Evaluating Initial Reach and Robustness of a Practical Randomized Trialof Smoking Reduction

Russell E. Glasgow and Paul A. EstabrooksKaiser Permanente Colorado

Alfred C. Marcus and Tammy L. SmithAMC Cancer Research Center/University of Colorado Denver

Health Sciences Center

Bridget GaglioKaiser Permanente Colorado

Arnold H. Levinson and Suhong TongAMC Cancer Research Center/University of Colorado Denver

Health Sciences Center

Objective: This study evaluated the reach, initial effectiveness, and potential moderators and mediatorsof results of a smoking reduction program. Design: A generally representative sample of 320 adultsmokers from an HMO, scheduled for outpatient surgery or a diagnostic procedure, were randomized toenhanced usual care or a theory-based smoking reduction intervention that combined telephone coun-seling and tailored newsletters. Main Outcome Measures: Self-reported number of cigarettes smokedand carbon monoxide levels. Results: The intervention enrolled 30% of known eligible smokers andproduced reductions of 3 cigarettes per day greater than enhanced usual care. Intervention participantswere significantly more likely than control participants to achieve at least a 50% reduction in self-reported number of cigarettes using complete cases, imputation analyses, and intent-to-treat procedures.Similar patterns were seen for carbon monoxide results but were significant only in complete caseanalyses. The intervention was generally robust across patient characteristics (e.g., education, ethnicity,health literacy, and dependence) and phone counselors. Conclusion: Initial results suggest that thisprogram has potential to reach and assist smokers who may not participate in cessation programs.Additional research is indicated to enhance intervention effects, assess maintenance, and evaluate publichealth impact.

Keywords: smoking reduction, mediators, randomized controlled trial, RE-AIM

Despite the considerable progress that has been made in reduc-ing smoking prevalence in the United States, there has been aplateau in terms of national smoking rates (Centers for DiseaseControl & Prevention, 2005; National Center for Chronic DiseasePrevention & Health Promotion, 2006). Although the majority ofcurrent smokers say they want to quit, most decline to participatewhen offered smoking cessation assistance, and only about 5% ofthose who attempt to quit in a given year do so successfully(National Center for Chronic Disease Prevention & Health Pro-motion, 2006). As a result, there has been recent interest insmoking reduction (Carpenter, Hughes, Solomon, & Callas, 2004;Hughes, 2000; Rodu & Godshall, 2006), stimulated not only by the

plateau in cessation rates, but also by data indicating that smokingreduction likely increases long-term cessation rates (Carpenteret al., 2004; Hughes, 2000; Hughes & Carpenter, 2006; Hughes,Cummings, & Hyland, 1999). Although data are mixed and incon-clusive, smoking reduction may produce sustained reductions intobacco exposure biomarkers (Glasgow, Morray, & Lichtenstein,1989; Hughes, Callas, & Peters, 2007).

Despite the potential of smoking reduction interventions, toour knowledge no previous study has integrated a behavioralsmoking reduction intervention into health care organizationalor “real-world” smoking programs (Hughes & Carpenter, 2006).Pawson, Greenhalgh, Harvey, and Walshe (2005) have suggestedthat it is important to further health research by “discerning whatworks, for whom, in what circumstances, in what respects, andhow” (p. S1:32). To provide the information identified by Pawsonet al., practical clinical trials are needed that have more “transpar-ency” in reporting of results (Des Jarlais et al., 2004; Tunis, Stryer,& Clancey, 2003) and assess the potential mechanisms and mod-erators of effectiveness in typical intervention delivery settings(Glasgow, Davidson, Dobkin, Ockene, & Spring, 2006; Glas-gow, Magid, Beck, Ritzwoller, & Estabrooks, 2005). Key char-acteristics of practical trials are that they are conducted withheterogeneous patients, are conducted in multiple settings (in ourcase, across many different clinics and specialty areas), use mul-tiple outcomes important to decision makers, and use standard or

Russell E. Glasgow, Paul A. Estabrooks, and Bridget Gaglio, Institutefor Health Research, Kaiser Permanente Colorado, Denver, Colorado;Alfred C. Marcus, Tammy L. Smith, and Arnold H. Levinson, AMCCancer Research Center/University of Colorado Denver Health SciencesCenter.

Funding was provided by National Cancer Institute Grant RO1 CA90974-01. Appreciation is expressed for support and collaboration by EricK. France, Chief of Preventive Medicine, and the Preventive MedicineDepartment of Kaiser Permanente Colorado.

Correspondence concerning this article should be addressed to RussellE. Glasgow, Kaiser Permanente Colorado, P.O. Box 378066, Denver, CO80237-8066. E-mail: russg@ris.net

Health Psychology Copyright 2008 by the American Psychological Association2008, Vol. 27, No. 6, 780–788 0278-6133/08/$12.00 DOI: 10.1037/0278-6133.27.6.780

780

alternative treatments as comparison conditions (Glasgow, Magid,et al., 2005; Tunis et al., 2003). Such trials can also increaseunderstanding of the external validity of health promotion interven-tions by documenting the reach of interventions—that is, the percent-age of eligible persons who will participate in a given program andhow representative they are of the target population (Glasgow,France, et al., 2006; Glasgow, Lichtenstein, & Marcus, 2003).

Our research group has recently developed a smoking reductionintervention that was designed to be broadly applicable and isintegrated into other smoking modification options in a largemanaged care organization (Glasgow, France, et al., 2006). Thefocus of the program is on behavioral approaches to reduce thenumber of cigarettes smoked, not on the use of alternative tobaccoproducts (Shiffman et al., 2002). A social–ecological theoreticalapproach described in detail elsewhere (Glasgow, France, et al.,2006), including risk perceptions, self-efficacy, problem solving,and environmental support, was used for intervention develop-ment. Nicotine replacement therapy was not used as part of theintervention because it had not been approved for use for smokingreduction (vs. cessation) in the United States.

Of late, social–ecological theories of human behavior havebecome popular bases for intervention development (Glasgow,Strycker, Toobert, & Eakin, 2000; Sorensen, Barbeau, Hunt, &Emmons, 2004; Spence & Lee, 2003). The primary assumption ofa social–ecological model is that the interaction between theenvironment and an individual is the primary determinant of

behavior (Bandura, 1997; McLeroy, Bibeau, Streckler, & Glanz,1988). Figure 1 depicts the simplified social–ecological interven-tion logic model used in this study. Briefly, personal psychosocial(e.g., self-efficacy beliefs), physical (e.g., gender), and behavioral(e.g., number of cigarettes smoked) factors are all important de-terminants of future behavior. Personal determinants are embeddedwithin an environmental context that will either facilitate or im-pede behavior change. Each environment—whether home, work,or some other place (e.g., church or nightclub)—can generally beseparated into physical and social structures. Physical structurecould include the presence of a designated smoking area or avail-ability of ashtrays. Social structure includes group norms forsmoking, work policies for smoking breaks, and supportive familymembers. Our study used this social–ecological structure to directintervention strategies toward personal and environmental factorswhile identifying potential moderators and mediators that mayaffect intervention success (see Figure 1).

The unique context of the study is that smokers scheduled foroutpatient surgery or an invasive medical procedure (e.g., colonos-copy) were approached before their operation or procedure. Therehas been a large amount of research on smoking cessation amonghospitalized and primary care patients. However, there has been nosuch research on outpatient surgery settings or on smoking reduc-tion provided at such a “teachable moment” (Cole-Kelly, 2006;McBride, Emmons, & Lipkus, 2003), and there are an increasingnumber of such outpatient and day surgeries scheduled.

Figure 1. Logic model of proposed intervention effects, mediators, and moderators. R � randomized; CO �carbon monoxide.

781SMOKING REDUCTION PRACTICAL RCT

The purposes of the present study were fourfold: (a) to evaluatethe reach of a smoking reduction program offered in conjunctionwith other smoking services of a large HMO; we hypothesized thata large percentage of smokers who are unwilling to attempt ces-sation would attempt reduction; (b) to determine the initial (3-month) effectiveness of the program relative to an enhanced usual-care condition in a practical, randomized controlled trial (Glasgow,Magid, et al., 2005); we hypothesized that those randomized to thesmoking reduction intervention would show greater reduction incigarettes per day and carbon monoxide levels than those in usualcare and that a larger percentage of intervention participants wouldachieve 50% reduction or more; (c) to assess the robustness of theprogram (moderator effects) across patient characteristics and in-tervention staff; we hypothesized that intervention effects wouldbe robust across potential moderator variables; and (d) finally, asecondary aim, to determine the mechanisms of intervention ef-fectiveness by evaluating potential theoretical mediators of treat-ment outcomes; we hypothesized that self-efficacy, number ofproblem-solving techniques used, risk perception, and level ofsocial–environmental support would mediate treatment outcomes.

Method

Recruitment

This study was conducted in the Kaiser Permanente Colorado(KPCO) health care system and received institutional review boardapproval. Participants were recruited from KPCO smokers about toundergo outpatient surgery, a gastrointestinal procedure (such as acolonoscopy or sigmoidoscopy), or a screening/diagnostic proce-dure (mammogram or pulmonary lung function test). KPCO’selectronic medical records system was used to identify all currentsmokers, ages 18 and older, scheduled for one of the proceduresabove. These 2,234 individuals were notified about the program bya personalized introductory letter from KPCO’s chief of preventivemedicine. A descriptive flyer was included to provide additionaldetails about the study. An informed consent form and HealthInsurance Portability and Accountability Act (HIPAA) statementwere included with the introductory letter, flyer, and an “opt-out”postcard. Smokers who did not wish to be contacted could declineby returning the postcard. One to two weeks after receiving theletter, participants who did not decline via postcard were contactedby trained telephone interviewers from the AMC Cancer ResearchCenter on behalf of the KPCO Preventive Medicine Department.After smoking status, interest, and eligibility criteria were con-firmed, potential participants received a detailed description of thestudy.

Patients were excluded if they (a) smoked fewer than 10 cigarettesper day, (b) could not read or understand English, (c) canceled orpostponed the medical procedure, or (d) were unavailable for thestudy duration. The medical procedure for each participant wasused as a motivational opportunity to make changes in theirsmoking habits. Participants were told that the study involvedreceiving phone calls and personalized mailed materials to helpthem to reduce the number of cigarettes smoked. The phoneinterviewer then asked whether the participant was interested inquitting or cutting down or not interested in working on his or hersmoking at that time. Patients who smoked fewer than 10 ciga-rettes per day or who had difficulty deciding between quitting or

cutting down were referred to their choice of several health planand state quit line cessation options. An attempt was made tocollect demographic and smoking history information on all thosecontacted to assess the representativeness of participants.

Patients who agreed to participate in the research study com-pleted an informed consent procedure and HIPAA authorization.Baseline survey assessments were completed before randomiza-tion using a computer-assisted telephone interview, after whichparticipants were randomly assigned to intervention conditionsusing a computer algorithm developed by the project statistician.Participants then came in for a brief in-person visit with KPCOresearch staff (different than intervention staff) to provide baselinesamples for the biochemical outcomes. Those randomized to in-tervention began the first phone counseling session immediately.Recruitment took place between November 2004 and April 2006.Follow-up assessments were completed between February 2005and September 2006.

Intervention

Intervention delivery. The intervention was delivered across a6-month time period and consisted of four telephone counselingsessions, four tailored newsletters, and one targeted newsletter.Three phone calls and three tailored letters were scheduled to bedelivered by the 3-month follow-up, as show in Figure 2. Tailoredintervention components produced individual messages based on acombination of smoking rate, type of procedure, gender, level ofsuccess, and self-efficacy. Targeted newsletters addressed issuescommon to smokers having had recent medical procedures andattempting to reduce their smoking. The intervention componentswere sequenced and gradually faded over time as depicted inFigure 2. The 3- and 12-month assessments were conducted viaself-administered questionnaire. (Twelve-month data will be re-ported separately when available.) This article concerns the3-month results, the program reach, and the mediator–moderatorresults.

Each counseling call used strategies to heighten participantself-efficacy to achieve and sustain reduced smoking levels. Anassessment, goal setting, and barriers identification and resolutionprotocol was used to help participants improve confidence in theirability to reduce the number of cigarettes they smoked. Partici-pants were encouraged to set an initial goal of a one third reductionin number of cigarettes smoked. On the basis of progress andself-efficacy, participants’ later reduction goals were individuallytailored, but we attempted to have most attempt at least a 50%reduction as recommended by Windsor (1999). This level wasselected a priori as a clinically significant measure of reduction(Jacobson & Truax, 1998). Those successful in achieving a 50% ormore reduction were then encouraged to consider cessation. News-letters included specific tailoring based on data collected duringthe preceding telephone counseling call and included content tar-geted to personal and environmental factors.

The reduction program consisted entirely of tailored behavioraland environmental change strategies. The telephone calls weretailored concurrently as participant data were collected, to respondappropriately to the individual’s progress toward his or her reduc-tion goals. Each newsletter that followed a counseling call wasthen tailored on the basis of the new information obtained fromthat call.

782 GLASGOW ET AL.

Figure 2. Timeline: Delivery of intervention components.

783SMOKING REDUCTION PRACTICAL RCT

Phone protocol and interviewers. The four interviewers whoconducted the counseling calls included both genders; three werecollege students in their early 20s, and one was a high schoolgraduate age 55. None had previous training or experience inpsychosocial or smoking counseling. Interviewer training beganwith an overview of the project, followed by a specific trainingsession focusing on the recruitment/baseline call. Additional train-ing sessions were conducted for each counseling call. Thecall-specific training sessions lasted 2 half-days and includedquestion-by-question analysis, independent review time followedby a group question-and-answer period, and supervised role-playing exercises. Supplementary training sessions were con-ducted in addition to the call-specific trainings to strengthen in-terviewer skills for building rapport and helping participants.Throughout the study, supervisory staff frequently monitored livecalls and provided real-time feedback for quality control andimprovement. When not conducting calls, interviewers were en-couraged to independently practice scenarios they had not recentlyencountered, review computer-assisted telephone interview callcontent, and practice call scenarios using their own language.

Newsletters. As shown in Figure 2, newsletters were graduallyfaded over time and were sent after each phone call. Each news-letter included a regular column by KPCO’s chief of preventivemedicine and a narrative story that addressed vicarious learningthrough the use of a like model (i.e., tailored on gender and levelof participant success) experiencing the intervention. The theoret-ical mediators were addressed throughout the content of the fivenewsletters, the first four of which were individually tailored andthe last of which was targeted (Kreuter, Farrell, Olevitch, &Brennan, 2000). Content focused on increasing self-efficacy, iden-tifying benefits and barriers to reduction, perceived risk or threat ofsmoking, and relapse prevention. In addition, the newsletters in-cluded generic content on other aspects of the social–ecologicalmodel such as tips for handling stress, preventing weight gain, andrearranging one’s social environment.

Enhanced usual care condition. Individuals randomized to the“usual care” condition participated in the recruitment/baseline call.Then three quarterly generic health education mailings about avariety of health promotion topics and generic behavior changetips were sent out on the same newsletter mailing schedule asthe first, third, and fifth newsletters for the intervention arm ofthe study. Usual care participants also participated in the sameassessments as the intervention participants, including the car-bon monoxide sample collections. On completion of the study,all individuals in the usual care arm received “end-of-studynewsletters,” which were three targeted reduction newsletterscontaining the same articles provided in the interventionnewsletters.

Measures

Outcomes. Both self-report and biochemical measures ofsmoking were collected at baseline and 3-month assessments byassessment staff separate from intervention staff phone callers. Theself-report measure asked participants separately about the numberof cigarettes smoked on (a) workdays and (b) nonworkdays theprior week. A composite self-report measure of cigarettes per daywas calculated on the basis of a weighted average [(5 � week dayrate) � (2 � weekend rate)/7].

Carbon monoxide (CO) in expired breath samples was assessedusing a Vitalograph CO monitor (Vitalograph Inc., Lenexa, KS).

Demographic and moderator variables. Potential backgroundand moderator variables included age, gender, number of chronicillnesses, ethnicity, education, income, years smoked, cigarettesper day, presence of other smokers in the home, whether thepatient’s physician had recommended quitting, whether cigaretteswere purchased by the carton (vs. smaller quantities), and somepsychosocial variables. Because of the priority on feasibility andreducing respondent burden, brief versions of all psychosocialmeasures were used. Psychosocial variables included depressivesymptoms as assessed by the two questions endorsed by the KaiserPermanente Care Management Institute National DepressionGuideline (sensitivity � 96%, specificity � 57%). Health literacywas assessed by the three-item health literacy scale of Chew,Bradley, and Boyko (2004), and nicotine dependency was assessedby five items from the Wisconsin Assay of Relapse and Depen-dence (Piper et al., 2004). These items assessed smoking afterwaking up, most friends smoke, most family smoke, urges uponawakening, and cigarettes per day. Finally, support from familyand friends was assessed via a single item asking participants torate the amount of support received for smoking reduction on a5-point scale.

Hypothesized mediating variables. As illustrated in Figure 1,we hypothesized that the smoking reduction intervention wouldachieve its outcomes through changes in four proposed mediatingvariables. Self-efficacy for smoking reduction was assessed viatwo measures: a 100-point rating scale of confidence that theparticipants could reduce their number of cigarettes smoked bytwo thirds (a la Bandura, 1997) and the average of 20 items fromthe situational efficacy scale of Condiotte and Lichtenstein (1981).Risk perception was assessed by a single item that asked par-ticipants to rate how much risk they personally faced by smok-ing 20 cigarettes per day. Use of reduction strategies was acomposite measure of the number of seven possible strategiesthat participants reported having used at the 3-month assessment.Finally, social–environmental support was assessed by means of a6-point rating scale that asked participants, “During the past 3months, when you were trying to reduce the number of cigarettesyou smoke, which of the following best describes the amount ofsupport or resistance you received from family, friends, andcoworkers?”

Analyses. We conducted descriptive analyses to evaluate dis-tributions and to evaluate reach and attrition. Repeated measuresanalyses were used to evaluate intervention effects and improve-ment from baseline to the 3-month follow-up on self-reported andbiochemical measures of smoking and on hypothesized mediatingvariables. In addition to complete-case and intent-to-treat analyses,we obtained a third set of estimates through multiple imputation(five multiples) of missing values using multivariate sequentialregression (IVEWare, University of Michigan; Raghunathan, Lep-kowski, Van Hoewyk, & Solenberger, 2001). The method assumesitem-missing values are missing at random (Rubin, 1976). Imputedvalues for each individual are conditioned on observed values forthat individual.

We conducted multiple regression analyses to identify moder-ator variables (using interaction terms) that altered the impact ofintervention on number of cigarettes and CO levels. Following therecommendations of Windsor et al. (1999), we adopted an a priori

784 GLASGOW ET AL.

criterion for clinical significance of achieving a reduction of atleast 50% of baseline level and conducted analyses of the percent-age of participants in each condition achieving this magnitude orgreater reduction on (a) self-reported cigarettes and (b) CO. Atwo-tailed alpha level of .05 was adopted for significance.

Results

Reach

As can be seen in Figure 3, a total of 391 participants (36.8% ofthe smokers confirmed as eligible) enrolled in the study. A moreconservative estimate, which assumed that the same percentage ofmembers who could not be contacted or who opted out beforeeligibility was determined were eligible as among those who com-pleted eligibility determination (64%; see Figure 3 and Table 1),produced an enrollment rate of 30.1%. As can be seen, this wasmore than three times the percentage of smokers who opted for acessation program (9.5% of known eligible smokers), despite avariety of cessation options being available and undecided smok-ers being encouraged to attempt cessation.

There was, however, considerable early attrition, and 71 of thosesmokers (36 intervention and 35 control smokers) who enrolled over

the telephone never completed the baseline biochemical sample col-lection or participated any further. Counting only those who com-pleted all baseline assessment activities as participants resulted in aparticipation rate of 30.1% of known eligible smokers (320 of 1,063)or 22.3% of all assumed eligible smokers (320 of 1,438).

Table 1 summarizes the characteristics of four groups of smok-ers: smoking reduction participants (n � 320), those who initiallyagreed but did not complete baseline (n � 71), those who declinedto participate (n � 167–572; many did not provide data on somemeasures in Table 1), and those who chose to quit (n � 101). Ascan be seen, there were few differences among members of theseself-selected groups. The majority of participants were more than50 years old, were female, smoked approximately a pack a day,and had smoked for more than 30 years. The only significantdifference was that Latinos were more likely to choose cessationover reduction or other options ( p � .05).

Implementation

The intervention was consistently implemented. Records keptby project staff indicated that by the 3-month follow-up, 94% ofintervention participants received two or more phone calls and100% received all the newsletters.

Figure 3. Participant flow diagram.

785SMOKING REDUCTION PRACTICAL RCT

Outcome Effects

There were no baseline differences between participants ran-domized to intervention versus control conditions, as can be seenin Table 2. As shown in Table 3, there were mixed results on thebehavioral and biochemical outcomes. A chi-square analysis re-vealed that significantly more intervention than control partici-pants achieved a 50% reduction in number of cigarettes per day( p � .008). Intent-to-treat analyses, assuming that those lost tofollow-up did not achieve 50% reduction, revealed a similar pat-tern (15.9% vs. 7.7%, p � .05) and were significant. Finally,imputation analyses using missing-at-random assumptions werealso significant (22.3% vs. 9.9%, p � .05).

Continuous repeated measures analyses of number of cigarettessmoked per day revealed a similar pattern, with intervention par-ticipants reporting reducing nonsignificantly more than controlparticipants, who also reduced their smoking, but this differencewas significant only in the complete cases analysis ( p � .05).

Analyses of biochemical data revealed a similar pattern. Anal-yses of the percentage of participants who achieved at least a 50%

reduction in CO levels were significant for the complete cases(chi-square, p � .05), but not for intent-to-treat or imputationanalyses (although relative risks did favor intervention partici-pants; relative risks � 1.9 and 2.4, respectively, ns). For thecontinuous CO measure, only the complete case analysis wassignificant. As can be seen in Table 3, in general the interventioncondition produced larger reductions than the control condition,and these differences reached conventional significance levels inabout half the analyses. Overall reductions were small to moderate.

Finally, there were no differences between conditions onquitting-related variables. Twenty-three percent of participants ineach condition reported having attempted to quit by the 3-monthfollow-up, but few of those were successful. As would be ex-pected, significantly more intervention participants (96%) thancontrol participants (81%) reported attempting smoking reduction( p � .0005).

Table 1Characteristics of Smokers Choosing Different Options

VariableSR participants: baseline

(n � 320)SR participants: no baseline

(n � 71)Declined

(n � 167�572)aWant to quit

(n � 101)

Female (%) 72.3 74.7 67.4 60.8Latino* (%) 5.0 5.6 9.9 14.9Age (M[SD]) 55.4 (10.8) 54.2 (10.5) 55.1 (12.7) 54.0 (10.9)Cigarettes/day (M[SD]) 20.6 (9.2) 20.2 (7.2) 21.6 (19.1) 17.7 (12.3)Years smoked 37.4 (11.4) 36.4 (9.9) 38.7 (12.7) 33.0 (12.5)

Note. SR � Smoking Reduction.a N varies because many did not provide data on some measures.*p � .05.

Table 2Baseline Characteristics of Intervention and ControlParticipants

Characteristic

Intervention(n � 164 )

Control(n � 156 )

M or % SD M or % SD

DemographicsMean age (years) 54.8 10.4 56.0 11.3Female (%) 73.2 71.8Latino (%) 3.7 6.5Mean years education 14.2 2.6 14.4 2.5

Smoking characteristicsMean cigarettes per day 21.2 9.4 20.1 9.0Mean years smoked 36.9 10.9 37.9 11.4Purchase by pack (%) 26.2 25.2Dr. recommend quitting (%) 83.5 81.5

Medical characteristicsMean no. medical conditions 5.6 0.8 5.6 0.9Have smoking-related condition (%) 26.8 31.0Positive on depression (%) 43.9 47.1

OtherMean health literacy score 14.1 1.6 13.7 1.9

Note. Between-condition comparisons were all nonsignificant.

Table 3Results by Condition on Reductions in Number Smoked andBiochemical Outcomes

Variable/time

Intervention Control

M or % SD M or % SD

Mean no. cigarettes/dayBaseline 21.1 9.4 20.6 9.63 months (complete)* 14.8 8.2 17.6 9.33 months (intent to treat) 17.2 9.6 17.3 8.73 months (imputed) 16.4 9.7 17.3 9.1

% greater than or equal to 50%reduction at 3 months

Complete* 19.8 RR � 2.15 9.2Intent to Treat* 15.9 RR � 2.06 7.7Imputed* 22.3 RR � 2.25 9.9

Mean carbon monoxideBaseline 29.8 14.2 31.1 15.03 months (complete)* 22.5 12.3 26.9 13.53 months intent to treat 25.5 13.5 26.3 13.23 months (imputed) 21.0 11.5 24.9 13.0

% greater than or equal to 50%carbon monoxide reduction

Complete cases* 20.0 RR � 3.45 5.83 months intent to treat 11.0 RR � 1.90 5.83 months imputed 22.7 RR � 2.40 10.7

Note. RR � relative risk.*p � .05.

786 GLASGOW ET AL.

Moderator Analyses

Interaction analyses (Moderator Variable � Treatment Condi-tion) were significant for 4 of the 14 potential moderator variablesdescribed in the Measures section for either number of cigarettessmoked (two significant moderators) or for CO level (four signif-icant moderators). However, the pattern of moderators was com-plex. Given space limitations, we discuss only the two moderatorsthat were significant in both analyses. Those having a chronicillness showed greater treatment effects on both number of ciga-rettes smoked ( p � .004) and CO level ( p � .009) than thosewithout a chronic illness. Second, those who reported purchasingcigarettes by the carton at baseline (vs. by the pack) showedgreater treatment effects on both cigarettes smoked ( p � .019) andCO levels ( p � .022). Intervention effects appeared robust acrossa variety of other potential moderators including income, ethnicity,health literacy, education, gender, age, the depression screener,and type of medical procedure.

Mediator Analyses

Of the four potential mediator variables analyzed, only numberof strategies used was significantly greater among interventionthan control participants ( p � .007). However, strategies were notsignificantly related to outcomes. Self-efficacy (100-point rating)significantly predicted improvements in all dependent variablesanalyzed, but the reduction intervention was not successful inenhancing efficacy more than the control condition, thus notfulfilling requirements for mediation (Baranowski, Anderson, &Carmack, 1998; Baranowski, Lin, Wetter, Resnicow, & Davis,1997). The other variables did not meet either of the abovemediation conditions.

Discussion

Relatively little is known about nonlaboratory applications ofbehavioral smoking reduction or about how and under what con-ditions they work (Hughes & Carpenter, 2006; Pawson et al., 2005;Shiffman et al., 2002). In general, our results found that thebehavioral smoking reduction intervention was successful inrecruiting smokers, even when offered in the context of acessation-oriented menu. A sizable proportion of smokers, 20%–30%, depending on the index, who were not willing to quit werewilling to attempt reduction in this research project. Because ofHIPAA and confidentiality concerns, relatively few data wereavailable on the representativeness of reduction participants. Fromavailable data, it appears that smokers selecting reduction wererelatively representative of health plan smokers, with the exceptionthat a smaller percentage of Latinos opted for reduction comparedwith cessation. Those selecting reduction appeared to be at least asheavy smokers as those choosing cessation.

The effectiveness of the intervention was more mixed. It appearsthat the program produced beneficial effects on both self-reportand biochemical measures of smoking exposure, but that theseeffects were modest and reduced in intent-to-treat analyses be-cause of attrition and large variability. Both measures of clinicallysignificant change (Jacobson & Truax, 1998; Windsor et al., 1999)of the percentage of participants achieving 50% or greater reduc-tion revealed significantly greater improvement for the interven-

tion than for the control condition. Somewhat surprisingly, anal-yses of the continuous outcome measures, which are usuallymore powerful, generally failed to reveal between-conditioneffects. One of the reasons for the modest intervention effectsmay have been contextual and historical factors that producedlarger change among control participants than might have oth-erwise been the case. Specifically, during this study the state ofColorado increased its tobacco tax significantly, implemented astatewide no-indoor-smoking ordinance, and the Colorado Quit-line started offering free nicotine replacement therapy, all ofwhich have been found to decrease consumption (Bartecchiet al., 2006; Hughes et al., 1999; Levy, Hyland, Higbee, Remer,& Compton, 2006).

The modest intervention effects appear to be moderately robustacross a variety of participant demographic, baseline smokingpattern, and medical characteristics. The moderator analyses thatwere significant suggested that those with chronic illness and thosewho purchased their cigarettes by the carton benefited more fromthe intervention. Given the importance of translating research intopractice (Glasgow, Davidson, et al., 2006; Glasgow et al., 2003;Glasgow, Magid, et al., 2005; Tunis et al., 2003), investigations ofgeneralizability or “robustness” of intervention effects are alsoimportant. Answers to questions such as for what groups does thisintervention, delivered by what types of staff, produce effects onwhat outcomes and under what conditions are essential (Glasgow,France, et al., 2006; Pawson et al., 2005).

The intervention was developed to reflect several key theoreticalprocesses. However, none of the four hypothesized mediatingvariables functioned as treatment mediators. This was generallybecause with the exception of number of reduction strategies used,the intervention was not more successful in producing improve-ments in these variables than the control condition. This did notappear to be due to insufficient intervention intensity or poorimplementation. With the exception of the previously validatedself-efficacy measures, the mediator measures may not have beensufficiently reliable or sensitive to change (Glasgow, Ory, et al.,2005). Because of the number of measures and respondent burdenconcerns, we chose to use brief measures of mediators, and thesemay not have been detailed or reliable enough to detect modestdifferences between conditions.

This study had both limitations and several strengths. Limita-tions include being conducted in only one health care setting andthe exclusion of non-English-speaking participants, which mayhave been partially responsible for fewer Latinos selecting reduc-tion, and the brevity of a number of the measures. Strengthsinclude the randomized design; a relatively large sample size forthe smoking reduction literature; the comprehensive measurescollected including reach, implementation, and biochemical expo-sure measures; and the theoretical basis of the study along withrelated moderation and mediation analyses.

In summary, it appears feasible to deliver smoking reductionservices on a relatively large-scale basis in a way that can reach animportant segment of smokers who would not otherwise receiveservices. Important directions for future research should includeprocedures to enhance retention in the intervention, ways toenhance the magnitude of intervention effects—such as addingnicotine replacement therapy; evaluation of longer term effects,including whether the reduction intervention leads to greatercessation (or less) than usual care over time (Hughes & Car-

787SMOKING REDUCTION PRACTICAL RCT

penter, 2006); and evaluation of intervention costs and costeffectiveness.

References

Bandura, A. (1997). Self-efficacy: The exercise of control. New York:Freeman.

Baranowski, T., Anderson, C., & Carmack, C. (1998). Mediating variableframework in physical activity intervention: How are we doing? Howmight we do better? American Journal of Preventive Medicine, 15,266–297.

Baranowski, T., Lin, L. S., Wetter, D. W., Resnicow, K., & Davis, H. M.(1997). Theory as mediating variables: Why aren’t community interven-tions working as desired? Annals of Epidemiology, 7, S89–S95.

Bartecchi, C., Alsever, R. N., Nevin-Woods, C., Thomas, W. M., Estacio,R. O., Bartelson, B. B., et al. (2006). Reduction in the incidence of acutemyocardial infarction associated with a citywide smoking ordinance.Circulation, 114, 1490–1496.

Carpenter, M. J., Hughes, J. R., Solomon, L. J., & Callas, P. W. (2004).Smoking reduction and motivational advice increase future cessationamong smokers not currently planning to quit. Journal of Counselingand Clinical Psychology, 72, 371–381.

Centers for Disease Control & Prevention. (2005, November 11). Cigarettesmoking among adults–United States 2004. Morbidity and MortalityWeekly Report, 54, 1121–1124.

Chew, L. D., Bradley, K. A., & Boyko, E. J. (2004). Brief questions toidentify patients with inadequate health literacy. Family Medicine, 36,588–594.

Cole-Kelly, K. (2006). The teachable moment. Academic Medicine, 81, 958.Condiotte, M. M., & Lichtenstein, E. (1981). Self-efficacy and relapse in

smoking cessation programs. Journal of Consulting and Clinical Psy-chology, 49, 648–658.

Des Jarlais, D. C., Lyles, C., Crepaz, N., & TREND Group. (2004).Improving the reporting quality of nonrandomized evaluations of behav-ioral and public health interventions: The TREND statement. AmericanJournal of Public Health, 94, 361–366.

Glasgow, R. E., Davidson, K. W., Dobkin, P. L., Ockene, J., & Spring, B.(2006). Practical behavioral trials to advance evidence-based behavioralmedicine. Annals of Behavioral Medicine, 31, 5–13.

Glasgow, R. E., France, E. K., Marcus, A., Riley, K. M., Levinson, A., &Bischoff, K. (2006). Do behavioral harm reduction approaches reachmore or different smokers? Different designs, similar answers. AddictiveBehaviors, 31, 518.

Glasgow, R. E., Lichtenstein, E., & Marcus, A. C. (2003). Why don’t wesee more translation of health promotion research to practice? Rethink-ing the efficacy to effectiveness transition. American Journal of PublicHealth, 93, 1261–1267.

Glasgow, R. E., Magid, D. J., Beck, A., Ritzwoller, D., & Estabrooks, P. A.(2005). Practical clinical trials for translating research to practice: De-sign and measurement recommendations. Medical Care, 43, 551–557.

Glasgow, R. E., Morray, K., & Lichtenstein, E. (1989). Controlled smokingvs. abstinence as a treatment goal: The hopes and fears may be un-founded. Behavior Therapist, 20, 77–91.

Glasgow, R. E., Ory, M. G., Klesges, L. M., Cifuentes, M., Fernald, D. H.,& Green, L. A. (2005). Practical and relevant measures of healthbehavior for primary care settings. Annals of Family Medicine, 3, 73–81.

Glasgow, R. E., Strycker, L. A., Toobert, D. J., & Eakin, E. G. (2000). Asocial-ecologic approach to assessing support for disease self-management: The chronic illness resources survey. Journal of Behav-ioral Medicine, 23, 559–583.

Hughes, J. R. (2000). Reduced smoking: An introduction and review of theevidence. Addiction, 95, S3–S7.

Hughes, J. R., Callas, P. W., & Peters, E. N. (2007). Interest in gradualcessation. Nicotine & Tobacco Research, 9, 671–675.

Hughes, J. R., & Carpenter, M. J. (2006). Does smoking reduction increasefuture cessation and decrease disease risk? A qualitative review. Nico-tine & Tobacco Research, 8, 739–750.

Hughes, J. R., Cummings, K. M., & Hyland, A. (1999). Ability of smokersto reduce their smoking and its association with future smoking cessa-tion. Addiction, 94, 109–114.

Jacobson, N. S., & Truax, P. (1998). Clinical significance: A statisticalapproach to defining meaningful change. In A. E. Kazdin (Ed.), Meth-odological issues and strategies in clinical research (2nd ed., pp. 631–648). Washington, DC: American Psychological Association.

Kreuter, M., Farrell, D., Olevitch, L., & Brennan, L. (2000). Tailoringhealth messages: Customizing communication with computer technol-ogy. Mahwah, NJ: Erlbaum.

Levy, D. T., Hyland, A., Higbee, C., Remer, L., & Compton, C. (2007).The role of public policies in reducing smoking prevalence in California:Results from the California tobacco policy simulation model. HealthPolicy, 82, 167–185.

McBride, C. M., Emmons, K. M., & Lipkus, I. M. (2003). Understandingthe potential of teachable moments: The case of smoking cessation.Health Education and Research, 18, 156–170.

McLeroy, K. R., Bibeau, D., Streckler, A., & Glanz, K. (1988). Anecological perspective on health promotion programs. Health EducationQuarterly, 15, 351–377.

National Center for Chronic Disease Prevention & Health Promotion.(2006). Smoking prevalence among U.S. adults—Tobacco informationand prevention source (TIPS) [Online]. Retrieved from http://www.cdc.gov/tobacco/data_statistics/index.htm

Pawson, R., Greenhalgh, T., Harvey, G., & Walshe, K. (2005). Realistreview: A new method of systematic review designed for complexpolicy interventions. Journal of Health Services Research and Policy,10, S21–S39.

Piper, M. E., Piasecki, T. M., Federman, E. B., Bolt, D. M., Smith, S. S.,Fiore, M. C., et al. (2004). A multiple motives approach to tobaccodependence: The Wisconsin Inventory of Smoking Dependence Motives(WISDM-68). Journal of Consulting and Clinical Psychology, 72,1395–1354.

Raghunathan, T. E., Lepkowski, J., Van Hoewyk, J., & Solenberger, P.(2001). A multivariate technique for multiply imputing missing val-ues using a sequence of regression models. Survey Methodology, 27,85–95.

Rodu, B., & Godshall, W. T. (2006). Tobacco harm reduction: An alter-native cessation strategy for inveterate smokers. Harm Reduction Jour-nal, 21, 37.

Rubin, D. B. (1976). Inference and missing data (with discussion). Bi-ometrika, 63, 581–592.

Shiffman, S., Gitchell, J. G., Warner, K. E., Slade, J., Henningfield, J. E.,& Pinney, J. M. (2002). Tobacco harm reduction: Conceptual structureand nomenclature for analysis and research. Nicotine & Tobacco Re-search, 4, S113–S129.

Sorensen, G., Barbeau, E., Hunt, M. K., & Emmons, K. (2004). Reducingsocial disparities in tobacco use: A social-contextual model for reducingtobacco use among blue-collar workers. American Journal of PublicHealth, 94, 230–239.

Spence, J. C., & Lee, R. E. (2003). Toward a comprehensive model ofphysical activity. Psychology of Sport and Exercise, 4, 7–24.

Tunis, S. R., Stryer, D. B., & Clancey, C. M. (2003, September 24).Practical clinical trials. Increasing the value of clinical research fordecision making in clinical and health policy. JAMA, 290, 1624 –1632.

Windsor, R. A., Li, C. Q., Boyd, N. R., Jr., & Hartmann, K. E. (1999). Theuse of significant reduction rates to evaluate health education methodsfor pregnant smokers: A new harm reduction behavioral indicator?Health Education Behavior, 26, 648–662.

788 GLASGOW ET AL.