secondary/vocational orless in the United Kingdom; ‘moderate’ meant commu-nity college/trade/technical school/some university(no degree) in Canada and the United States, college/university (no degree) in the United Kingdom ortechnical/trade/some university (no degree) in Australia;and ‘high’ referred to those who completed university orpostgraduate studies in all countries. Household incomewas also grouped into ‘low’ [less than $30 000 (country-specific dollars) (£30 000 in the United Kingdom) peryear], ‘moderate’ [$30 000–59 999 (£30 000–44 999in the United Kingdom)], and ‘high’ categories [equalto or greater than $60 000 (£45 000 in the UnitedKingdom)].

Cigarettes per day (CPD) was asked at each wave andrecoded to: ‘already quit’, ‘1–10 CPD’, ‘11–20 CPD’ and‘21+ CPD’. Respondents were also asked at each waveabout their endorsement for regulation of tobacco pro-ducts via the following statement: ‘Tobacco productsshould be more tightly regulated’. Response optionswere ‘strongly agree’, ‘agree’, ‘neither agree nor dis-agree’, ‘disagree’ or ‘strongly disagree’.

Data analysis

All analyses were conducted using STATA version 10.1.Both bivariate and multivariate analyses were conductedto examine (i) the association between reported exposureto POS anti-smoking warnings and smokers’ interest inquitting in the same wave (i.e. cross-sectional associa-tion) and (ii) the association between POS exposure andsubsequent quit attempts (i.e. longitudinal association, inwhich quit attempts were measured one wave after theexposure to POS warnings). We first used c2 tests to lookat the unadjusted association between POS exposure andquit interest/attempts for each wave, and then used mul-tivariate analyses to look at the fully adjusted associationwhile considering data from all waves.

Taking into consideration the correlated nature of thedata within people and across survey waves, we used thegeneralized estimating equations (GEE) approach tocompute parameter estimates. All GEE models included aspecification for an unstructured within-subject correla-tion structure, and parameter estimates were computedusing robust variance. Our large sample size allowed usto assume an unstructured correlation structure in GEEwhich helped us to estimate all possible correlationsbetween within-subject responses and included them inthe estimation of the variances. Both main outcomeswere dichotomous, so the GEE models included a specifi-cation for the binomial distribution of the dependentvariables. For the outcome of quit attempts a ‘forward’specification was used in the analysis so longitudinalassociation was examined. In GEE modelling we firstexplored the country effect and its interactions forboth outcome variables. Because country effect and itsinteraction were significant, we constructed separateGEE models for each country. In each of the models weincluded the following covariates: age, sex, ethnicity,baseline education, baseline income and also the follow-ing time-varying covariates reported at each wave:endorsement of regulation for tobacco products, overallsalience of anti-smoking information, reported exposureto anti-smoking warnings in eight places other than POSand cigarettes per day.

Ethics approval

The study protocol was approved by the institutionalreview boards or research ethics boards of the University

428 Lin Li et al.


of Waterloo (Canada), Roswell Park Cancer Institute(United States), University of Strathclyde (UnitedKingdom), University of Stirling (United Kingdom), TheOpen University (United Kingdom) and The CancerCouncil Victoria (Australia).

RESULTS

Reported exposure to POS anti-smoking warnings bywave and country

Figure 2 shows that smokers in Australia reported higherlevels of noticing POS anti-smoking warnings comparedto their counterparts in the other three countries, andthis was consistent over time. It is also notable thatthe reported exposure to POS warnings in Australiaincreased over waves. There was also a smaller increase inthe United Kingdom, but no clear evidence of an increasein either Canada or the United States.

Interest in quitting and quit attempts by wave

Table 2 provides the reported quitting interest andattempts over waves in each of the four countries. It canbe seen that the proportions of reported interest in quit-ting were generally higher than those of quit attempts inall the countries. Those who were exposed to POS anti-smoking warnings were more likely to report having quit-ting interest in Australia, but this was not the case in theother three countries. Similar results were found for quitattempts with a positive effect in Australia and no clearpattern in the other three countries, although the patternof significant results by year in Australia was less clearthan for quitting interest (Table 2).

Associations between exposure to POS warnings andquitting interest and attempts: GEE modelling results

GEE modelling results show that over the waves inAustralia (but not in the other three countries) there wasa significantly positive association between exposureto POS warnings and quitting interest [adjusted oddsratio = 1.139, 95% confidence interval (CI) 1.039–

1.249, P < 0.01], in that those who were exposed to POSwarnings were 13.9% more likely at the same wave toreport having quitting interest than those who werenot exposed (Table 3). Similarly, in Australia those whowere exposed to POS warnings at one wave were 21.6%more likely to report having made quit attempts at thenext wave than those who were not exposed (adjustedodds ratio = 1.216, 95% CI 1.114–1.327, P < 0.001,Table 4). No significant association between exposureto POS warnings and outcomes was found in the otherthree countries.

DISCUSSION

POS anti-smoking warnings are designed to warn of thedangers of smoking and encourage smokers to quit. Thisstudy shows clearly that the stronger, larger AustralianPOS warnings are noticed and are associated withincreased interest in quitting and prospectively with quitattempts for smokers in Australia. However, the weaker ornon-existent warnings in the other countries appear tohave no measurable effects. In Australia, the positiveassociation between exposure to POS warnings and out-comes remained even when exposure to anti-smokingwarnings in other domains and overall salience of anti-smoking information were controlled for.

The findings raise two important questions. First, doesonly finding a relation with quitting in Australia suggestsome threshold effect, and secondly, how likely is it thatthe relation we found for quitting are causal? The Austra-lian warnings are required to be placed in prime POSpositions—they are not incidental to signs promoting orannouncing tobacco sales, and they clearly promote theQuitline number. They are typically at least A4 in size andblack-on-white text, although there is variation betweenstates [12]. The small state of Tasmania now mandatescoloured graphic warnings using the same images as areon the cigarette pack warnings [20]), but the sample sizeis too small and the introduction too recent to be able totest for an independent effect of these warnings. The dataare clearly consistent with some threshold of exposure

0.0%

50.0%

100.0%

Per

cen

t(ye

s)

Australia

Canada

UK

USAustralia 39.3% 41.2% 42.4% 44.9% 52.4% 51.1% 57.8%

Canada 26.0% 21.1% 24.4% 24.9% 22.2% 21.2% 27.9%

UK 14.1% 14.0% 16.6% 15.5% 17.6% 19.4% 21.2%

US 18.6% 20.6% 21.8% 22.7% 22.0% 19.0% 23.5%

W1-2002

W2-2003

W3-2004

W4-2005

W5-2006

W6-2007

W7-2008

Figure 2 Noticing point-of-sale (POS)anti-smoking warnings in shops by wavein International Tobacco Control FourCountry Survey (ITC-4) countries (notes:‘W1’ means ‘wave 1’ of the survey; thisapplies to the other waves. Proportionsreported here were based on weighteddata. A significant linear trend over years/waves was found in Australia, but not inthe other three countries)



that is related to warning message design and position.However, the failure to find any evidence of effects of thenewly introduced POS anti-smoking warnings in someCanadian provinces weakens the argument for a thresh-old effect, and is more consistent with between-countrydifferences. It is also consistent with their warnings notevoking the broader anti-smoking efforts as do Australianwarnings. In other studies of the ITC-4 cohort, we havenot found between-country effects for the impact ofhealth warnings on cigarette packs [14], even though inthe United States they are far smaller and less prominentthan in the other countries. One possibility for thispattern of findings is that pack warnings are availableto smokers all the time, and smokers are prone to noticethem from time to time, even when they are tucked awayon the side of the pack. By contrast, little time is spent instores, so they need to be prominent enough to be noticedwith some regularity to have any real effect. This suggeststhat POS warnings should be designed and placed in sucha way that they are large and prominent enough to benoticed.

One unique and relevant aspect of Australia’stobacco control programme is its comprehensiveness.It has integrated communications from mass-mediacampaigns, pack warnings and POS warnings. Recentresearch into the different measures designed to reducesmoking rates in Australia found that increases intobacco taxes and greater exposure to televised massmedia campaigns contributed to the decreased smokingprevalence among Australian adults [21], and this isconsistent with international evidence showing thatstrong disease-related messages are potent motivators ofmaking quit attempts [8]. The overall broad and multi-location strategy of using the same prominent warningson the cigarette packs and at POS, as well as via massmedia would probably provide additional opportunitiesfor smokers to remember the warnings and this, in turn,might lead to an increase in the likelihood that the warn-ings would lead to action. Our data indicate that overallsalience of anti-smoking information was associatedwith interest in quitting in all countries but Canada (forreasons that are unclear) and quit attempts in Australiaand the United Kingdom. A recent study by Brennanet al. [13] found that graphic health warnings on ciga-rette packets and on television operated in a complemen-tary manner to positively influence Australian smokers’awareness of the health risks of smoking and motivationto quit [13], suggesting that such a multi-location strat-egy may enhance effects. It is worth reflecting on thefact that in most jurisdictions where POS is critical fortobacco marketing, the POS pack displays have tradition-ally served as a prompt for smokers to purchase tobaccorather than avoid it, and to pose a risk for relapse forthose who have quit [22,23].Ta

ble

2R

epor

ted

inte

rest

inqu

itti

ng

and

quit

atte

mpt

sov

erti

me,

byco

un

try

and

expo

sure

stat

us

topo

int-

of-s

ale

(PO

S)an

ti-s

mok

ing

war

nin

gs.

Wav

e1

(20

02

)W

ave

2(2

00

3)

Wav

e3

(20

04

)W

ave

4(2

00

5)

Wav

e5

(20

06

)W

ave

6(2

00

7)

Wav

e7

(20

08

)

Exp

osed

Not

expo

sed

Exp

osed

Not

expo

sed

Exp

osed

Not

expo

sed

Exp

osed

Not

expo

sed

Exp

osed

Not

expo

sed

Exp

osed

Not

expo

sed

Exp

osed

Not

expo

sed

Au

stra

lia%

Inte

nde

da7

6.9

75

.37

5.6

72

.37

7.5

**7

2.2

77

.8**

70

.97

6.7

**6

9.8

77

.6**

*6

8.6

71

.7**

64

.9%

Att

empt

edb

34

.63

4.1

47

.4*

41

.95

0.4

**4

1.9

50

.64

7.5

46

.14

6.7

49

.64

9.1

––

Can

ada

%In

ten

ded

82

.38

1.1

80

.38

0.7

74

.97

7.2

75

.87

6.2

74

.47

5.1

78

.67

4.5

74

.87

2.9

%A

ttem

pted

45

.84

3.2

46

.54

3.9

43

.04

2.9

43

.44

2.8

41

.94

2.6

45

.54

6.9

––

Un

ited

Kin

gdom

%In

ten

ded

59

.06

5.5

*5

4.3

61

.3*

58

.06

3.0

66

.96

3.9

79

.9**

60

.16

1.5

56

.76

2.6

*5

7.9

%A

ttem

pted

30

.23

1.8

34

.33

8.3

44

.14

4.9

39

.04

1.1

40

.54

1.3

40

.04

0.8

––

Un

ited

Stat

es%

Inte

nde

d7

1.4

75

.66

8.4

72

.06

9.0

69

.26

4.2

69

.9*

71

.27

1.2

71

.86

9.6

70

.96

8.5

%A

ttem

pted

37

.33

6.3

44

.73

8.9

47

.24

1.1

43

.44

0.3

41

.64

4.7

48

.14

6.1

––

a Perc

enta

geof

part

icip

ants

wh

ore

port

edh

avin

gso

me

inte

nti

onto

quit

.Exp

osu

rean

din

tere

stin

quit

tin

gva

riab

les

wer

efr

omth

esa

me

wav

e(i

.e.c

ross

-sec

tion

alas

soci

atio

n).

b Perc

enta

geof

part

icip

ants

wh

ore

port

edh

avin

gm

ade

atle

asto

ne

quit

atte

mpt

betw

een

two

con

secu

tive

surv

eys.

Qu

itat

tem

pts

occu

rred

inth

en

extw

ave/

saf

ter

expo

sure

(i.e

.lon

gitu

din

alas

soci

atio

n).

‘–’s

tan

dsfo

r‘n

otap

plic

able

’,be

cau

seth

ew

ave

8at

tem

ptda

taw

ere

not

avai

labl

eat

the

tim

eof

wri

tin

g.*D

iffer

ence

sw

ere

sign

ifica

nt

atP

<0

.05

,Pea

rson

’sc2

test

(exp

osed

vers

us

not

expo

sed;

*is

plac

edn

ext

toth

eh

igh

erfig

ure

);**

P<

0.0

1;*

**P

<0

.00

1.

430 Lin Li et al.


Table 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-value

AustraliaPOS warnings 1.139 1.039–1.249 <0.01Seen in other health warning domains (number) 1.065 1.034–1.097 <0.001Overall salience of anti-smoking information 1.090 1.052–1.129 <0.001

CanadaPOS warnings 0.919 0.826–1.022 0.122Seen in other health warning domains (number) 1.096 1.064–1.129 <0.001Overall salience of anti-smoking information 1.035 0.997–1.075 0.071

United KingdomPOS warnings 1.014 0.908–1.132 0.806Seen in other health warning domains (number) 1.072 1.044–1.101 <0.001Overall salience of anti-smoking information 1.063 1.031–1.097 <0.001

United StatesPOS warnings 0.911 0.825–1.007 0.069Seen in other health warning domains (number) 1.065 1.037–1.094 <0.001Overall salience of anti-smoking information 1.083 1.045–1.121 <0.001

Exposure 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-value

AustraliaPOS warnings 1.216 1.114–1.327 <0.001Seen in other health warning domains (number) 0.995 0.968–1.023 0.718Overall salience of anti-smoking information 1.042 1.006–1.079 0.023

CanadaPOS warnings 0.909 0.820–1.007 0.067Seen in other health warning domains (number) 1.011 0.983–1.039 0.450Overall salience of anti-smoking information 0.997 0.961–1.034 0.854

United KingdomPOS warnings 0.949 0.835–1.079 0.425Seen in other health warning domains (number) 1.005 0.976–1.035 0.738Overall salience of anti-smoking information 1.062 1.025–1.100 0.001

United StatesPOS warnings 0.987 0.900–1.082 0.774Seen in other health warning domains (number) 1.022 0.997–1.048 0.081Overall salience of anti-smoking information 0.959 0.928–0.992 0.016

Exposure 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.



Although some weak effect of POS exposure on inter-est in quitting was identified (in bivariate analysis) in acouple of waves (i.e. waves 5, 7) in the United Kingdom,overall the relation between POS warning exposureand quit interest/attempts in other three countries (i.e.Canada, the United Kingdom and the United States) wasnot significant or consistent. As these countries have alsostrengthened POS regulations governing tobacco display/health warnings more recently, it will be interesting forfuture research to use the upcoming ITC-4 survey datato assess further their impacts on cessation.

This study has some strengths and weaknesses. One ofthe main strengths is its longitudinal design, whichallowed for changes in exposure and outcomes over timeto be assessed for the outcome of quit attempts, especiallyusing GEE modelling which allowed us to combinerespondents from all waves while accounting for inherentwithin-person correlation, thereby increasing our samplesize and power to detect effects. This study covered a7-year period, providing unique and rich informationto examine the association between POS anti-smokingwarnings and smokers’ interest in quitting and theirsubsequent quit attempts. Our use of nationally represen-tative samples of smokers and essentially identical mea-sures and parallel data collection methods across all fourcountries [18,24] strengthened the validity of conclu-sions about the similarities or differences of the impactof POS warnings.

This study did not examine the nuances ofenactment/implementation differences between states/provinces within the countries. The subnational differ-ences might be important, and should be explored infuture research. The study relies on self-reports of expo-sure to anti-smoking warnings and it is possible thatsensitivity to warnings, rather than their actual pres-ence in the person’s environment, might be more impor-tant. Thus smokers more interested in quitting mightbe noticing the warnings more and then these are thesmokers who go on to make more quit attempts, makingthe results an artefact. We think this is unlikely, aswe controlled for interest in quitting in the analysesof quit attempts, and this explanation cannot accountfor the effects only being present in Australia. Respon-dents’ recall might be imprecise, especially when a longperiod of time (e.g. 6 months or 1 year) was involved,and they could be affected by external factors such asamount of competing material in the environment.However, there is no evidence to suggest that self-reportis systematically inaccurate in population-based studiesof this kind. The important point here is to exercisecaution when making inference about differences acrossthe countries.

In summary, the results of this study indicate thatthe effects of POS anti-smoking warnings on interest in

quitting and quit attempts were not homogenous acrossthe four western countries, and that greater exposure toPOS warnings was associated positively with increasedinterest in quitting and quit attempts only for smokers inAustralia over time. The results support the use of moreprominent, comprehensive and consistent anti-smokingwarnings at the POS, especially if they are employed inmultiple locations and multiple channels as has beencarried out in Australia. Doing so should encourage moresmokers to take action to quit.

Declarations of interest

None identified.

Acknowledgements

The authors would like to thank other members of theITC Four Country Survey team for their support. Wewould also like to thank Jessica Longbottom of CancerCouncil Victoria and Karima Ladhani of University ofWaterloo for helping obtain some point-of-sale policiesand relevant materials. We are grateful to the anony-mous reviewers and editors who provided useful sugges-tions on an earlier draft of this paper. The researchreported in this paper was funded by grants from theNational Cancer Institute of the United States (P50CA111326, P01 CA138389 and R01 CA100362),Robert Wood Johnson Foundation (045734), CanadianInstitutes of Health Research (57897 and 79551),National Health and Medical Research Council ofAustralia (265903 and 450110), Cancer Research UK(C312/A3726), Canadian Tobacco Control Research Ini-tiative (014578) and the Ontario Institute for CancerResearch, with additional support from the Propel Centrefor Population Health Impact, Canadian Cancer Societyand a Prevention Scientist Award from the CanadianCancer Society Research Institute. Sara C. Hitchmanreceives additional funding from a Canadian Institutes ofHealth Research Doctoral Research Award. The fundingsources had no role in study design; in the collection,analysis and interpretation of data; in the writing of themanuscript; nor in the decision to submit the manuscriptfor publication.

References

1. Saffer H. Tobacco advertising and promotion. In: JhaP., Chaploupka F., editors. Tobacco Control in DevelopingCountries. Oxford: Oxford University Press, Inc.; 2000,p. 215–36.

2. Chitanondh H. Thailand country report on tobacco advertisingand promotion bans. Geneva: World Health Organization;2003.

3. Harris F., MacKintosh A. M., Anderson S., Hastings G.,Borland R., Fong G. T. et al. Effects of the 2003 advertising/

432 Lin Li et al.


promotion ban in the United Kingdom on awareness oftobacco marketing: findings from the International TobaccoControl (ITC) Four Country Survey. Tob Control 2006; 15:iii26–33.

4. Chapman S., Wakefield M. Tobacco control advocacy inAustralia: reflections on 30 years of progress. Health EducBehav 2001; 28: 274–89.

5. Paynter J., Edwards R., Schluter P., McDuff I. Point of saletobacco displays and smoking among 14 15 years olds inNew Zealand: a cross-sectional study. Tob Control 2009; 18:268–74.

6. Paynter J., Edwards R. The impact of tobacco promotion atthe point of sale: a systematic review. Nicotine Tob Res 2009;11: 25–35.

7. Li L., Yong H.-H., Borland R., Fong G. T., Thompson M. E.,Jiang Y. et al. Reported awareness of tobacco advertisingand promotion in China compared to Thailand, Australiaand the USA. Tob Control 2009; 18: 222–7.

8. National Cancer Institute. The role of the media in promotingand reducing tobacco use. Tobacco control monograph no. 19.NIH Publication no. 07-6242, June 2008. Bethesda, MD:US Department of Health and Human Services, NationalInstitutes of Health, National Cancer Institute; 2008,p. 211–81.

9. Federal Trade Commission. Cigarette Report for 2004 and2005. Washington, DC: Federal Trade Commission; 2007.

10. New South Wales Parliamentary Counsel’s Office. PublicHealth (Tobacco) Regulation 1999. 1999. Available at: http://www.legislation.nsw.gov.au/xref/inforce/?xref=Type%3Dsubordleg%20AND%20Year%3D1999%20AND%20No%3D468&nohits=y (accessed 3 March 2011; archived byWebcite at http://www.webcitation.org/617U6kZoP).

11. Victorian Government Department of Human Services.Tobacco Retailer Guide. Melbourne; 2006. Available at:http://www.health.vic.gov.au/tobaccoreforms/downloads/retailers_guide.pdf (accessed 18 January 2011).

12. Freeman B., Chapman S. 11.4. State and territorylegislation. In: Scollo M., Winstanley M., editors. Tobacco inAustralia: Facts and Issues, 3rd edn. Melbourne: The CancerCouncil Victoria; 2011, p. 2–4. Available at: http://www.tobaccoinaustralia.org.au/chapter-11-advertising/11-4-state-and-territory-legislation (accessed 17 February2011).

13. Brennan E., Durkin S., Cotter T., Harper T., Wakefield M. A.Mass media campaigns designed to support new pictorialhealth warnings on cigarette packets: evidence of a comple-mentary relationship. Tob Control 2011; 20: 412–18.

14. Borland R., Wilson N., Fong G., Hammond D., CummingsK., Yong H. et al. Impact of graphic and text warnings oncigarette packs: findings from four countries over five years.Tob Control 2009; 18: 358–64.

15. Government of Ontario. Smoke-Free Ontario Act. 2006.Available at: http://www.cctc.ca/cctc/EN/lawandtobacco/byregion/on/smoke-free_ontario_act (accessed 28 July2009; archived by Webcite at http://www.webcitation.org/617UXBJVT).

16. Government of Quebec. Tobacco Act. 2010. Availableat: http://www2.publicationsduquebec.gouv.qc.ca/dynamicSearch/telecharge.php?type=2&file=/T_0_01/T0_01_A.html (accessed 28 February 2010; archived byWebcite at: http://www.webcitation.org/617Ujc4KN).

17. Government of the UK. The Tobacco Advertising and Promo-tion (Point of Sale) Regulations 2004; 2004. Norwich: TheStationery Office Limited.

18. Fong G. T., Cummings K. M., Borland R., Hastings G.,Hyland A., Giovino G. A. et al. The conceptual framework ofthe International Tobacco Control (ITC) Policy EvaluationProject. Tob Control 2006; 15: iii3–11.

19. Thompson M. E., Fong G. T., Hammond D., Boudreau C.,Driezen P., Hyland A. et al. Methods of the InternationalTobacco Control (ITC) Four Country Survey. Tob Control2006; 15: iii12–8.

20. Bicevskis M. Graphic point-of-display tobacco health warningsin Tasmania: Expected and unexpected benefits. 2006. Avail-able at: http://2006.confex.com/uicc/wctoh/techprogram/P5024.HTM (accessed 28 July 2010; archived by Webciteat: http://2006.confex.com/uicc/wctoh/techprogram/P5024.HTM).

21. Wakefield M., Durkin S., Spittal M. Impact of tobaccocontrol policies and mass media campaigns on monthlyadult smoking prevalence. Am J Public Health 2008; 98:1–8.

22. Germain D., McCarthy M., Wakefield M. Smoker sensitivityto retail tobacco displays and quitting: a cohort study.Addiction 2009; 105: 159–63.

23. Wakefield M., Germain D., Henriksen L. The effect of retailcigarette pack displays on impulse purchase. Addiction2008; 103: 322–8.

24. International Agency for Research on Cancer (IARC). IARCHandbooks of Cancer Prevention: Tobacco Control. Volume 12.Methods for Evaluating Tobacco Control Policies. Lyon, France:International Agency for Research on Cancer; 2008.



Health & Medicine

Ll etal itc 4 pos warnings and quitting-addiction_published