Study protocol: A cluster randomised controlled trial of implementation intentions to reduce smoking initiation in adolescents.

0.3 MB·English

by  Conner, Mark

Conneretal.BMCPublicHealth2013,13:54
http://www.biomedcentral.com/1471-2458/13/54
STUDY PROTOCOL Open Access
Study protocol: A cluster randomised controlled
trial of implementation intentions to reduce
smoking initiation in adolescents
Mark Conner1*, Sarah Grogan2, Rebecca Lawton1, Christopher Armitage3, Robert West4, Kamran Siddiqi5,
Brenda Gannon4, Carole Torgerson6, Keira Flett2 and Ruth Simms-Ellis1
Abstract
Background: The current literature suggests that forming implementation intentions (simple ‘if-then’plans) about
how to refuse the offerof a cigarette may be an effective intervention to reduce smoking initiation inadolescents.
Thisstudy is a pragmatic trial to test theeffectivenessand cost-effectiveness of such an interventionin reducing
smoking initiation in a sample of UK adolescents.
Methods/Design:A cluster randomised controlledtrial with atleast 36 schools randomised to receive an
implementation intention intervention targeting reducing smoking initiation (interventiongroup) or increasing
homework (control group). Interventionswill be conducted atthe classroom level and be repeated every six
months for four years (eight interventions). Objectively assessed (carbon monoxide monitor) and self-reported
smoking plus smoking related cognitions (e.g., smoking intentions, attitudes, norms and self-efficacy) will be
assessed atbaselineand 12, 24,36 and 48 months post baseline. Objectively assessed smoking at48 months post
baseline will be theprimary outcome variable. Health economic analyses will assess life years gained.
Discussion: The results ofthe trial will provideinformationon the impact of a repeated implementation intention
for refusing offers ofcigarettes onrates of smoking initiation in adolescents.
Trial registration: ISRCTN27596806
Background that awareness of the health consequences of smoking is
Both in the UK and internationally, tobacco smoking now widespread, even among the young. In the UK,
continues to be an important cause of morbidity and while the rates of regular smoking at 11 years of age are
mortality. For example, smoking related illnesses are only 0.5%, this rapidly rises to 15% by 15 years of age,
estimated to kill over 100,000 people in the UK each and then more gradually to around 20% among young
year [1,2], usually later in life. Yet smoking is a behav- adults [7]. Therecent UK Department of Health [2] plan
iour that is, in general, taken up between the ages of 10 is to reduce the rates of regular smoking in 15 year olds
and 20 years. Avariety of different research studies have to 12% or less by the end of 2015. Two important ways
supportedtheideathatthevastmajorityofsmokerstake to tackle smoking-related harm are interventions to help
upthishabit asadolescents[3-5]with40%ofadult smo- individuals (usually adults) quit smoking and interven-
kers having started before they reached 16 years of age tions to help individuals (usually adolescents) not to ini-
[1]. The General Household Survey [6] reported that tiate smoking. The present paper reports the protocol
38% of adult regular smokers took up the habit before for an intervention designed to reduce smoking initi-
the age of 15 years. This appears to be the case despite ation in adolescents as potentially the most effective way
the fact that health promotion messages have ensured toreduce smoking-related harm.
A worthwhile intervention to reduce smoking initiation
inthegroupsmostlikelytotakeupthishabit(i.e.,adoles-
*Correspondence:[e-mail]
cents)wouldneedtohaveatleastthreeimportantcharac-
1InstituteofPsychologicalSciences,UniversityofLeeds,LeedsLS29JT,UK
Fulllistofauthorinformationisavailableattheendofthearticle teristics: strong effects on reducing initiation; wide reach;
©2013Conneretal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreative
CommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedtheoriginalworkisproperlycited.
Conneretal.BMCPublicHealth2013,13:54 Page2of9
http://www.biomedcentral.com/1471-2458/13/54
and low cost. Promising initial data have been collected test multi-modal interventions [13], although there are far
[8] (i.e., an explanatory trial) on the efficacy of one inter- fewersuchstudiesinthisarea.Afurtherproblemhasbeen
vention technique that may have all of these characteris- that manyof theseinterventions havebeenquite intensive
tics. The study outlined here would aim to collect further and costly to implement. For example, the Hutchinson
data on the outcome effectiveness and cost-effectiveness Smoking Prevention Project [15] included a total of 65
ofthistechniqueintheformofaclusterrandomisedcon- lessons provided over a period of 8 years. Similarly, the
trolledtrial(i.e.,apragmatictrial).Giventhenatureofthe ASSIST intervention involved two-day training events for
existingdatainthisarea,thishasbeencastasapragmatic thepeersupporters.Suchinterventionsmaybecostlyand
trial [9] that could inform the potential roll out of this difficult to scale up into population level interventions
interventionforwidespreaduse.Theinterventioninques- that could be conducted in all schools across the UK
tionistheformationofrepeatedimplementationintentions at reasonable cost and with relative ease. In contrast, an
about how to refuse offers of cigarettes. Implementation interventioninvolvingformingarepeatedimplementation
intentions are simple ‘if-then’ plans [10] about how to re- intention about how to refuse offers of cigarettes would
spondtoenvironmentalcuesinordertohelpachieveagoal appear to be efficacious in reducing smoking initiation
such as not taking up smoking, e.g., If offered a cigarette I rates [8,12] and is also simple, easy to administer, cheap,
will say ‘no thanks, I don’t smoke’. Such plans can be and would appeartohave the potential tobereadilyscal-
formed before the opportunity to act presents itself and able up into a population-level intervention delivered in
havebeenfoundtobeaneffectivemeanstochangearange thevastmajorityofUKclassrooms.
of behaviours [11]. Our research has shown that the Gollwitzer[10]hasdefinedanimplementationintention
repeated formation of implementation intentions about as aplanof how,whereandwhentocommitabehaviour
how to refuse offers of cigarettes can have a strong effect (see [11] for a review). This type of plan establishes an
on reducing smoking initiation rates among adolescents ‘if-then’ link between a situation and a planned behaviour
[8,12]. In addition, this is a simple intervention that (e.g., between the offer of a first cigarette and a refusal
could be deployed across most schools (i.e., has wide strategy). Implementation intentions have proved to be
reach) in order to help tackle smoking initiation in ado- effectiveyetsimplemeansofchangingarangeofdifferent
lescent groups. Finally, this intervention is relatively health behaviours [17]. Gollwitzer and Sheeran’s [11]
low cost, requiring around 30 minutes per session to meta-analysis showed that across 94 independent studies
implement (including anti-smoking messages and com- in both laboratory and field settings, implementation
pleting an implementation intention questionnaire). intentions had an average effect size of d+ = 0.65, al-
Such an intervention could offer ‘value for money’ and thoughonly6outofthe94studiesreviewedbyGollwitzer
be rolled out to significant proportions of the adoles- and Sheeran [11] investigated health-risk behaviours and
cent population in the UK. none examined smoking. More recently, research has
Adolescence is the period during which the vast major- begun to emerge to suggest that implementation inten-
ity of smokers take up this habit in the UK [1,3]. As such tions may be effective in promoting smoking cessation.
it represents a key period during which to intervene. A For example, in two field experiments, Armitage [18,19]
considerable number of interventions have been tested in found that implementation intentions caused significantly
this age group. The vast majority of such interventions more smokers to quit (up to 19% quit) compared with
havebeenschool-basedbuthavemetwithonlymixedevi- smokers randomly allocated to active control conditions
denceofeffectiveness(forreviewssee[13,14]).Inpartthis (2% quit), suggesting that the technique has utility in this
maybeattributabletolackofthoroughevaluation.Never- domain.
theless, even among high quality randomised controlled Implementation intentions have also been used in
trialsofschool-basedinterventionstheevidence ismixed. relation to reducing smoking initiation. Two studies
Forexample,information-basedinterventionshavegener- have assessed the impact of forming an implementation
ally been ineffective, while social influence interventions intention about what to say to refuse the offer of a
have tended to show rather mixed effectiveness [13]. The cigarette on subsequent smoking initiation. Higgins &
mosthighlyregardedandlongesttrialofthistypeofinter- Conner [12] reported that among adolescents who
vention was the Hutchinson Smoking Prevention Project formed such an implementation intention (on a single
whichreportednoevidenceforeffectivenessoveraperiod occasion) 0% went on to initiate smoking in the next
of 8 years [15]. The ASSIST project is another well con- two months, whereas 6% did so in a control condition.
ducted trial carried out in the UK to evaluate a peer-led Conner & Higgins [8] more recently reported the results
intervention to reduce smoking initiation [16]. This RCT of an explanatory cluster randomised controlled trial of
did report support for a peer-led intervention over both a repeated implementation intentions in a more appropri-
one-andatwo-yearperiod.Moremixedevidencehasbeen ately powered study. Classes of children were randomly
reportedforinterventionsthattargetsocialcompetenceor allocated to complete implementation intentions about
Conneretal.BMCPublicHealth2013,13:54 Page3of9
http://www.biomedcentral.com/1471-2458/13/54
how to refuse offers of cigarettes (intervention) or We will ask head teachers to sign a written consent
complete their homework (control) on 7 occasions be- form agreeing that their schools take part in the trial
tweentheagesof11–12and13–14years(bothgroupsalso and agreeing not to introduce any new anti-smoking
read simple anti-smoking messages on each occasion). interventions during the time period of the trial without
Researchassistantsdeliveredtheinterventioninclassroom informing the trial staff. We will seek to record carefully
time to whole classrooms of children in approximately 30 any anti-smoking initiatives conducted in each school
minute sessions. After controlling for various known pre- during the trial period. Through the schools we will seek
dictors of smoking initiation (e.g., gender, attitudes to to obtain parental consent (an opt-out procedure) for
smoking, friends and family smoking) and the multi-level participation in the RCT (intervention plus data collec-
nature of the data, rates of self-reported and objectively tion) through lettersfrom theparticipatingschools.
assessed smoking at age 15–16 years were significantly
lower in the intervention compared to the control condi- Studypopulation
tion(d+=0.24and1.04forself-reportedsmokingandob- Proposedsamplesize
jectively assessed smoking respectively). For self-reported In selecting a sample size we have based our estimates
smoking, the unadjusted reduction in smoking initiation on six factors: 1) the size of effect observed in our previ-
was approximately 7% between the intervention and con- ous research; 2) the fact that school will be the unit of
trol conditions, while the unadjusted difference for object- randomisation and adolescents will be clustered within
ivelyassessedsmokingwasapproximately10%. schools; 3) a likely dropout rate based on our previous
research; 4) an estimate of the intra-class correlation co-
Researchquestions efficient (ICC); 5) choice of significance level at 5% and
The following research questions will be addressed in sought power of at least 90%; and 6) a correction for
thistrial: imbalances in recruitment or drop-out rates across
schools. Based on these values we have estimated the
(cid:1) Canrepeatedimplementationintentionsrelatedto power of our study to detect differences between inter-
refusingoffersofcigarettesreducesmoking vention and control arms of the RCT using our pro-
initiationratesin11–16yearoldsrelativeto a posed sample sizes. In relation to effect size, our
control groupofadolescents? explanatory trial [8] indicated that the intervention
(cid:1) Whatisthe cost effectiveness ofsuchan reduced objectively assessed smoking rates from around
intervention? 14% to around 4% (in unadjusted data) in 15–16 year
olds. In the analysis controlling for other predictors of
Methods/Design smoking and the multilevel nature of the data (children
Design clustered within classes, classes clustered within schools)
This study will use a pragmatic cluster randomised con- this equated to a large effect size with an odds ratio of
trolledtrialdesign(seeFigure1).Theunitofrandomisa- 0.15 (95%CI: 0.03–0.80). We have conservatively pow-
tion will be schools. Individual adolescents will be the ered the present study using a value at the upper end of
unit of analysis although we will take account of cluster- this range, i.e., to detect a difference of 5% between the
ingbyschools. intervention (9% rate of smoking initiation) and control
(14% rate of smoking initiation) arms at the end of the
Samplingandrecruitmentofschoolsandadolescents study (i.e., equivalent to an odds ratio of 0.61). In rela-
All secondary schools (except independent schools) in tion to dropout rates we have estimated that 85% of
the Leeds and Staffordshire areas will be invited to take those who begin the study will be available to follow-up
part. There will be a total of at least 36 schools recruited at 48 months [8]. For ICC we have estimated a value of
intotheRCT(seeFigure2).Wewillapproachsecondary 0.01 based on previous research [8]. In relation to sig-
schools in the West Yorkshire and Staffordshire Local nificance level we have used an alpha of 0.05 using a
Education Authorities (the participating regions) and the two-sidedtestinordertobeopentotestingthepossibil-
surrounding areas to request participation in the pro- ity thattheintervention actuallyincreased smoking initi-
posed research. Through previous work and that of our ation rates. In relation to a correction factor to allow for
collaborators we have good experience of recruiting and imbalances in recruitment or dropout we estimate this
working with schools in both regions. Based on our pre- willnot reduce ourstatistical powerbymorethan1%.
vious work [8,12] we anticipate at least 50% of schools Using these assumptions we have estimated that we
will agree to participate. Within each school all available requireatotalof36schoolswith18randomlyallocatedto
classes in year 7 (11–12 year olds) will participate (esti- the intervention arm and 18 to the control arm of the
mated to be at least 4 classes of 30 adolescents per RCT.Withineachschool,weestimatethatweneedtoap-
school). proachatleast120adolescentsatbaseline(approximately
Conneretal.BMCPublicHealth2013,13:54 Page4of9
http://www.biomedcentral.com/1471-2458/13/54
Intervention Condition Control Condition
Smoking Prevention Pro-homework
Data collection (baseline) Data Data collection (baseline)
September 2012– January 2013 collection September 2012–January 2013
sessions
Questionnaire
Teacher First teacher delivered session & carbon First teacher delivered session Teacher
delivered December 2012–February 2013 monoxide December 2012–February 2013 delivered
sessions: (CO) test sessions:
Students completion. Students
read Second teacher delivered session Second teacher delivered session read pro-
smoking May–June 2013 Duration: May–June 2013 homework
prevention approximately persuasive
persuasive 60 minutes. messages
messages and
and DSaetap tceomllbeecrt–ioDne c(1e2m mbeorn 2t0h1s3) DSeaptate cmoblleerc–tiDonec (e1m2bm eor n2t0h1s3) complete
complete implementation
implementation intention –
intention – worksheets.
worksheets. Third teacher delivered session Third teacher delivered session
November–December 2013 November–December 2013 Duration: 30
Duration: 30 minutes.
minutes.
Fourth teacher delivered session Fourth teacher delivered session
May–June2014 May–June2014
Data collection (24 months) Data collection (24 months)
September –December 2014 September –December 2014
Fifth teacher delivered session Fifth teacher delivered session
November–December 2014 November–December 2014
Sixth teacher delivered session Sixth teacher delivered session
May – June 2015 May – June 2015
Data collection (36 months) Data collection (36 months)
September–December 2015 September–December 2015
Seventh teacher delivered session Seventh teacher delivered session
November–December 2015 November–December 2015
Eighth teacher delivered session Eighth teacher delivered session
May–June2016 May–June2016
Data collection (48 months) Data collection (48 months)
September–December 2016 September–December 2016
Figure1FlowChartofStudyProceduresforbothGroups.
4 classes per school). Based on an 85% recruitment rate
(due to refusal to participate and non-attendance on test-
ing days) the overall number of adolescents available for
analysis will be: 0.85 * 18 * 120 = 1836 per group. The
effectivenumberofadolescentspergroupwillbereduced
Eligible Schools furtherduetoadesigneffect(i.e.,adolescentsclusteredby
school): Design effect = 1 + (m – 1)*ICC, where m is the
number of participants per cluster. This yields a design
Randomization Excluded
effect of 2.01 and reduces the effective sample size per
N>36 schools
groupto913(1836/2.01).Thiseffectivesamplesizeyields
a statistical power of 90.65% to detect a 5% difference in
smokinginitiationbetweentheintervention(9%initiation)
and control (14% initiation) arms using a two-sided test
Intervention Group Control Group
Smoking prevention Pro-homework condition with an alpha of 0.05. Adding our correction factor for
condition
N >18 schools potential imbalances in recruitment would still provide a
N >18 schools
powerofalmost90%.Thereforeourproposedsamplesize
Figure2Randomizationflowchart.
is 4320 adolescents from 36 schools (18 intervention; 18
Conneretal.BMCPublicHealth2013,13:54 Page5of9
http://www.biomedcentral.com/1471-2458/13/54
control), with approximately 120 adolescents being the Committee as reason to stop the trial. All adverse
approachedperschool. eventswould alsobereportedtotheTSC.
Discontinuationcriteria Randomisation
Analyses will follow intention-to-treat (ITT) principles The unit of randomisation will be schools and our sample
as far as possible. Therefore we will include in the ana- size calculations and analyses takes account of this cluster-
lyses all schools and children initially randomised who ing of the data. We believe randomising by school before
provide data at baseline, including them for the purpose the intervention begins is most appropriate in order to re-
of analysis in the group originally allocated to them. To ducethelikelihoodofcontaminationbetweentheinterven-
maximise the power of such analyses all reasonable and tion and control conditions (randomising by individual or
ethical steps will be taken to ensure the completeness of classroom risks such contamination). TheTrial Statistician
follow-up measures. Where data are missing we will use (RW) will use a random number generator to randomly
multiple imputation based on baseline data to replace allocate schools to the control or intervention arms of the
missing data [20]. Children missing at baseline but trial. Schools have requested feedback on which condition
present at one or more subsequent data collection time they are randomised to before giving final agreement to
pointswillbetreatedas‘otherparticipants’andexcluded participateandagreeingtosendoutletterstoparents.The
from theintention-to-treatanalyses. trial will therefore deviate from best practice in recruiting
childrentothetrialafterratherthanbeforerandomisation.
Schoolswithdrawal Wewill monitor anyconsequential effectson participation
If a school wishes to withdraw from the study we will rates. As we hope to recruit all classes in a school year
attempt to collect data on the reasons for withdrawal. We movementbetweenclasseswithinaschoolshouldnotbea
will explorewiththeschoolthepossibilityofusingall data problem. However, adolescents moving school during the
collected up until that point and the possibility of still RCTare likely to have missing data. Of those initially par-
collecting data from pupils at the school at the final time ticipating in the study at age 11–12 years we anticipate a
point in order to maximise the chance of having more cumulative loss to follow-up at age 15–16 years of 15%
complete data on our key outcome measures (an ITT (basedonourpreviousresearch,[8]).Wewillmakeefforts
principle).Provideddataisavailablefrombaseline,multiple to minimize drop out (e.g., attempting to ensure each
imputation will be used to replace missing data. Where no school is visited on more than one day to collect smoking
baseline data are available all data will be excluded from measures)andanypotentialbiasesattributabletodropout
analyses as there will be insufficient data to appropriately willbeexploredstatistically.
performanymultipleimputationofmissingvalues.
Compliancewithgoodpractice
Childwithdrawal
Parents will have the opportunity to have their child not (cid:1) Allstatisticalanalysesofprimaryand secondary
included in the study before commencement. Parents outcomes willbe carriedoutunderthe guidance of
will also have the opportunity to withdraw their child thetrialstatistician(RW).
from the study at any time point by informing the (cid:1) HTAguidelinesformissingdatainRCTswillbe
school. Further data will not be collected from children followed[20].
withdrawn from the study (missing data will be replaced (cid:1) CONSORTguidelines[21,22]forpresentationof
based on multiple imputation from available data). As resultsfromclusterrandomisedtrialswillbe
data will be collected anonymously, it will not be pos- followed.
sible to withdraw data already collected. Children with- (cid:1) Presentationofresultswillconform togoodpractice
drawn from the study may not be left out of any whole forpresentationofcomplex interventions [23].
class activities that form part of the intervention as to (cid:1) Theflowofclusters and individuals through the
do so might involve taking the child out of the class trial,fromassignment toanalysis,willbepresented
whilsttheseactivities wereoccurring. usingaflowchart,inaccordancewith CONSORT
guidelines [21,22].
Interimanalysisandstoppingrules (cid:1) Intra-classcorrelationcoefficientsfrom the
Overallsmokingratesinthetwoconditionswillbecom- multilevelanalyseswillbepresentedfollowinggood
puted and presented to the Independent Trial Steering practice forclusterrandomised trials.
Committee (TSC; and associated data monitoring com-
mittee) each year. Evidence of considerably higher rates Ethicalconsiderations
of smoking initiation in the intervention condition com- Ethical approval was obtained through the University of
pared to the control condition could be considered by LeedsInstituteofPsychologicalSciencesethicscommittee
Conneretal.BMCPublicHealth2013,13:54 Page6of9
http://www.biomedcentral.com/1471-2458/13/54
(Referencenumber12–0155).TheUniversityofLeedswill Intervention
act as trial sponsor. Written informed consent will be This will be a phase III pragmatic randomised controlled
obtainedfromallparticipatingschools.Theinformedcon- trial conducted in two geographically distinct regions of
sent of participants’ parents will be soughtby letter using England (West Yorkshire and Staffordshire) that will pro-
anopt-outapproach,i.e.parentswillhavetheopportunity vide the information on which to base a decision about
to“optout”ofthestudyifthey donotwishtheirchild to rolloutoftheintervention.Theinterventioninvolvesask-
take part. Children will receive information sheets about ingadolescentstoreadsimpleanti-smokingmessagesand
the study and will have the opportunity to not complete then to form an implementation intention by planning
themeasures. how, where, and when to resist smoking (i.e., refuse the
offer of a cigarette). This intervention will be repeated
every six months over a period of four years between the
Datacollectionmethods agesof11–12and15–16years(8occasions;seeflowchart
Theprimaryoutcomemeasureissmokinginitiationbased of study procedures in Figure 1). The anti-smoking mes-
onobjectivemeasuresat48months.Asecondarymeasure sagesfocusonsimple-to-understandnegativeconsequences
will be self-reported smoking. These measures will also be of starting smoking and the positive consequences of
assessed at baseline and at 12, 24, and 36 months post- remaining a never-smoker. The implementation inten-
baseline in order to explore the time periods over which tion intervention is designed to give adolescents simple
theinterventionmightbebesttargeted.Thesedatacollec- responsesforhowtorefuseacigaretteandtolinkthemto
tion sessions will be completed in classroom time by situationswhereacigarettemightbeoffered.Fiveoptions
the researchers and take approximately 30 minutes. Self- will be provided for how they could refuse the offer of a
reported smoking and measures of cognitions (e.g. atti- cigarette or resist the temptation to smoke (e.g., ‘No
tudestosmoking,intentionstoresistsmoking,self-efficacy thanks, I don’t want to smoke; No thanks, I don’t want
overresisting,familysmokingandfriendssmoking)willbe the habit; No cancer sticks for me; No thanks, smoking
collected via questionnaire completed individually but in a makesyousmellbad;No,it’sbadforyourhealth’;see[8]).
classroomsetting.Anobjectivemeasureofsmokingwillbe Participantsarerequiredtochecktheoptionstheyplanto
completed individually outside a classroom setting. A useortowriteinanadditionalresponse.Similarlypartici-
numberofmeasuresofobjectivesmokingarefeasible.The pants will be required to check where they would not
two most common measures are breath carbon monoxide smoke(e.g.,‘Iwillnotsmokeatschool;Iwillnotsmokeat
measures and saliva cotinine measures. Although saliva home; I will not smoke at a party; I will not smoke with
cotinine may be more effective in detecting occasional my friends; I will not smoke if offered a cigarette’) and
smoking(duetothegreaterhalf-lifeof16–18hoursversus whentheywouldnot smoke(e.g.,‘IthinkIcanmakesure
4–6 hours in breath carbon monoxide) it also has a num- I don’t smoke this term’) and signal agreement with their
ber of disadvantages. These include lower acceptability to plan not to smoke by ticking a box. The implementation
adolescents, greater cost, and loss of ability to feed back intentionwillbepresentedasapaperandpencilquestion-
theresultstoparticipantsimmediately.Forthisreasonour naire that adolescents read and complete in classroom
objectivemeasureofsmokingwillbeobtainedfrombreath time. Previous evidence has suggested that implementa-
W
carbon monoxide monitors (Micro+ Smokerlyzer CO tion intentions are more effective among individuals who
Monitor, Bedfont Scientific Limited, Kent, England). This are motivated to perform the behaviour [17]. Therefore
instrument gives a measure of carbon monoxide in the combining the implementation intention with motivatio-
breath in parts per million (ppm) accurate to within 2% nal messages is likely to be more effective than using
based upon exhaling one breath into the device and has implementation intentions alone. This was the combin-
been adapted for use in adolescent samples. Although a ation used in our previous research [8,12]. The research
number of factors influence carbon monoxide in the has broadly followed the MRC framework for developing
breath, recent smoking should significantly elevate levels. and evaluating complex interventions to improve health
Carbon monoxide in the breath can be used as a reliable [29,30]. The pre-clinical through phases II testing of this
and valid measure of exposure to cigarette smoking [24], repeatedimplementationintentioninterventionhavebeen
comparable in accuracy to blood carboxyhaemoglobin completed as part of this previous work. In particular,
levels [25]. The Bedfont devices have been demonstrated significant workhas been completedtodevelop the inter-
to give reliable and valid assessments of smoking status vention and test its feasibility and acceptability in appro-
[26] and have been employed by various researchers with priate samples of schoolchildren (see [12]). In the control
adolescent samples [8,27]. Our secondary outcome meas- condition participants will read persuasive messages en-
ure, self-reported smoking, will be assessed by standard couragingthemtocompletealltheirhomeworkandthey
measures (as used in national surveys and our previous willformanimplementationintentioninrelationtocom-
research;[8,28]). pletingalltheirhomework.
Conneretal.BMCPublicHealth2013,13:54 Page7of9
http://www.biomedcentral.com/1471-2458/13/54
The intervention will take place in classroom time and imputation, based on regression methods, will be under-
will be led by a teacher using materials provided by the taken to ‘complete’ the data and ensure efficient analysis.
researchers. The materials will include written simple This will be done for all participants for whom baseline
anti-smokingmessages(asinour previousresearch, these data is available (to provide meaningful imputations) and
will be varied between sessions to reduce boredom) and will take into consideration the two-level hierarchical
paper versions of the implementationintentions sheets to nature of the data. With this approach the results will be
be completed. The time required will be no longer than unbiased given the general assumption of missing at ran-
60minutestoincludeanintroductionbytheteacher,brief dom (MAR) rather than the more restrictive assumption
classroom discussion, reading of the anti-smoking mes- of missing completely at random (MCAR). Results from
sagesandcompletionofthepaperandpencilimplementa- the multiple imputations will be combined using Rubin’s
tion intention task. We will negotiate with schools about rules.
theleastdisruptiveandmostappropriateteachingsession Subgroup analyses will be used to examine any differ-
inwhichtoimplementthesesessions(e.g.,morningregis- ences between high and low deprived schools, between
tration).Teacherswillbeprovidedwithaone-to-twohour regions (West Yorkshire versus Staffordshire), and be-
training session to introduce them to implementation tween boys and girls. We will use mediation analyses to
intentions and anti-smoking messages, and to provide explore the extent to which any intervention effects are
them with advice and written information on how to run mediated by changes in the measured cognitions as
the intervention sessions. This training will be conducted tested in our previous work (e.g., attitudes to smoking,
by the research team and teachers will be paid for their intentions to resist smoking, self-efficacy to not smoke,
time.Wewillnegotiatewithschoolsonwhetherthesame etc.; [31-34]).
or different teachers will run the sessions throughout
the trial and it may be necessary to train some additional Economicanalyses
teacherseachyear. We will also conduct an economic evaluation of the
intervention. The study will take the perspective of the
Interventionfidelity service providers, health and social care sectors. Costs
Wewillassessadherencetotheintervention intwo ways. will include the resources used in service provision and
First, approximately 10% of intervention sessions will be will be collected using an adapted version of a form
assessed for adherence to protocol (a mixture of self- developed by the Health Economics Group at the
report by teachers and observation by researchers). Sec- University of Leeds for an on-going trial with a popula-
ond, all written implementation intentions sheets will be tion of adolescents. Costs will also include intervention
collectedbyteachersandpassedtoresearchstaffandthey time, resources used on training and materials. First, the
will be analysed for completeness. In our previous re- economic sub-study aims to identify the within study
search [8] 73% of adolescents completed implementation incremental cost effectiveness ratio (ICER); the costs
intentions sheets on at least 6 occasions (out of a total of and benefits of the intervention compared to the control
8 occasions). In addition, the number of occasions imple- group. The analysis will use smoking rates in the two
mentation intentions sheets were completed was unre- arms ofthe trial, and ‘cost per non-initiation of smoking’
lated to reduced smoking initiation. In the present trial willbeestimated. Second, it willadopt alifetime horizon
the impact of number of complete and partially complete and cost per life year gained will be estimated from a
implementation intentions sheets on rates of objective decision analytic cost effectiveness model. As far as pos-
and self-reported smoking will be examined statistically. sible parameters in the model will be specified using
Both sets of data will inform the basis of an intervention data collected within the trial. Data on resource use
fidelityanalysis. costs and quality of life will be collected. Other para-
meters, such as the long term ‘natural history’ will be
Analyses parameterised using the published literature, and where
Analysesofprimaryandsecondaryoutcomemeasures necessary formally elicited expert opinion. The outcome
The study uses a two group experimental design with measure for these analyses will be life years gained [35].
schools of adolescents (all aged 11–12 years at outset) The calculation of QALY’s will also be explored, using a
randomly allocated to one of two conditions: a repeated measure of quality of life. In the analysis, the non-
implementation intention condition targeting smoking or parametric bootstrap method will be used to produce a
a control condition. The primary outcome will be object- within trial probabilistic sensitivity analysis of the ICER.
ively assessed smoking and the secondary outcome meas- A scatterplot of the cost effectiveness plane, the 95%
ure will be self-reported smoking. A two-level logistic cost effectiveness elipse and cost effectiveness accept-
regression will be used to model smoking outcomes at ability curve will be presented. Discounting will be
48 months. Given that drop outs are expected, multiple undertakenusingtherate recommendedatthe time.
Conneretal.BMCPublicHealth2013,13:54 Page8of9
http://www.biomedcentral.com/1471-2458/13/54
Discussion 8. ConnerM,HigginsA:Long-termeffectsofimplementationintentionson
Reducingsmokinginitiationispotentiallythemosteffect- preventionofsmokinguptakeamongadolescents:Aclusterrandomized
controlledtrial.HealthPsychol2010,29:529–538.
ive way to reduce smoking-related harm. The present
9. RolandM,TorgersonDT:Understandingcontrolledtrials:Whatare
paperreportstheprotocolforaninterventiondesignedto pragmatictrials?BrMedJ1998,316:285.
reducesmokinginitiationinadolescents.Theintervention 10. GollwitzerP:Goalachievement:Theroleofintentions.EurRevSocPsychol
1993,4:141–185.
involves the repeated formation of implementation inten-
11. GollwitzerPM,SheeranP:Implementationintentionsandgoal
tions (i.e. if-then plans) about how to refuse offers of achievement:Ameta-analysisofeffectsandprocesses.AdvExpSoc
cigarettes Promising initial data have been collected on Psychol2006,38:69–119.
12. HigginsA,ConnerM:Understandingadolescentsmoking:Theroleofthe
the efficacy of this intervention [8]. The current cluster
TheoryofPlannedBehaviourandimplementationintentions.Psychol
randomisedcontrolledtrialwillcollectfurtherdataonthe HealthMed2003,8:177–190.
effectivenessoftheintervention.Thisisasimpleinterven- 13. ThomasRE,PereraR:School-basedprogrammesforpreventingsmoking.
CochraneDatabaseSystRev2006,3(Art.No.CD001293).doi:10.1002/
tion that could be deployed across most schools (i.e., has
14651858.CD001293.pub2.
wide reach) in order to help tackle smoking initiation in 14. WieheSE,GarrisonMM,ChristakisDA,EbelBE,RivaraFP:Asystematic
adolescents. In addition, this intervention is relatively low reviewofschool-basedsmokingpreventiontrialswithlong-termfollow-
up.JAdolescHealth2005,36:162–169.
cost, requiring around 30-60 minutes per session to im-
15. PetersonAV,KealeyKA,MannSL,MarekPM,SarasonIG:Hutchinson
plement,andcanbeimplementedbyteachers.Shouldthe
smokingpreventionproject:Long-termrandomizestrialinschool-based
intervention prove effective in changing smoking out- tobaccouseprevention–resultsonsmoking.JNatlCancerInst2000,
comes and prove ‘value for money’ it might usefully be 92:1979–1991.
16. CampbellR,StarkeyF,HollidayJ,AudreyS,BloorM,Parry-LangdonN,
rolled out to significant proportions of the adolescent
HughesR,MooreI:Aninformalschool-basedpeer-ledinterventionfor
populationintheUK. smokingpreventioninadolescence(ASSIST):Aclusterrandomizedtrial.
Lancet2008,371:1595–1602.
Competinginterests 17. SheeranP,MilneS,WebbTL,GollwitzerPM:Implementationintentions
Theauthorsdeclarethattheyhavenocompetinginterests. andhealthbehaviours.InPredictinghealthbehaviour.2ndedition.Edited
byConnerM,NormanP.Maidenhead:OUPress;2005:276–323.
Authors’contributions 18. ArmitageCJ:Efficacyofabriefworksiteinterventiontoreducesmoking:
MCconceivedthestudy.MC,SG,RL,CA,RW,KS,BGandCTwereapplicants Therolesofbehavioralandimplementationintentions.JOccupHealth
forthefundingandinvolvedindesigningthestudy.MCledthedraftingof Psychol2007,12:376–390.
theprotocolwithinputfromallauthors.Allauthorsreadandapprovedthe 19. ArmitageCJ:Avolitionalhelpsheettoencouragesmokingcessation:
finalprotocol. Arandomizedexploratorytrial.HealthPsychol2008,27:557–566.
20. CarpenterJR,KenwoodMG:Missingdatainclinicaltrials-apracticalguide.
Acknowledgements Birmingham:NationalInstituteforHealthResearch;2008.PublicationRM03/
ThepresentresearchisfundedbyagrantfromtheMedicalResearch JH17/MK.
Council/NationalPreventiveResearchInitiative(referenceMR/J000264/1)to 21. CampbellM,ElbourneD,AltmanD:CONSORTstatement:Extensionto
MarkConner,SarahGrogan,RebeccaLawton,ChristopherArmitage,Robert clusterrandomisedtrials.BrMedJ2004,328:702–708.
West,KamranSiddiqi,BrendaGannon,andCaroleTorgerson.KeiraFlettand 22. CampbellM,PiaggioG,ElbourneDR,AltmanDG:CONSORT2010statement:
RuthSimms-Ellisareemployedonthegrant. Extensiontoclusterrandomisedtrials.BrMedJ2012,345:e5661.
23. PereraR,HeneghanC,YudkinP:Graphicalmethodsfordepicting
Authordetails randomisedtrialsofcomplexinterventions.BrMedJ2007,334:127–129.
1InstituteofPsychologicalSciences,UniversityofLeeds,LeedsLS29JT,UK. 24. StookeyGK,KatzBL,OlsonCA,DrookSL,CohenSJ:Evaluationof
2DepartmentofPsychology,SportsandExercise,StaffordshireUniversity, biochemicalvalidationmeasuresindeterminationofsmokingstatus.
Stoke-on-TrentST42DF,UK.3DepartmentofPsychology,Universityof JDentRes1987,66:1597–1601.
Manchester,ManchesterM139PL,UK.4InstituteofHealthSciences, 25. JarvisMJ,Tunstall-PedoeH,FeyerabendC,VeseyC,SaloojeeY:Comparison
UniversityofLeeds,LeedsLS29JT,UK.5DepartmentofHealthSciences, oftestsusedtodistinguishsmokersfromnon-smokers.AmJPublic
UniversityofYork,YorkYO105DD,UK.6SchoolofEducation,Durham Health1987,77:1435–1438.
University,DurhamDH11TA,UK. 26. IrvingJM,ClarkEC,CrombieIK,SmithWCS:Evaluationofaportable
measureofexpired-aircarbonmonoxide.PrevMed1988,17:109–115.
Received:14December2012Accepted:11January2013 27. ZackM,BelsitoL,ScherR,EissenbergT,CorrigallWA:Effectsofabstinence
Published:19January2013 andsmokingoninformationprocessinginadolescentsmokers.
Psychopharmacology2001,153:249–257.
References 28. JarvisL:Smokingamongsecondaryschoolchildrenin1996.England.London:
1. CancerResearchUK:Smokingstatistics.;2012.[On-line]Available:http://info. HMSO;1997.
cancerresearchuk.org/cancerstats/types/lung/smoking/(accessed12.11.12). 29. CampbellM,FitzpatrickR,HainesA,KinmonthAL,SandercockP,
2. DepartmentofHealth:Healthylives,healthypeople–Atobaccocontrolplan SpiegelhalterD,TyrerP:Frameworkfordesignandevaluationofcomplex
forEngland.London:HMSO;2011. interventionstoimprovehealth.BrMedJ2000,321:694–696.
3. ASH:Factsataglance.;2011.[On-line].Available:http://www.ash.org.uk/ 30. CraigP,DieppeP,MacintyreS,MichieS,NazarethI,PetticrewM:
files/documents/ASH_93.pdf(accessed28.4.11). Developingandevaluatingcomplexinterventions:ThenewMedical
4. PaavolaM,VartiainenP,PuskaP:Predictingadultsmoking:Theinfluence ResearchCouncilguidance.BrMedJ2008,337:979–983.
ofsmokingduringadolescentandsmokingamongfriendsandfamily. 31. ConnerM,GroganS,FryG,GoughB,HigginsAR:Direct,mediatedand
HealthEducRes1996,11:309–315. moderatedimpactsofpersonalityvariablesonsmokinginitiationin
5. RoyalCollegeofPhysicians:Smokingandtheyoung.Suffolk:Lavenham adolescents.PsycholHealth2009,24:1085–1104.
Press;1992. 32. ConnerM,SandbergT,McMillanB,HigginsA:Roleofanticipatedregretin
6. GeneralHouseholdSurvey:2007report.London:ONS;2009. adolescentsmokinginitiation.BrJHealthPsychol2006,11:85–101.
7. FullerE,SanchezM:Smoking,drinkinganddruguseamongyoungpeoplein 33. GroganS,ConnerM,FryG,GoughB,HigginsA:Genderdifferencesin
Englandin2009.London:NHSInformationCentreforHealthandSocial smoking:Alongitudinalstudyofbeliefspredictingsmokingin11–15
Care;2010. yearolds.PsycholHealth2009,24:301–316.
Conneretal.BMCPublicHealth2013,13:54 Page9of9
http://www.biomedcentral.com/1471-2458/13/54
34. LawtonR,ConnerM,ParkerD:Beyondcognition:Predictinghealthrisk
behaviorsfrominstrumentalandaffectivebeliefs.HealthPsychol2007,
26:259–267.
35. RaikouM,McGuireA:Areviewofthecost-effectivenessofinterventions
(specificallypointofsalesmeasuresandmassmedia)topreventtheuptakeof
smokinginyoungpeopleunder18yearsold,NICE.;2007.
doi:10.1186/1471-2458-13-54
Citethisarticleas:Conneretal.:Studyprotocol:Aclusterrandomised
controlledtrialofimplementationintentionstoreducesmoking
initiationinadolescents.BMCPublicHealth201313:54.
Submit your next manuscript to BioMed Central
and take full advantage of:
• Convenient online submission
• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at
www.biomedcentral.com/submit