Smoking is the single most preventable cause of illness and premature death in the world and is an important risk factor for 6 of the 8 leading causes of death, including several types of cancer, cardiovascular diseases, and respiratory diseases [1]. Consequently, smoking is related to a reduced quality of life and places a burden of €4 to €7 billion on health care [2-4]. Thus, quitting smoking is important, not only to improve individual and population health, but also to reduce smoking-related health care costs.

Extensive evidence exists on the clinical effectiveness of behavioral interventions for smoking cessation [5-7]. Brief advice from a general practitioner is one of these effective smoking cessation interventions [8]. However, general practitioners and practice nurses often report a lack of time and skills to provide their patients with elaborate smoking cessation advice [9,10]. Another behavioral intervention that has proven to be effective in increasing smoking cessation rates up to 13 months is computer tailoring [5,7,11-16]. Computer tailoring entails the adaption of the content of an intervention to participants’ individual characteristics by using computer programs [17]. Most often, a questionnaire is used as a screening instrument [12,18,19]. The answers smokers provided on the questions in this screening instrument are accumulated into a large data file and are subsequently matched with relevant feedback messages that are ultimately combined into a tailored feedback letter. Tailored interventions are more effective in attracting and keeping the smoker’s attention [17,20], resulting in better processing of information [21]. A single tailored feedback message is successful in increasing cessation rates [15], but dynamically tailored feedback provided on multiple occasions can be even more effective [11,12,22]. Due to the automatic generation of the tailored feedback and the fact that computer-tailored interventions are increasingly delivered online [23,24], the integration of an Internet-based computer-tailored program in the general practice setting might limit the burden on health professionals and patients, reduce facility and administrative costs, and could potentially be time- and cost-saving. As a combination of effective interventions was expected to achieve higher abstinence rates than either of the 2 alone [25], our research team has developed a smoking cessation intervention consisting of Internet-based multiple computer tailoring and a single tailored counseling session by a practice nurse.

Despite the proven clinical effectiveness of behavioral smoking cessation interventions, information about their relative cost-effectiveness is limited. Previously, several cost-effective smoking cessation interventions have been developed [26-28]. However, the interventions studied all involved the use of smoking cessation medication and investigated the cost-effectiveness of referrals to intensive counseling combined with pharmacotherapy [27], reimbursement of smoking cessation support [28], or the smoking cessation drug varenicline [26]. With regard to behavioral smoking cessation interventions, a computer-based smoking cessation intervention for primary care professionals was successful in increasing abstinence rates and quality-adjusted life years (QALYs) among patients [29]. In addition, in response to a call in the Journal of Medical Internet Research for research to economically evaluate eHealth interventions [30], cost-effectiveness and cost-utility studies of Internet-based interventions aimed at alcohol reduction [31] and depressive symptom treatment [32,33] have been initiated. However, to our current knowledge, no such studies have yet been conducted concerning the cost-effectiveness and cost-utility of an Internet-based smoking cessation intervention.

Therefore, the objective of the present study was to compare the cost-effectiveness and cost-utility of (1) an Internet-based multiple computer-tailored smoking cessation program combined with a single tailored counseling session by a practice nurses, (2) only an Internet-based multiple computer-tailored smoking cessation program, and (3) care as usual, defined as practice nurses’ standard care regarding smoking cessation.

Economic evaluation studies aim to determine the costs and effects associated with an intervention and to compare these with the costs and effects of other interventions and/or current practice [34]. They usually consist of 5 steps [35], which are listed in Textbox 1. For a more extensive and detailed description of each of these steps, refer to Multimedia Appendix 1.

The present economic evaluation study was trial-based, embedded in a randomized controlled trial (RCT) that tested the effectiveness of an Internet-based multiple computer-tailored smoking cessation program and tailoring counseling by practice nurses. This 3-armed RCT was conducted among Dutch adult smokers and had a follow-up period of 12 months. From May 2009 to June 2010, 91 practice nurses working in different Dutch general practices throughout the Netherlands recruited smoking patients for participation in the RCT. To aid recruitment, several recruitment materials were provided (eg, desk displays, posters, and business cards). Smokers interested in participation could sign up for the study on the study website. There, information was provided about the objectives of the study, the randomization procedure, and the incentive provided when respondents completed all questionnaires (ie, a €10 gift voucher). When signing up, participants were able to choose their own username and password and were informed that no one but the research team would be able to retrieve these passwords. After providing informed consent, participants were randomized into 1 of the 2 intervention groups (multiple tailoring and counseling or multiple tailoring only) or into the usual care control group. Randomization took place at the participant level by means of a computer software randomization device.

The trial design was approved by the Medical Ethics Committee of Maastricht University and the University Hospital Maastricht (MEC 08-3-037; NL22692.068.08), and is registered with the Dutch Trial Register (NTR1351). A more detailed description of the study design has been published elsewhere [25].

Participants were eligible for participation if they smoked, were motivated to quit within 6 months, were 18 years or older, and were able to read and understand Dutch sufficiently to read study materials and participate in the trial. Moreover, they had to have access to the Internet. This resulted in a total of 414 eligible smokers.

Figure 1 presents an overview of the intervention components in each of the study groups.

The Internet-based multiple computer-tailored smoking cessation program was based on a previously developed single computer-tailored intervention [12,15] for which the I-Change model (ICM) formed the theoretical framework [36]. As was its predecessor, the attitude-social influence-efficacy model [37], the ICM is a theory of behavioral change which incorporates theoretical concepts from several sociocognitive models, including the transtheoretical model [38], the theory of planned behavior [39], social cognitive theory [40], and the health belief model [41]. According to the ICM, the most proximal predictor of behavior is the intention to perform this behavior. Intention is predicted by 3 motivational constructs, attitude, perceived social influence, and self-efficacy, which in turn can be predicted by several premotivational factors, such as awareness, previous experience with the same and related behaviors, biological factors, and sociocultural factors. To overcome barriers that increase the well-known gap between intention and behavior (eg, [42]), the ICM proposes ability factors, such as an individual’s skills to refrain from smoking, and the formation of action plans. The ICM has been used successfully to develop several other effective computer-tailored programs [12,15,18]. While filling out the first online questionnaire (ie, baseline questionnaire), all participants were asked to set a date within the next 4 weeks at which they would attempt to quit smoking. They received a total of 4 feedback letters: at baseline, 2 days after the quit date they had set for themselves at baseline, after 6 weeks, and after 6 months. Feedback was personalized and tailored to several participant characteristics: gender, attitude, social influence and self-efficacy, intention to quit smoking, action planning, and smoking behavior. Feedback letters were iterative: the second, third, and fourth feedback letters did not only concern the participant’s present state, but also referred to changes participants had made since they were included in the program. Most feedback letters consisted of 4 to 5 pages and 7 components: (1) introduction, including specific feedback on the respondent’s smoking behavior and on his/her intention to quit smoking and to maintain nonsmoking; (2) feedback on the respondent’s attitude, including perceived advantages (pros) and disadvantages (cons) about smoking and quitting smoking; (3) feedback on perceived social influence (not) to smoke; (4) feedback on the respondent’s reported self-efficacy to refrain from smoking in specific situations, including suggestions on how to cope with these situations; (5) feedback on the extent to which respondents were planning to undertake specific actions (action plans) while preparing their quit attempt; (6) feedback on how to cope with situations in which it might be difficult not to smoke (coping plans), including the formulation of personal plans in the shape of if-then statements [28]; and (7) ending. Participants could access their feedback letters directly online after questionnaire completion. Additionally, feedback letters were sent to the participant by email. In both cases, feedback letters could be printed. An example of a tailored feedback message is provided in Multimedia Appendix 2.

After receiving the first tailored feedback, participants in the multiple tailoring and counseling group were prompted to schedule a counseling meeting with their practice nurse within 6 to 8 weeks. They received this counseling session instead of the third tailored feedback letter at the 6-week follow-up. A counseling protocol was provided to assist practice nurses in guiding these counseling sessions. This protocol consisted of 3 chapters guiding on 3 different types of participants: smokers who had quit successfully, smokers who had quit but relapsed, and smokers who had not yet quit. The content of the counseling session was developed to be as similar as possible to the content of the computer-tailored feedback and was also tailored to the participant characteristics mentioned previously. After 6 months, practice nurses were instructed to call their patients to ask them about their progress toward permanent cessation and, if needed, to provide them with additional cessation support.

Participants randomized in the usual care group received smoking cessation guidance according to participating practice nurses’ standard practice, which can vary from a brief intervention consisting of a single recommendation to stop smoking to more intensive interventions [43,44].

Self-reported online questionnaires were used to assess both costs and effects. Questionnaires were administered at baseline and at 6-week, 6-month, and 12-month follow-ups. When follow-up questionnaires were not completed by 1 week after the invitation, an email reminder was sent. At 12-month follow-up, this email reminder was followed by a phone call to collect data.

All analyses were conducted according to the intention-to-treat principle. Missing data for costs, EQ-5D items, overall tobacco consumption, and addiction level were replaced by mean imputation by using participants’ scores on the previous and next measurement. When mean imputation was not possible because of missing data on multiple measurement points, missing data were replaced using the last observation carried forward (preferred choice) or next observation carried backward method. Missing data for smoking abstinence were replaced using a negative scenario; participants lost to follow-up were considered still smoking.

Of the 414 participants who were eligible for participation, 163 were randomized into the multiple tailoring and counseling group, 132 into the multiple tailoring group, and 119 into the usual care group. No baseline differences were found between the 3 groups (Table 1). After 12 months, 231 (55.8%) of the 414 participants could be followed up. Participants lost to follow-up were significantly younger than those who remained in the study (P=.01). Of the 46 participants who reported prolonged abstinence at the 12-month follow-up, 30 successfully completed a cotinine test and all cases of self-reported abstinence were confirmed. After imputation of missing values, total cost data were available for 370 participants (89.2%) whereas effect data were available for 414 (100%) participants for abstinence measures, 409 (98.8%) for addiction level, and 384 (92.8%) for QALYs.

A significant difference was found between the multiple tailoring and counseling and the usual care groups with regard to annual health care–related costs per patient, with significantly higher costs in the multiple tailoring and counseling group (Table 2). No differences were found between the 3 groups with regard to any specific type of health care–related costs (Table 2), nor regarding effects on abstinence, QALYs, or addiction level assessed at 12-month follow-up (Table 3).

Table 4 shows that for participants in the multiple tailoring and counseling group costs were higher, whereas effects were lower than in the usual care and multiple tailoring groups; thus, multiple tailoring and counseling was dominated by the other 2 treatments. Comparing multiple tailoring with usual care showed that multiple tailoring resulted in more costs, but also in more effects. Compared with usual care, €5100 had to be paid within the multiple tailoring group for each additional abstinent participant (Table 4).

The CEA showed that until a threshold value for the WTP of €5100 per abstinent participant, usual care was probably the most efficient treatment. However, from a WTP of €5100 or higher, multiple tailoring was probably most cost-effective (Table 5). With the accepted Dutch cutoff point of €18,000 per QALY for preventive interventions [57], multiple tailoring would be the preferable treatment. These results are visually displayed in the cost-effectiveness acceptability curve (CEAC), showing the probability of each treatment being preferable to the other 2 treatments for varying levels of the WTP per additional abstinent participant (Figure 2). Sensitivity analyses supported these results (Table 5).

Results from secondary analyses showed that with 7-day point prevalence abstinence, a high probability was found (ie, 88%, with a WTP of €18,000 per abstinent participant) that multiple tailoring was the most cost-effective treatment. Regarding the level of addiction, however, it was most probable that multiple tailoring would be least efficient (Table 5).

With regard to QALYs gained, Table 4 shows that multiple tailoring was dominated by usual care because this treatment was both more expensive and less effective. Furthermore, cost-utility analyses showed that multiple tailoring and counseling was more expensive, but also more effective than usual care and multiple tailoring in increasing the number of QALYs gained. This resulted in an incremental cost of €40,300 per QALY gained when comparing multiple tailoring and counseling with usual care, and in an incremental cost of €18,367 per QALY when comparing multiple tailoring and counseling to multiple tailoring.

With a WTP of €18,000 per abstinent participant, the CUA showed that usual care would probably (ie, 64%) be the most efficient treatment (Table 5). Although decreasing this threshold value to €0 led to an increased probability that usual care would be most efficient, increasing this threshold led to a lower probability of usual care being most preferable. With a WTP of almost €40,000, usual care and multiple tailoring and counseling would be equally preferable. These results are further illustrated in the cost-utility acceptability curve (CUAC) (Figure 3). Sensitivity analyses showed similar results (Table 5).

To our current knowledge, this was the first study to determine the cost-effectiveness and cost-utility of a behavioral smoking cessation intervention consisting of Internet-based computer tailoring with and without counseling by a practice nurse. The results presented suggest that participants who received the Internet-based multiple computer-tailored program and tailored counseling by their practice nurse reported significantly more annual health care-related costs than participants who received care as usual. A potential explanation for this finding might be that smokers were prompted by the tailored feedback they received to ask for more smoking cessation guidance, eg, additional counseling sessions with the practice nurse or a prescription for smoking cessation medication. Although participants who received the Internet-based program only might have had the same tendency, the practice nurses for participants in the multiple tailoring and counseling group might have been prompted by their patients’ visit to offer them more smoking cessation help. Because the current smoking cessation guidelines in the Netherlands recommend more than 1 counseling session [44,58], this is not unlikely. Interestingly, though the significant difference in total health care-related costs could not be explained by a difference in any particular type of health care-related costs. For instance, no significant differences were found with regard to costs spent on general practitioner or practice nurse consultations or on the use of smoking cessation aids.

Furthermore, the present study showed that the Internet-based multiple computer-tailored smoking cessation program would probably be the most cost-effective of the 3 treatments under study. Although no similar studies yet exist within the field of smoking cessation, this finding is in-line with findings from recent studies toward the cost-effectiveness of Internet-based interventions aimed at other health-related behaviors or health problems [31,32]. Compared with current practice, the incremental costs per abstinent participant associated with the Internet-based multiple computer-tailored smoking cessation program were €5100. This is slightly more than what was found in previous studies [59,60]. However, one of these studies only included costs directly related to the interventions received [59], whereas we conducted our economic evaluation from a broader societal perspective. Nevertheless, the interpretation of the incremental costs per abstinent participant is difficult because no information exists on the amount of money that society is willing to pay per abstinent participant. Although a WTP of €18,000 per QALY is an accepted Dutch cutoff point [55], no such cutoff point exists with regard to smoking abstinence rates. To enable the interpretation of the incremental costs per abstinent participant, future research should aim at identifying an acceptable cutoff point for the WTP per abstinent participant.

Regarding cost-utilities, the results suggest that care as usual would probably be the most preferable of the treatments studied. A potential explanation for this finding might be that the follow-up period of 12 months was not sufficiently long for the beneficial effects of the intervention on smoking abstinence to be translated into detectable changes in quality of life, as recent ex-smokers are known to suffer from withdrawal symptoms [61]. A potential solution would be to use short-term trial data as input for a model predicting the effects of smoking cessation interventions on long-term quality of life, a technique used in several recent studies [28,29,60]. Although trial data served as input for these models, several assumptions had to be made to build these models, bringing about additional uncertainty in the results presented [34, 60]. Another solution would be to lengthen the follow-up period of clinical trials to gather longer-term data on quality of life. Although this might imply an increased burden on the participant, it might be needed to establish the intervention’s cost-utility in an as certain as possible way. However, as previous studies were able to detect a positive association between quitting smoking and quality of life during a 12-month follow-up period [62,63], additional explanations for this finding need to be sought. One such explanation might be that within Dutch general practices, care as usual for smoking cessation is rather intensive. Anecdotal evidence suggests that practice nurses usually offer 4 to 6 consultations as part of smoking cessation care (unpublished). Although participants in the multiple tailoring and multiple tailoring and counseling groups received 4 and 3 tailored feedback letters, respectively, a primarily Internet-based program might have been perceived as more distant and less intense than care as usual. As a result, participants in the usual care group might have established a better or stronger social bond with their practice nurse than participants in the multiple tailoring or multiple tailoring and counseling groups. This assumed social bond [64] might subsequently have resulted in positive effects on quality of life among usual care participants.

The present study aimed to contribute to the literature by examining the cost-effectiveness and cost-utility of an Internet-based smoking cessation intervention, something that has not been done before to our knowledge. In the present study, in addition to generic quality of life, disease-specific effects of the intervention (ie, smoking abstinence) were taken into account. To facilitate the comparison of the cost-effectiveness of interventions targeting different diseases, effects are usually assessed in terms of quality of life. However, to compare smoking cessation interventions more specifically, disease-specific effect measures might be more informative.

Nevertheless, the present study also had its limitations. First, it suffered from relatively high dropout rates. High rates of attrition seem to be inherent to many Internet-based interventions and dropout rates of 44% are not uncommon [24,65-67]. As a consequence, however, there was not sufficient power for us to conduct a complete-case analysis as part of the sensitivity analyses. Secondly, because we expected higher attrition rates in the multiple tailoring and counseling group, slightly more participants were randomized into this intervention group at baseline. However, attrition rates appeared to be similar among the groups, resulting in a skewed distribution of participants with 163 participants in the multiple tailoring and counseling group, 132 in the multiple tailoring group, and 119 in the usual care group. Although we do not expect that this has biased our results, this finding of no selective attrition can be considered valuable in the design of future trials.

The Internet-based multiple computer-tailored program seemed to be the most cost-effective treatment when smoking abstinence was the outcome measure. However, the cost-utility, using quality of life as the outcome measure, was probably the highest with care as usual. To enable the interpretation of the incremental costs per abstinent participant found in the cost-effectiveness analyses, future research should aim at identifying an acceptable cutoff point for the WTP per abstinent participant.