The study was designed as a two-armed, unblinded cluster-randomized controlled trial. At the beginning of the study, around 2000 Bavarian general practitioners (GPs) received a fax by the Bavarian Association of General Practitioners with information about the research project. All interested GPs were sequentially registered for randomization. After giving written consent, the participating practices were randomized either to the intervention or the control arm. The sequence of randomization (allocation 1:1) was provided by a methodologist, who did not participate in the execution of the study, via the program Research Randomizer [11]. Randomization was concealed by using sequentially numbered, opaque, sealed envelopes held by the study coordinator. Before starting the recruitment of patients, physicians and practice nurses received detailed instructions by the research team on the study process (both intervention and control group) and on the coaching program (only intervention group).

Physicians assigned to the control arm were asked to change nothing in their usual way of counseling and to treat participants in the same manner as if they would have been non-participants. There was no structured documentation of the care provided. The patients recruited by intervention practices received free access to the Web-based coaching program. The patients of practices participating in the control arm were advised by the GPs in their individual way of usual care to quit smoking. The study was approved by the Medical Ethics Committee of the Technische Universität München (April 19, 2011) and was in accordance with ethical standards for human experimentation established by the Declaration of Helsinki. All participants gave written informed consent. A data and safety monitoring board was established before the beginning of the study.

Participating physicians were general practitioners in Bavaria, Germany. The GPs were requested to recruit individuals with the desire for smoking cessation (see Multimedia Appendix 2). Individuals at least 18 years old and with Internet access were potentially eligible. Exclusion criteria were aged younger than 18 years, insufficient German language skills, and lack of Internet access. Further exclusion criteria were psychiatric disorders and posttraumatic stress disorder.

After the GP decided that patient should participate, an information form was given and discussed with the patients, and a participation form had to be signed. At the same time, the baseline data acquisition took place. All participants were asked to fill in a standardized questionnaire together with the GP. The standardized questionnaire comprised the following information: age, sex, height, weight, physical activity, years of tobacco use, number of cigarettes smoked per day, number of previous quit attempts, use of current nicotine replacement therapy (NRT), and reasons for smoking. Participants of the intervention group received a free Web-code. The physician filled in a form together with the patient with information about the potential existence and grade of a chronic obstructive pulmonary disease. This form with the Internet code was used by the patient for specification during the registration process of the Internet program. Participants of both groups were requested after 12 weeks to document the follow-up evaluation together with their physician. The follow-up comprised again information about smoking status, weight, physical activity, number of cigarettes smoked per day, and use of current NRT. At follow-up, a biochemical validation of the self-reported cessation status was also implemented through a cotinine urine test. Cotinine is detected in the urine for 2-4 days after the use of tobacco.

Physicians in the intervention group received €50 per participant for time and effort. Physicians in the control group received €25. Participants in the intervention group received free access to the smoking cessation program, which usually costs €79. Participants in the control group received €10 as an incentive to come into practice for follow-up investigation after 12 weeks. No methodical changes were made during the entire study period.

Anamnestic and health data were documented in a structured registration form including information about the potential existence and grade of a chronic obstructive pulmonary disease from the GP. The patient received a copy of this form in order to use the health data for subscribing via Internet into the HausMed coaching program [12] at home. A specific program was installed to allow the participants to log in without charge. After completion of a pre-assessment, the program generated individual coaching based on the given information of the physicians (registration form), the physical characteristics, and the everyday behavior of the participants.

The coaching program is based on the generally accepted principles of cognitive behavioral therapy and combined psychoeducation and motivational techniques with behavioral-therapeutic elements [13]; for example, education, realistic goal-setting, and individual resources, and in particular, the behavioral change theory targeted to smoking cessation by using inexpensive Internet and mobile technologies in combination with existing health care resources of GPs. The content of the coaching program aimed at achieving a lasting change of behavioral patterns with the help of individualized education, motivation, exercise guidance, daily SMS text messaging (short message service, SMS) reminders, self-monitoring via Internet and, finally, through active monitoring and approximately three telephone calls during the 12 weeks by the GPs or their staff. The framework of the program is based on the idea by Buchkremer and Batra, of the Department of Psychiatry and Psychotherapy, University of Tübingen, Germany [14,15]. The development and implementation of the Web-based smoking cessation program was carried out by WeCARE GmbH, Göttingen, Germany. The coaching program was subdivided into 12 different constitutive modules. Each module was performed for one week and contains particular tasks (Textbox 1), which were supported by corresponding daily SMS reminders. The participant had to perform a specific task each day and received a daily SMS regarding that specific task.

The reminder contained adapted information to maintain motivation, impart daily tips, and encourage daily performance of the respective task. The specific daily tasks were offered on the first day of each module. The coaching program also offered a variety of printed material (emergency plan, relaxation exercises, questionnaires, information, self-agreements, etc), which was connected to the respective task and included interactive buttons, video clips, and learning progress quizzes to examine learning success (Figures 1 and 2).

At the end of each week, participants were asked to give feedback via the Internet concerning their condition and level of motivation and whether or not they did their weekly tasks. Participants could also communicate among themselves on a forum or asked a HausMed team member in case they had any questions. The active monitoring (or rather supervising) of the entire 12-week coaching course was carried out by the GP through a separate login account in a secured physician area (motivation, condition, and status of the module exercise). In addition to that, three specified telephone calls from the GP or a qualified practice nurse (Weeks 2, 4, and 12) were implemented to motivate and support the participants. If either a participant’s motivation or condition declined notably at any point during the coaching period or if the module exercise was not completed, additional counseling from the GP or practice nurse was given over the telephone. There was no limitation to the frequency of website use, but participants were given a goal of using the website at least once a week and GPs were advised to log in into the program twice a week. No changes were made to the coaching program within the entire study period.

The primary outcome measure was biochemically confirmed smoking status 12 weeks after inclusion into the study by use of a cotinine urine test. Secondary outcome measures were self-reported smoking status, number of NRTs, weight in kilograms, number of smoked cigarettes per day, physical activity (range from 0-4), and breathing difficulties (range from 0-4). A higher number on the scale refers to a more frequent physical activity and to more breathing difficulties. No changes were made on the trial outcome measures within the entire study period.

Sample size calculation was performed with G*Power 3 correcting for the cluster design (intracluster correlation coefficient=.05, average cluster size=3) for two-sided testing (alpha is 5% and a power of 80%). For the expected effect, abstinence rates of 30% versus 10% were assumed. Using these assumptions, the calculated total sample size for primary outcome smoking cessation was 152 participants. Taking expected attrition into account, we aimed at recruiting a total of 180 participants in about 80 general practices.

Baseline data are presented descriptively. Group differences were calculated for all participants whose smoking status was available at baseline and follow-up (completer collective). Sensitivity analysis was performed by an intent-to-treat analysis assuming that participants with missing values had no smoking status change at all. The strongly variable cluster size caused major numerical problems in the linear mixed model analysis. As it was not possible to adjust for intracluster correlations properly and because of the high variability of patients in practices, it was decided to perform the main analysis using Fisher’s exact test without accounting for the clusters. For the main outcome smoking status, we also performed secondary analyses based on logistic regression analyses with adjustments for age, gender, height, and number of cigarettes smoked per day. We further conducted generalized estimating equations as sensitivity analysis to account for practices as patient clusters. All analyses were performed using SPSS version 19.0.

We found that the Web-based coaching program in combination with telephone counseling and monitoring in general practice was not effective for achieving smoking cessation compared to usual care.

Some trials have already shown that smoking cessation programs can be delivered effectively via the Internet [8,16-18], although Web-based interventions are often accompanied by a high attrition rate [19,20]. The evidence has been summarized in a recent meta-analysis by Chen et al [9]. They concluded that computer and other electronic aids increase to a small extent the likelihood of prolonged smoking cessation compared to no intervention. Another meta-analysis of Web- and computer-based smoking cessation programs indicates that there is currently sufficient evidence to support their use [21]. They stated within a meta-analysis of 22 randomized controlled trials (RCTs) that Web- or computer-based smoking cessation programs led to a 1.5 times higher abstinence rate compared to control groups. The effect of Web- or computer-based programs was therefore similar to that of counseling interventions. Their pooled cessation rate for Web- or computer-based programs over the long term (12 months) was 9.9% (95% CI 8.9-10.9). The advantage of a Web-based intervention compared to a control group cannot be confirmed by the present results, although our abstinence rates from both groups in the short term were comparable to theirs. In a Cochrane review, Civljak et al [10] also detected no consistent effects from trials that compared Internet interventions with usual care.

For example, Muñoz et al [22] suggested that quit rates obtained by using Internet interventions for smoking cessation are comparable with quit rates reported from smoking cessation therapies or smoking cessation groups. Our present result supports this fact as there was no significant difference between the intervention and the usual care in a primary care setting. Previous findings with a biochemical confirmation of the self-reported cessation support our results. Patten et al [23] revealed in adolescent smokers no significant treatment differences between a brief office intervention and a Web-based intervention. Their smoking abstinence rates after 24 weeks were 12% versus 6% for the brief office intervention and Internet-based intervention, respectively. The present discrepancy of the self-reported smoking status (17.6%) and the biochemical validation (9.8%) was noticeable even if this conspicuousness is based on only three individuals in the treatment group with a disconfirmed smoking status. It has been reported that biochemical validation does not modify the conclusions of low-intensity interventions trials [24]. Glasgow et al observed a disconfirmation rate between 4-5% over all groups. Compared to the present disconfirmation, we identified a similar rate only in the intervention group of 3.5-5.9%. In the control group, all conducted cotinine tests were negative so there was no disconfirmation to be documented. The self-reported cessation rate from 17.6% (complete-case) of the intervention group compared to the confirmed rate from 9.8%, beside the missing validation of 2% (n=1), leads to the suggestion of a certain overestimation from the present self-reported cessation rate. Nevertheless, this discrepancy could have still occurred by chance due to the small number of disconfirmed results.

The present study underlines how difficult it is to achieve smoking cessation in primary care and how difficult it is to motivate patients to quit smoking. Based on a parallel trial conducted with comparable conditions, we identified that a Web-based intervention with telephone counseling led to a significantly greater weight reduction [25] compared to usual care. This advantage was not shown regarding smoking cessation. Many reasons have been already identified to explain the low effectiveness of smoking cessation interventions in general practice. Patients’ low motivation to quit [26,27], patients’ low compliance [26,28], patients’ resistance to speak about smoking [29,30], lack of time [31,32], lack of economical reimbursement [28,33], lack of skills and low self-efficacy [32,34], consideration of smokers’ other problems [30,35], and an uneasy feeling when talking about smoking cessation [31,32] have been already identified as barriers. Notwithstanding, even interventions with a small effect are capable of substantially decreasing diseases that are associated with nicotine use [36].

Strengths of the present study were the embedding of the study in a realistic primary care setting and the use of a biochemical validation of the self-reported cessation status. However, some important methodological aspects for the interpretation of the study results need to be considered. First, the randomization of the present study was conducted at the practice level before individual participants were included. Thus, physicians knew whether they recruited patients for the intervention or the control group, which could lead to bias. Second, due to the highly variable cluster sizes the statistical analysis of our data was not straightforward. Classical linear mixed models taking the cluster design into account could not be used because of numerical problems. Therefore, we used simple Fishers’ exact test (which ignores intracluster correlation) and an additional multilevel analysis (which runs into problems when cluster sizes differ) as sensitivity analysis. Third, according to our power calculations the target number of participants was not completely reached due to a slow recruitment of participants, and at a certain point the study had to be stopped, which may have reduced the study’s ability to detect significant differences between the groups. Fourth, the proportion of participants without follow-up values was undoubtedly higher in the intervention than in the usual care group. This could be partly due to the fact that participants in the control group received a small financial incentive while those in the intervention group did not. Participants in the intervention group might also have been less willing to have an additional practice visit after completing the program than those in the control group who had little practice contact otherwise. Therefore, our complete-case analysis with the self-reported cessation might overestimate the rates to some extent. Within the intent-to-treat analysis, where the missing post values were replaced without a change of smoking status (baseline carried forward), the cessation rates were clearly smaller. Fifth, the content of the usual care was not further evaluated. The practitioners of the control group were asked to change nothing in their usual way of counseling and to treat their participants in the same manner as usual. There was no additional documentation of their counseling provided.

Our findings suggest that the tested Web-based coaching program in combination with telephone counseling and monitoring in general practice was not effective for achieving smoking cessation compared to usual care. The effect was similar to usual primary care and comparable to other Web-based interventions. We identified a discrepancy of self-reported smoking cessation and the biochemical validation, which should be reconsidered for further studies. The limited statistical power and the high drop-out rate may have reduced the study’s ability to detect significant differences between the groups. Further RCTs are needed in order to investigate long-term outcomes and interventions in larger populations, as well as in the contents of usual primary care.