RESEARCH REPORT

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction,

98

, 837–846

Blackwell Science, Ltd

Oxford, UK

ADDAddiction

1360-0443© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

98Original Article

Personalized cessation adviceRon Borland et al.

Correspondence to:

Ron BorlandVicHealth Centre for Tobacco ControlThe Cancer Council Victoria11 Rathdowne StreetCarlton,Victoria 3053AustraliaTel: 61 3 9635 5185Fax: 61 3 9635 5440E-mail: ron.borland@cancervic.org.au

Submitted 23 March 2002;initial review completed 29 July 2002;

final version accepted 3 March 2003

RESEARCH REPORT

The effectiveness of personalized smoking cessation strategies for callers to a Quitline service

Ron Borland

1

, James Balmford

1

, Catherine Segan

1

, Patricia Livingston

2

& Neville Owen

3

VicHealth Centre for Tobacco Control, Victoria,

1

Centre for Behavioural Research in Cancer, The Cancer Council Victoria

2

and Cancer Prevention Research Centre,

University of Queensland, Australia

3

ABSTRACT

Aim

To assess the effectiveness of a program of computer-generated tailoredadvice for callers to a telephone helpline, and to assess whether it enhanced aseries of callback telephone counselling sessions in aiding smoking cessation.

Design

Randomized controlled trial comparing: (1) untailored self-help mate-rials; (2) computer-generated tailored advice only, and (3) computer-generatedtailored advice plus callback telephone counselling. Assessment surveys wereconducted at baseline, 3, 6 and 12 months.

Setting

Victoria, Australia.

Participants

A total of 1578 smokers who called the Quitline service andagreed to participate.

Measurements

Smoking status at follow-up; duration of cessation, if quit; useof nicotine replacement therapy; and extent of participation in the callbackservice.

Findings

At the 3-month follow-up, significantly more (

c

2

(2)

=

16.9;

P

<

0.001) participants in the computer-generated tailored advice plus tele-phone counselling condition were not smoking (21%) than in either the com-puter-generated advice only (12%) or the control condition (12%). Proportionsreporting not smoking at the 12-month follow-up were 26%, 23% and 22%,respectively (NS) for point prevalence, and for 9 months sustained abstinence;8.2, 6.0, and 5.0 (NS). In the telephone counselling group, those receiving call-backs were more likely than those who did not to have sustained abstinence at12 months (10.2 compared with 4.0,

P

<

0.05). Logistic regression on 3-monthdata showed significant independent effects on cessation of telephone counsel-ling and use of NRT, but not of computer-generated tailored advice.

Conclusion

Computer-generated tailored advice did not enhance telephonecounselling, nor have any independent effect on cessation. This may be due topoor timing of the computer-generated tailored advice and poor integration ofthe two modes of advice.

KEYWORDS

Computer tailoring, quitlines, smoking cessation, telephone

counselling, tobacco control, transtheoretical model.

INTRODUCTION

Services to smokers can range from broadly directedencouragement to quit in mass media campaigns,through provision of printed self-help materials, to inten-

sive group or individual-based cessation therapies. Asintensity increases, effectiveness increases, but at somepoint interventions are likely to become too expensiveand/or too demanding on potential users to be dissemi-nated widely. For example, barriers to participation in fee-

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction,

98

, 837–846

838

Ron Borland

et al.

paying services will result in lower rates of uptake, whileintensive services provided by government or other bodiesat no cost to users may not be sustainable.

Strategies to deliver effective cessation services tolarge numbers of smokers at a low cost can provide amajor public health benefit. The development of suchstrategies can be guided by stepped-care models for smok-ing cessation, which identify a range of different intensityinterventions that can be applied in response to the needsof different subgroups of smokers (Reid

et al

. 1992;Abrams 1993; Abrams, Clark & King 1999). Low-ratesmokers may be able to quit with minimal assistance(Owen

et al

. 1995), while higher-rate smokers mayrequire more intensive forms of help (Wilson

et al

. 1992).Those who have a history of repeated failure may alsorequire more assistance than those only recently moti-vated to try. Complementary to this, many smokers iden-tify a preference for personalized assistance (Owen &Davies 1990; Clarke

et al

. 1993), but few seek intensivehelp such as face-to-face counselling or group-based pro-grammes (Brown & Owen 1990; Owen 1989). This studylooks at the potential of repeated computer-generatedtailored advice to callers to a reactive smoking cessationhotline service as an alternative and complement to pro-active telephone callbacks, both interventions with someevidence of efficacy.

Reactive smoking cessation helplines, offering per-sonal counselling and self-help materials to a large pop-ulation, appear to be efficacious and useful as a publicintervention (Orleans

et al

. 1991; Ossip-Klein

et al

. 1991;Stead & Lancaster 2002). Recent descriptive evaluationsof telephone Quitlines in Australia and the United King-dom have shown high quit rates (point prevalent absti-nence in excess of 20% at 12 months) and strongevidence of positive behavioural change among continu-ing smokers (Platt

et al

. 1997; Wakefield & Miller 1999;Owen 2000). Provision of a callback service is moreexpensive than a reactive helpline, but can increase quitrates over single-call services, although not invariably(Stead & Lancaster 2002). There is heterogeneity in theresults, suggesting some programs may work while oth-ers are ineffective; perhaps the quality and/or timelinessof the calls may affect outcomes. Zhu

et al

. (1996, 2000)found that a dedicated callback protocol approximatelydoubled cessation over controls who received only onecounselling session. Callbacks were relapse-sensitive, tai-lored to the client’s needs and decreased in frequency fol-lowing successful cessation. Subsequent work confirmsthe effectiveness of the Zhu

et al.

program, which is pro-vided as part of a service in California (Zhu

et al.

2000,2002). We also tested a callback protocol, provided toVictorian (Australia) smokers, in which counsellingcalls were provided at strategic points around quitting(Borland

et al

. 2001); we found a significant benefit of

callback counselling on sustained abstinence above aninitial call-in. This seemed to be mainly due to reducedrelapse over the period the callbacks were provided.

Recent advances in computer technology have intro-duced the potential for tailoring written advice, to meetthe specific needs of smokers who are trying to quit with-out the assistance of a counsellor. There is evidence thatinformation tailored in this way can be more appealingand is generally attended to more closely than non-tailored information (Brug 1999).

Prochaska

et al

. (1993; also Velicer

et al

. 1993) pio-neered the use of sophisticated computer-tailored feed-back for smoking cessation and demonstrated that itcould be particularly effective in drawing smokerstowards quitting. Recently, Lancaster & Stead (2000)reviewed eight randomized trials of tailored printed mate-rials compared with standard self-help materials, calcu-lating a combined odds ratio of 1.41 (1.14–1.75) forcessation. Such interventions can clearly work. A prob-lem with the use of meta-analysis in this context is that itcan combine programmes with quite different levels ofquality of advice and/or schedules of advice.

Several studies (Prochaska

et al

. 1993; Velicer

et al

.1993) have demonstrated the effectiveness of tailoringadvice to personal characteristics of the recipient, accord-ing to stage of change and other aspects of the Trans-theoretical model of behaviour change (Prochaska,DiClemente & Norcross 1992). It is unclear what aspectsof tailoring are important in determining the effective-ness of tailored materials. Velicer

et al

. (1999) investi-gated the effect of multiple tailored messages deliveredover repeated contacts, compared with a single tailoredmessage. Similar increases in abstinence rates were foundover a stage-based self-help manual, regardless of thenumber of contacts. No clear evidence, however, wasfound for a dose–response relationship as the number ofcontacts increased from one to six, suggesting that thebeneficial effect of tailoring is not a result of repeated con-tact. None the less, Dijkstra and colleagues (1998, 1999)have shown a beneficial effect of multiple tailoring com-pared with single tailoring among precontemplativesmokers, with regard to stage transition and intention toquit.

The efficacy of tailored materials as a supplement tonicotine replacement therapy (NRT) has also been evalu-ated by Shiffman

et al

. (2000, 2001). In both trials theaddition of a computer-tailored intervention (consistingof six tailored mailings sent over a 12-week period) signif-icantly improved quit rates over nicotine replacementtherapy and self-help materials alone. In the earlier trial,Shiffman et al. (2000) found no added benefit of one out-bound telephone counselling call 2 days after the partic-ipant’s quit date.

The aim of this research was to compare the impact of

Personalized cessation advice

839

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction,

98

, 837–846

standardized Australian printed self-help cessation mate-rials (as a control condition) with a modified version ofthe personalized computer-generated advice program ofProchaska

et al

. (1993) used in conjunction with thestandardized printed materials. We also examinedwhether adding personalized telephone callback counsel-ling increased the effectiveness of the package includingthe computer-generated tailored advice.

METHODS

The participants were callers to the Victorian Quitlinetelephone counselling and advice service, which isavailable to people who are interested in quitting or inmaking inquiries about the process of quitting. The ser-vice is promoted primarily through various media (tele-vision, radio and print) and by health service providers,such as general practitioners. Callers eligible for thestudy were adult cigarette smokers able to communicatein English and with no obvious psychiatric or neurologi-cal problems.

The study was a randomized controlled trial. Callerswho were eligible to participate in the study wererecruited into the trial after their proximal reasons forcalling were dealt with, and after the quit services avail-able to them, which included the callback service (but notthe computer advice), had been outlined. All callers wereoffered a set of printed self-help materials. The small num-ber of callers who sought the callback service proactivelyat this point were not eligible for the study. Eligible smok-ers were then invited into a study of factors that affectsmoking cessation and were told that it involved complet-ing a short survey at the time, and three others over thenext year.

Unlike traditional placebo-blinded trials, participantswere not told about any intervention unless they were ina condition where they were to be offered one, and werethen told only about what they were offered. The disad-vantage of this procedure is that a proportion of partici-pants do not take up the offer, potentially diluting effectsizes. The strength is that there is no need to tell partici-pants in other groups what they are missing out on. Webelieve, and our Institutional Ethics Committee con-curred, that it is both ethically and methodologicallyproblematic to tell participants about all conditions whenone or more seems to be less effective. It can create exactlythe kinds of expectancies placebo double-blinded trialswere designed to overcome.

Once consent to participate in the trial was obtained,the baseline smoking assessment interview was adminis-tered. On completion of the 15-minute interview, smok-ers were allocated randomly to one of three levels ofservice: (i) no extra help, so only the printed self-help ces-

sation materials that are routinely available; (ii) the addi-tional offer of a series of three computer-generatedtailored letters; or (iii) the additional offer of the series ofthree computer-generated tailored letters, and the call-back counselling service. Randomization occurred byshuffling questionnaires; there was no opportunity forthe interviewers to influence the choice of condition. Thisprocedure resulted in all study participants being offeredprinted self-help cessation materials (a ‘Quit pack’). TheQuit pack consisted of a 30-page full-colour A5

Can QuitBook

, which is organized around stage of change; leafletswhich promoted the Quitline callback service and group-based ‘Fresh Start’ smoking cessation courses; and a busi-ness-sized card with strategies to resist cravings. The typ-ical initial call (before recruitment into the trial) rangedfrom the extremely brief (all they wanted was printedmaterial) to 10–15 minutes of reactive counselling focus-ing usually on ambivalence or perceived challenges. Wedo not have records as to who received what. Subsequentwork suggests that less than half would have received anycounselling.

The computer-generated tailored advice was basedon the Transtheoretical model of behaviour change(Prochaska

et al

. 1992, 1993), on which all callers wereassessed during the baseline 15-minute telephone inter-view, and by mail at 3 months and 6 months. Personal-ized letters were mailed as soon as possible following thebaseline, 3- and 6-month assessments. Responses to theassessment were entered on scan forms, which were thenscanned for entry into the tailoring program. The pro-gram used was a modified version of the program devel-oped by Prochaska

et al

. (1993) and shown to be effectivein a number of randomized trials (Prochaska

et al

. 1993;Velicer

et al

. 1999; Prochaska

et al

. 2001). The mostnotable change was the deletion of the social liberationadvice (as the scale did not behave as theorized in devel-opmental work). Pilot studies were used to adapt theassessment tool and computer-generated advice for use inan Australian context

.

Where existing phrases were per-ceived to be colloquial, some wording changes were madeto give the material an Australian feel (Borland, Segan &Velicer 2000). The computer-generated advice took theform of a two- to three-page, single-spaced letter thatdescribed the caller’s stage of change, and provided feed-back and advice on the levels of the caller’s cons of smok-ing, use of up to six change processes and temptations tosmoke relative to normative (compared to other smokersin the same stage of change) and ipsative (compared tothe previous assessment) data (Prochaska

et al

. 1993;Velicer

et al

. 1993). Qualitative evaluation of the tailoredadvice in a pilot study on the same population (callers toa Quitline) showed that it was generally well received andread.

The callback counselling program was a cognitive-

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction,

98

, 837–846

840

Ron Borland

et al.

behavioural, stage-based intervention which involved aseries of calls at negotiated preidentified points in the daysand weeks before and after a quit attempt. For example,the first callback often occurred just before or after aplanned quit day or, for those without immediate quittingplans, in a week’s time to allow the participant to receiveand read the service’s written materials. While calls wererarely made or requested beyond a few weeks after cessa-tion, a few callers did receive follow-up calls for up to 6months, largely because of problems in initiating a quitattempt and/or problems with relapse. Counselling callswere brief, generally of 10–15 minutes duration. The ser-vice was delivered by sessional staff proficient in tele-phone counselling and trained in delivering smokingcessation advice. Approximate data on costs of the call-back service are provided in Borland

et al

. (2001).All participants were mailed a smoking assessment

survey at 3 months, 6 months and 12 months after thebaseline assessment interview. Non-respondents to thefollow-up surveys were telephoned, resulting in approxi-mately half of the research surveys being administered bytelephone. The same smoking assessment survey was usedat each data collection point, except for the 12-month fol-low-up which used a shorter version of the survey.

Measures

The smoking assessment survey included smoking status(are you currently a cigarette smoker?), cigarette con-sumption, time to first cigarette of the day, self-efficacy anddetermination to quit, time since last quit attempt, lengthof current attempt, longest period quit in previous12 months, pros and cons of smoking, temptations tosmoke, use of change processes, household smoking anduse of a range of quit smoking services in the last 3–6months. Participants were categorized into three stages ofchange, as defined by DiClemente

et al

. (1991). Those inPrecontemplation were not seriously considering quittingin the next six months. Those in Contemplation were seri-ously considering quitting in the next 6 months, or wereplanning to quit in the next 30 days but had not made aquit attempt in the last year. Those in Preparation wereplanning to quit in the next 30 days, and had made a quitattempt in the last year. At baseline, additional measureswere also collected: gender, age, education level, agestarted smoking regularly, longest period ever quit for, anduse of pipes and cigars. At the 12-month follow-up, datawere collected using a briefer assessment tool which omit-ted the sections measuring the pros and cons of smoking,temptations to smoke and use of change processes, andadded a measure of current use of cigars or pipes. Themajor outcome measures were self-reported smoking

status at the 3-, 6- and 12-month follow-ups, both pointprevalence and 9 months’ sustained abstinence.

Statistical analysis

Data collected were analysed using SPSS

©

. Analyses arereported principally by intention to treat; that is, based onexperimental condition, regardless of the extent to whichparticipants received the intervention. We also report all-cases analysis where missing cases are treated as smok-ers. In addition, for those callers who received callbacks,subsidiary analyses are reported of the relationshipbetween the number of callback sessions received andoutcomes.

RESULTS

Recruitment and participant attrition

Overall, 1578 current smokers were recruited into thetrial, representing approximately 40% of eligible callers tothe Quitline service between October 1996 and August1997. Of those asked to participate, approximately 15%refused. Socio-demographic characteristics are shown inTable 1. Almost all the sample were in either Preparation(45%), that is, planning to quit in the next 30 days, andhad made a quit attempt in the last year, or Contempla-tion (54%). Given that they had all called the Quitline,those technically in Contemplation are possibly closer toPreparation than the typical Contemplator. Indeed, 65%of the Contemplators planned to quit in the next 30 daysand were in Contemplation because they did not have therequisite quit attempt in the past year.

Participants smoked an average of 22.8 cigarettes perday. There were no significant group differences in meannumber of cigarettes per day (

F

2,1575

=

1.58,

P

=

0.21) ormean minutes to the first cigarette of the day(

F

2,1575

=

1.35,

P

=

0.26).There was no significant difference in attrition across

the three treatment groups between baseline and the 12-month follow-up (

c

2

(2)

=

3.54;

P

=

0.17), but a trend tolose more in the intervention groups (see Table 1). Anal-yses of dropouts indicated that attrition decreased signif-icantly with age (

c

2

(2)

=

62.63;

P

<

0.01) with moreyoung people (aged 15–29 years) dropping out. No sig-nificant differences were found on gender (

c

2

(1)

=

2.01;

P

=

0.16) or educational attainment (

c

2

(2)

=

3.41,

P

=

0.18).Of the 528 smokers offered the callback service, 129

(24.4%) rejected it and another 41 (7.8%) did not receiveany callbacks, leaving only 358 (67.8%) of the samplereceiving callbacks. Overall, those allocated to the call-back service received an average of 3.3 callbacks, but of

Personalized cessation advice

841

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction,

98

, 837–846

those who actually received callbacks, the mean numberwas 4.8, with 30% receiving one or two, 47% from threeto six and 23% receiving seven or more calls.

Cessation attempts and outcomes

The number of smokers who made a quit attempt andamong those, the percentage who succeeded (self-reported not smoking at the next follow-up period), isshown in Table 2. Slightly more of the participants whoreceived the computer-generated tailored advice plus call-

back telephone counselling made an attempt betweenbaseline and 3-month follow-up compared to the othertwo groups, and significantly more of this group whotried succeeded, i.e. were quit at the 3-month follow-up(

c

2

(2)

=

13.95,

P

<

0.01).At the 3-month follow-up, significantly less people in

the computer-generated tailored advice only group(14.7%) reported having used nicotine replacement ther-apies (NRT) (gum or patches) than for the control(20.9%) or computer-generated tailored advice pluscallback telephone counselling (23.9%) groups

Table 1

Demographic and other features of participants at baseline (

n

=

1578).

Demographic characteristicsControl

group(

n

=

527) %

Computer-generatedtailored advice only(

n

=

523) %

Computer-generatedtailored advice and callbacktelephone counselling(

n

=

528)

%

P

-value

Gender% male 43.3 48.4 47.2 0.24

Age (years)15–29 31.4 37.4 31.8 0.2430–49 48.3 43.7 48.750

+

20.3 19.0 19.5

Educationup to and year 10 29.7 28.2 26.9 0.28years 11 and 12 41.5 43.2 48.0university or college 28.8 28.6 25.0

Level of addictioncigarettes per day 22.5 22.5 23.6 0.21mean minutes to first cigarette 11.5 13.5 12.0 0.26

Stage at baselineprecontemplation 0.9 1.9 0.6 0.29contemplation 52.9 54.5 54.5preparation 46.1 43.6 44.9

Quit attempt in past year (yes) 55.4 57.0 56.3 0.88

Retention3 months 93.0 92.4 90.3 0.266 months 88.2 86.4 83.0 0.0412 months 78.9 76.9 74.1 0.17

Table 2

Percentages of participants who made a quit attempt and of those, the percentage who succeeded in the 3 months followingrecruitment by intervention and condition (

n

=

1312).

Condition typeMade a quit attempt

a

(%, 95% CI)

Successful quit attemptof those who tried

b

(%, 95% CI)

Control group 57.4 (52.5–61.5) 23.2 (18.1–28.3)Computer-generated tailored advice only 56.0 (51.3–60.7) 24.7 (19.3–30.1)Computer-generated tailored advice plus callback telephone counselling 62.8 (58.2–67.4) 36.6 (30.8–42.4)

a

c

2

(2)

=

4.52;

P

=

0.10,

b

c

2

(2)

=

13.95;

P

=

0.001.

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs

Addiction,

98

, 837–846

842

Ron Borland

et al.

(

c

2

(2)

=

9.57,

P

<

0.01). This difference in use persisted atthe 6-month follow-up, but was no longer significant atthe 12-month follow-up.

The point prevalence of smoking status by condition ispresented in Table 3. At the 3-month follow-up, signifi-cantly more of the computer-generated tailored adviceplus callback telephone counselling group had quit, com-pared to the computer-generated advice only and thecontrol groups. The difference in point prevalencebetween the groups declined and was not significant atthe 12-month follow-up, due to a non-significant trendfor more new quitting in the groups who did not receivecallbacks.

Sustained abstinence rates showed that significantlymore participants in the computer-generated tailoredadvice plus callback telephone counselling group hadabstained for at least 3 months at the 6-month follow-up.The difference between the three groups declined andwas not significant for 9-month continuous abstinence atthe 12-month follow-up (see Table 3).

Alternative analyses of assigning smokers whosesmoking status could not be ascertained at the 12-monthfollow-up as smokers (

n

=

369) was undertaken. Bothpoint and period prevalence estimates did not differ sig-nificantly between groups (see Table 3).

The greater decline in the difference in cessation ratesfor the point prevalence measures over sustained absti-nence suggests that the callback telephone counsellingintervention may have inhibited participants’ confidenceto make subsequent quit attempts. To check this, weexamined self-efficacy levels at the 6-month follow-upamong participants who were smoking at that time. Nosignificant relationship was found between self-efficacy

and group (

c

2

(8)

=

6.81;

P = 0.56). Furthermore, exami-nation of subsequent quitting activity by group amongthose smoking at the 3-month follow-up showed no evi-dence of any trend for differences between experimentalconditions.

It is also possible that the results might reflectfaster relapse in the counselling group. Further analy-ses were conducted to assess the average days torelapse for participants who had made a quit attempt.No significant difference or trend was found betweenthe number of days quit among those who had triedto quit and relapsed by intervention group at eitherthe 3-month follow-up (F2,523 = 0.84, P = 0.43), 6-month follow-up (F2,337 = 0.64, P = 0.94) or 12-monthfollow-up (F2,385 = 1.51, P = 0.22).

To assess the relationship between receipt of the call-back counselling intervention and outcome, we com-pared point prevalence abstinence rates in the callbackcounselling group, among those who received callbacksversus those who refused or otherwise did not receive anycallbacks. We found evidence of a benefit of callbacks inthe first 3 months, translating into improved long-termabstinence, but effects on point prevalence were not clearbeyond 3 months (see Table 4). Those allocated to thecallback condition who did not receive callbacks had sim-ilar abstinence rates to the other two conditions.

Given the finding that NRT use was related to experi-mental condition, the potential effect of NRT on smokingstatus at 3 months was analysed using logistic regression.The regression included NRT use (yes, no) along withallocation to counselling (yes, no) and/or computer-gen-erated tailored advice (yes, no) to predict cessation at 3-month follow-up. This showed that NRT significantly

Table 3 Smoking cessation rates (point and period prevalence) at 3 months, 6 months and 12 months by condition.

nControl group %(95% CI)

Computer-generatedtailored advice only %(95% CI)

Computer-generatedtailored advice pluscallback telephonecounselling %(95% CI) P-value

Point prevalence3-month follow-up 1450 12.2 (9.3–15.1) 12.4 (9.5–15.3) 20.5 (16.9–24.1) c2(2) = 16.3; P < 0.016-month follow-up 1355 15.9 (12.6–19.2) 19.5 (15.8–23.2) 25.1 (21.0–29.2) c2(2) = 12.0; P < 0.0112-month follow-up 1209 22.1 (18.1–26.1) 22.6 (18.5–26.7) 25.6 (21.3–29.9) c2(2) = 1.5; P = 0.46

Period prevalence3-month abstinence at

6-month follow-up1355 7.3 (5.0–9.6) 9.3 (6.6–12.0) 13.5 (10.3–16.7) c2(2) = 9.7; P < 0.01

9-month abstinence at12-month follow-up

1209 5.0 (2.9–7.1) 6.0 (3.7–8.3) 8.2 (5.5–10.9) c2(2) = 3.4; P = 0.18

All casesPoint prevalence at 12 months 1578 17.5 17.4 18.9 c2(2) = 0.5; P = 0.76Period prevalence: 9 months at

12 months1578 4.0 4.6 6.1 c2(2) = 2.6; P = 0.27

Personalized cessation advice 843

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs Addiction, 98, 837–846

improved outcomes [P < 0.001; OR = 2.04; (1.22–2.73)],as did callback telephone counselling [P < 0.001; OR =1.84 (1.36–3.02)], but computer-generated tailoredadvice had no effect [P = 0.63; OR = 1.11 (0.72–1.72)].

DISCUSSION

The results of this study bring into question the value ofcomputer-generated tailored cessation materials but arelargely consistent with other research showing somebenefit of callback counselling.

This study relied exclusively on self-report, providedgenerally on an optically scannable form. Self-report isgenerally valid in conditions where there is no face-to-facecontact (Velicer et al. 1992) and we see no reason tobelieve it would not be valid here. It is true that those in thecounselling condition had also had advice calls, but we donot believe that this created any demand characteristics todistort reporting. Indeed, as both intervention conditionswere to receive feedback on their progress based on theirresponses (at 3 and 6 months, but not 12 months), webelieve the incentives are to respond honestly.

While the trials of Prochaska and colleagues reportedpositive effects of their expert advice system (Prochaskaet al. 1993; Velicer et al. 1999; Prochaska et al. 2001), wedid not find evidence of a beneficial effect. Given that thetrend was slightly positive, the results are consistent withthe meta-analysis which shows only a small net effect ofcomputer-generated advice. However, we believe thatflaws in the program make it likely that a better-designedprogram would have delivered better results. The majordifferences between our trial and others was that ourswas conducted in Australia (the others were in the USA),and we used smokers who were reactively telephoning aQuitline service, while other trials used community vol-unteers or proactively recruited smokers. We do notbelieve the location had any effect. We believe the differ-ence lies with the motivation to quit, or stage of changedistribution, of the sample. Our participants were moti-vated uniformly and were virtually all in the Preparationor Contemplation stages (as defined by the Transtheoret-ical model). All had called for help, many were on thebrink of a quit attempt, and those classified as in Contem-plation were arguably very close to the Preparation stage.

By contrast, the other samples (in the Prochaska et al.studies) contained many in the Precontemplation stage,as well as more typical Contemplators and Preparers.

Subsequent to discovering the disappointing resultsfor the computer-generated tailored advice we reviewedits content, focusing on what was offered at each stage.The advice provided to Precontemplators and Contem-plators to encourage them towards quitting is strong.However, beyond this problems emerge. Over a third ofour sample received information at baseline that wasmatched to their Transtheoretical model stage, not totheir quitting intentions. This is the group of Contempla-tors who plan to quit in the next 30 days, but are not con-sidered Preparers because they report no quit attempt inthe past year. This group were not given advice directed atquitting in the short term, rather the advice was directedat getting them to move towards quitting. This may havediscouraged some who would have quit from doing so atthe time. We tested for this by analysing the 3 month out-come data as a function of the baseline stage, separatingthe two groups of Contemplators. We found a significantdifference in 3 month point prevalence abstinencebetween the two groups of Contemplators in the printedself-help materials condition (P = 0.02), but not in thetwo intervention conditions (P = 0.41 and P = 0.38 fortailored advice only and tailored advice plus counselling,respectively), suggesting that the mismatched tailoredadvice did inhibit cessation among Contemplators whowere planning to quit within the next 30 days.

The material provided from Preparation onwards isless comprehensive and was potentially provided at thewrong time, given that feedback was only provided on a3-month cycle. Preparation advice is mainly aboutencouraging quit attempts, and only when in Action isdetailed relapse prevention advice provided. This is aproblem for our sample as many were already on thebrink of a quit attempt, and the vast majority of thosewho did make a quit attempt did so within 2 weeks of call-ing the Quitline. Those who made quit attempts at thistime received no support in the early stages of cessationwhen relapse rates peak (Zhu & Pierce 1995). For the call-ers who made a quit attempt and managed to stay quit,their first advice on how to cope with quitting wasreceived approximately 2.5 months later. At this point itwould be of little use, as the risk of relapse would have

Table 4 Abstinence measures for participants offered callback telephone counselling according to whether or not callbacks were received.

Point prevalence 3-monthfollow-up (%)a

Period prevalence (9 months)at12 months (%)b

Point prevalence12-month follow-up (%)c

No callbacks 15.4 4.0 23.8Callbacks received 22.9 10.2 26.4

ac2 = 3.47, P = 0.06; bc2 = 4.40, P = 0.04; cc2 = 0.31, NS.

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs Addiction, 98, 837–846

844 Ron Borland et al.

reduced substantially. The poor timing of advice aroundPreparation and Action is a serious flaw of the programwhen used with smokers actively seeking help who aretypically at this point. It is less of a problem whenapproaching smokers proactively, where relatively feware poised to quit, perhaps explaining the difference inresults from other studies (e.g. Velicer et al. 1999).

Postcessation, the Transtheoretical model is under-developed. It postulates two postcessation stages: Action,the first 6 months, where dealing with challenges is par-amount; and Maintenance, the period after that until astable non-smoking life-style has been established. Moreattention needs to be given to the complex changes thatoccur in the weeks and early months postcessation, i.e.within the Action stage (Shiffman 1989). In addition, theboundary between Action and Maintenance has beencriticized for being set arbitrarily at 6 months (Sutton1996). Indeed, there is no clear differentiation within thecomputer-generated advice, which also brings into doubtthe utility of this stage boundary, at least for purposes ofdirecting what forms of advice are appropriate.

It is also possible to argue that the period prevalenceestimates used in this study are biased against the tailoredadvice as the intervention persists until after the 6-monthfollow-up. However, as there is no evidence of any pointprevalence effect, or of the tailored computer advice onlycondition performing better than the other two over the3–6 or 6–12 month intervals, there is no evidence thatusing different outcomes would have made any differenceto the findings.

In summary, the computer-generated tailored adviceprogram used in this study was not effective, most prob-ably because it does not meet the needs of smokersactively seeking help. To be maximally effective amongthis group of smokers, tailored programs need to be deliv-ered in timely ways (i.e. more frequent advice for those onthe brink of a quit attempt), and more focus is needed onthe challenges facing a smoker once they make a commit-ment to try to quit.

The combination of the callback telephone counsel-ling and computer-generated tailored advice showed asignificant benefit initially, but it drifted into non-signifi-cance by the end of the trial. This seems to be due largelyto equivalent rates of relapse reducing the absolute differ-ence. The magnitude of effect found was somewhatsmaller than for a previous trial of telephone counsellingalone (Borland et al. 2001). While this could simply be achance effect and thus suggest that the magnitude ofeffect is a little smaller than the earlier study found, thedifferences could also be due to the lower uptake rate forthe callback service in this trial and/or to interferencebetween the two forms of advice. Advice was not inte-grated between the two forms of assistance, and given theproblems with the tailored advice, there were often con-

flicts in advice given. In considering the effect size for thecounselling and tailoring condition (probably bestthought of as counselling with some extra unhelpfuladvice), it is also important to consider the design. Over30% of those allocated to receive callbacks did not receiveany, reducing the potential effect size. This sub-group hadquit rates similar to the groups who were not offered calls(a similar outcome was found in Borland et al. 2001),suggesting that those who did not get calls do not differfrom those that did in their inherent capacity to quit suc-cessfully. If this is so, then the real effect size for counsel-ling is underestimated by about 30%.

The finding that counselling and use of NRT wereindependent predictors of cessation provides naturalisticevidence supporting Shiffman et al.’s (2000, 2001) find-ing that counselling and NRT make largely independentcontributions to quitting success.

The results suggest that combining two behaviouralinterventions (computer advice and telephone counsel-ling) was ineffective, as the quit rates were marginallylower than for a similar trial of the callback service alone(Borland et al. 2001), but again this conclusion is prema-ture on the basis of these two studies. Not only was thetailored advice not effective with this population, but alsointegration of the two forms of help was flawed in somecases (as noted above). It is not enough to have two formsof help that are based on the same conceptual model,without adequate integration. In our opinion, if tele-phone calls are interpolated with computer advice, thetelephone adviser needs to ensure that their advice is builton past computer advice and also to ensure that whathappens in the telephone session is coded into the com-puter in a way that subsequent computer-generatedadvice can take it into account.

The above analysis suggests that extreme cautionshould be taken in combining behavioural interventionssuch as these into categories for such things as meta-analysis. As this trial suggests, populations that differ inimmediate commitment to quitting may need differenttypes of help, so combining studies across different popu-lations may be problematic. Secondly, as noted already,effect sizes may be underestimated when use of the inter-vention is optional rather than contracted as part of thetrial. Finally, trials of behavioural interventions oftenoccur without the extensive preliminary testing that isnormal for pharmacotherapies, and therefore more bador inadequately developed interventions are likely to besubject to trial. In this study, the combination of com-puter advice and callback counselling was an experimen-tal intervention and, as noted above, we found problemswith the expert advice program that were potentiallyproblematic with the population used in this study. Thereis also a need for support for more formative testing ofinterventions, comparable to Phase 1 and Phase 2 clini-

Personalized cessation advice 845

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs Addiction, 98, 837–846

cal trials where issues of dosage (number and length ofsessions) and timing can be refined, ensuring that opti-mal interventions subject to formal trials are as good asthey can be.

Our findings are consistent with the conclusions of ameta-analytical review of studies that investigated theeffectiveness of smoking cessation telephone counselling(Stead & Lancaster 2002). The telephone counsellingcondition in this trial had the greatest initial impact oncessation rates compared to other conditions. We cannotrule out some of this effect being due to response bias, butbelieve this unlikely (Velicer et al. 1992). The results ofthis trial provide further evidence that a proactive call-back telephone counselling service can be a useful inter-vention that helps smokers to quit and stay quit, at leastfor the period over which the callbacks are being deliv-ered. We found significant short-term (at both 3 and6 months) effects of the service. While quit rates contin-ued to increase in all groups, the deterioration in size ofeffect by 12 months is of some concern. We might expectthat if the callbacks genuinely helped build skills inrelapse prevention that the absolute difference betweenthis group and the others would have been maintained.Development of more effective relapse prevention strate-gies is clearly a priority.

A major challenge for public health programmes is topromote behaviour change services that can be madeavailable to large numbers of people at low cost, andwhich promote lasting change among participants. Withincreasing uptake of relevant information technologysystems, it should be feasible for smoking cessation sup-port of the type we have described here to be offered morewidely. Our findings do suggest caution about what toexpect from computer-generated personalized advice.However, the field is in its early stages, and furtherresearch is required to establish what are the potentialsand limitations of these systems, and how advice is mostusefully personalized. Combining personalized resourceswith the efficient and appropriate use of telephone coun-selling has the potential to deliver high quality cessationservices to large numbers of smokers at a reasonable cost(Borland et al. 2001). None the less, the challengesinvolved in integrating the two in a way that will provideconsistent and useful advice for clients should not beunderestimated. To be maximally effective, tailored pro-grams, like any intervention, need to be appropriate tothe audience needs, be delivered in a timely way and beconsistent with any other advice.

ACKNOWLEDGEMENTS

The authors acknowledge the callers to the Quitline whomade this research possible, the contribution made by the

counsellors and staff of the Quitline service, as well as thedata entry staff. This research was funded by an NHMRCproject grant to Ron Borland, Neville Owen and others,supplemented by extra support from The Cancer CouncilVictoria. Callbacks were funded by Quit Victoria.

REFERENCES

Abrams, D. (1993) Smoking treatment issues: towards a steppedcare approach. Tobacco Control, 2(Supplement), S517–S537.

Abrams, D., Clark, M. & King, T. (1999) Increasing the impact ofnicotine dependence treatment: conceptual and practical con-siderations in a stepped care plus treatment-matchingapproach. In: Tucker, J. A., Donovan, D. M. & Marlatt, G. A.,eds. Changing Addictive Behaviour: Bridging Clinical and PublicHealth Strategies, pp. 307–330. New York: Guilford Press.

Borland, R., Segan, C., Livingston, P. M. & Owen, N. (2001) Theeffectiveness of callback counselling for smoking cessation: arandomized trial. Addiction, 96, 881–889.

Borland, R., Segan, C. J. & Velicer, W. F. (2000) Testing the tran-stheoretical model for smoking change: Victorian data. Aus-tralian Journal of Psychology, 52, 83–88.

Brown, S. L. & Owen, N. (1990) Clinical versus population per-spectives on smoking behaviour. Behavior Change, 7, 120–125.

Brug, J. (1999) Dutch research into the development and impactof computer-tailored nutrition education. European Journal ofClinical Nutrition, 53, S78–S82.

Clarke, V., Hill, D., Murphy, M. & Borland, R. (1993) Factorseffecting the efficacy of a community based quit smoking pro-gram. Health Education Research, 8, 537–546.

DiClemente, C. C., Prochaska, J. O., Fairhurst, S. K., Velicer, W.F., Velasquez, M. M. & Rossi, J. S. (1991) The process of smok-ing cessation: an analysis of the precontemplation, contem-plation and preparation stages of change. Journal of Consultingand Clinical Psychology, 59, 295–304.

Dijkstra, A., De Vries, H. & Roijackers, J. (1999) Targeting smok-ers with low readiness to change with tailored and nontai-lored self-help materials. Preventive Medicine, 28, 203–211.

Dijkstra, A., De Vries, H., Roijackers, J. & van Breukelen, G.(1998) Tailored interventions to communicate stage-matchedinformation to smokers in different motivational stages. Jour-nal of Consulting and Clinical Psychology, 66, 649–657.

Lancaster, T. & Stead, L. F. (2000) Self-help interventions forsmoking cessation. Cochrane Database Systematic Review, 2,CD001118.

Lichtenstein, E., Glasgow, R. E., Lando, H. A., Ossip–Klein, D. J. &Boles, S. M. (1996) Telephone counselling for smoking cessa-tion: rationales and meta-analytic review of evidence. HealthEducation Research, 11, 243–257.

Orleans, C. T., Schoenbach, V. J., Wagner, E. H., Quade, D.,Salmon, M. A., Pearson, D. C., Fiedler, J., Porter, C. Q. &Kaplan, B. H. (1991) Self-help quit smoking interventions:effects of self-help materials, social support instruction andtelephone counselling. Journal of Consulting and Clinical Psy-chology, 59, 439–448.

Ossip-Klein, D. I., Giovino, G. A., Megahed, N., Black, P. M.,Emont, S. L., Stiggins, T., Shulman, E. & Moore, L. (1991)Effects of a smokers’ hotline: results of a 10-county self-help.Journal of Consulting and Clinical Psychology, 59, 325–332.

Owen, L. (2000) Impact of a telephone helpline for smokers who

© 2003 Society for the Study of Addiction to Alcohol and Other Drugs Addiction, 98, 837–846

846 Ron Borland et al.

called during a mass media campaign. Tobacco Control, 9,148–154.

Owen, N. (1989) Behavioural epidemiology research and inter-vention studies to improve smoking-cessation services. HealthEducation Research, 4, 145–153.

Owen, N. & Davies, M. J. (1990) Smokers’ preferences for assis-tance with cessation. Preventive Medicine, 19, 424–443.

Owen, N., Kent, P., Wakefield, M. & Roberts, L. (1995) Low-ratesmokers. Preventive Medicine, 24, 80–84.

Platt, S., Tannahill, A., Watson, J. & Fraser, E. (1997) Effective-ness of antismoking telephone helpline: follow-up survey.British Medical Journal, 314, 1371–1375.

Prochaska, J. O., DiClemente, C. C. & Norcross, J. (1992) Insearch of how people change—applications to addictivebehaviours. American Psychologist, 47, 1102–1114.

Prochaska, J. O., DiClemente, C. C., Velicer, W. F. & Rossi, J. S.(1993) Standardized, individualized, interactive and person-alized self-help programs for smoking cessation. Health Psy-chology, 12, 399–406.

Prochaska, J. O., Velicer, W. F., Fava, J. L., Rossi, J. S. & Tsoh, J. Y.(2001) Evaluating a population-based recruitment approachand a stage-based expert system intervention for smoking ces-sation. Addictive Behaviors, 26, 583–602.

Reid, D. J., Killoran, A. J., McNeil, A. D. & Chambers, J. S. (1992)Choosing the most effective health promotion options forreducing a nation’s smoking prevalence. Tobacco Control, 1,185–197.

Shiffman, S. (1989) Conceptual issues in the study of relapse. In:Gossop, M., ed. Relapse and Addictive Behaviour, pp. 149–179.London: Routledge.

Shiffman, S., Paty, J. A., Rohay, J. M., Di Marino, M. E. & Gitchell,J. (2000) The efficacy of computer-tailored smoking cessationmaterials as a supplement to nicotine polacrilex gum therapy.Archives of Internal Medicine, 160, 1675–1681.

Shiffman, S., Paty, J. A., Rohay, J. M., Di Marino, M. E. & Gitchell,J. (2001) The efficacy of computer-tailored smoking cessationmaterials as a supplement to nicotine patch therapy. Drug andAlcohol Dependence, 64, 35–46.

Stead, L. F. & Lancaster, T. (2002) Telephone counselling forsmoking cessation (Cochrane review). In: The CochraneLibrary, Issue 3 2002. Oxford: Update Software.

Sutton, S. (1996) Can ‘stages of change’ provide guidance inthe treatment of addictions? a critical examination ofProchaska and Diclemente’s model. In: Edwards, G. & Dare,C., eds. Psychotherapy, Psychological Treatments and theAddictions, pp. 189–205. Cambridge, UK: Cambridge Univer-sity Press.

Velicer, W. F., Prochaska, J. O., Bellis, J. M., DiClemente, C. C.,Rossi, J. S., Fava, J. L. & Steiger, J. H. (1993) An expert systemintervention for smoking cessation. Addictive Behaviors, 18,269–290.

Velicer, W. F., Prochaska, J. O., Fava, J. L., LaForge, R. G. & Rossi,J. S. (1999) Interactive versus noninteractive interventionsand dose–response relationships for stage-matched smokingcessation programs in a managed care setting. Health Psychol-ogy, 18, 21–28.

Velicer, W. F., Prochaska, J. O., Rossi, J. S. & Snow, M. (1992)Assessing outcome in smoking cessation studies. PsychologicalBulletin, 111, 23–41.

Wakefield, M. & Miller, C. (1999) Evaluation of the NationalQuitline service. In: Hassard, K., ed. Australia’s NationalTobacco Campaign: Evaluation Report, Vol. 1. Canberra: Com-monwealth Department of Health and Aged Care.

Wilson, D., Wakefield, M., Owen, N. & Roberts, L. (1992) Char-acteristics of heavy smokers. Preventive Medicine, 21, 311–319.

Zhu, S.-H. & Pierce, J. P. (1995) A new scheduling method fortime-limited counselling. Professional Psychology: Research andPractice, 26, 624–625.

Zhu, S.-H., Stretch, V., Balabanis, M., Rosbrook, B., Sadler, G. &Pierce, J. (1996) Telephone counselling for smoking cessa-tion: effects of single-session and multiple session interven-tion. Journal of Consulting and Clinical Psychology, 64, 202–211.

Zhu, S.-H., Anderson, C. M., Tedeschi, G. J., Rosbrook, B.,Johnson, C. E., Byrd, M. & Gutierrez-Terrell, E. (2002) Evi-dence of real-world effectiveness of a telephone Quitline forsmokers. New England Journal of Medicine, 347, 1087–1093.

Zhu, S.-H., Tedeschi, G., Anderson, C. M., Rosbrook, B., Byrd, M.,Johnson, C. E. & Gutierrex-Terrell, E. (2000) Telephone coun-selling as an adjuvant treatment for nicotine replacementtherapy in a ‘real world’ setting. Preventive Medicine, 31, 357–363.