Forty-six cardiac wards of urban, leading clinical and academic hospitals throughout the Netherlands were approached for potential participation. Thirteen of these wards that did not offer any form of smoking cessation intervention to cardiac patients (unlike care as usual) were recruited. Five of these wards were still rejected by the research team because of incomparability with the other wards in terms of their constitution (i.e. they had a combined heart and lung unit). The remaining eight cardiac wards were accepted for the study.

In line with the Dutch Medical Research Ethics Committee (MERC) rules [66] the study protocol was submitted for approval to the MERC of the VU medical center Amsterdam. After approval by this MERC (MEC 2009/215; NL27637.029.09), the MERCs and/or Board of Directors of each hospital endorsed execution of the study. The study was registered with the Dutch Trial Registration (NTR2144).

An experimental study using cross-over randomization at the ward level with a baseline measurement at hospital admission and post-measurements at six and 12 months after baseline was conducted. All cardiac wards first provided care as usual and subsequently the two experimental conditions one after the other. Table 1 shows a schematic representation of the serially implemented conditions. After completion of care as usual, four of the cardiac wards started implementing TC, and after its completion they implemented FC. The other four cardiac wards implemented the experimental conditions in reverse order. After each condition there was a one month period without inclusion of any patient.

Power and sample size calculations were conducted for the main outcome, seven-day point prevalence abstinence (PPA) after 12 months. Power calculations (one-sided; p < 0.05, Power = 0.80) showed that 193 patients per condition were needed to detect significant differences in PPA and possible interaction effects of the interventions with motivation to quit and SES. Advanced statistical methods, however, allowed adjustment for potential baseline confounding variables and improved the statistical power. Therefore, the required sample size was reduced by 30% [67], resulting in the need for N = 135 in each condition. Assuming 15% attrition at the 12-month follow up, 155 patients per condition needed to be recruited, totalling 465 patients for the three conditions.

After approval was received from the Board of Directors and Ethical Review Committees of each hospital, the face-to-face counsellors (16 in total, i.e. two nurses from each of the cardiac wards that participated in the study) received training to provide FC. Ward nurses and cardiologists received written and oral instructions of the study with a written stepwise protocol and information sheet on the inclusion process of eligible patients, a flowchart with instructions and addresses for referral. Over an 18-month period beginning in December 2009, the nurses asked all patients admitted to the cardiac ward if they had smoked in the month prior to admission. If the answer was yes, the nurses reviewed the information sheet with the inclusion criteria, approached medically stable patients at the bedside, provided the necessary information about the study to patients, and invited eligible smoking patients to take part in the study. If patients agreed to take part, the nurses asked them to sign an informed consent form. Subsequently, at hospital admission, nurses administered the questionnaire to patients and registered relevant patient data on a separate form. Nurses also produced a summary of patients’ prescribed medication from their case history.

Inclusion criteria for patients were: smoking on average ≥5 cigarettes per day in the month prior to admission or, if not smoking, having quit smoking less than four weeks before admission; being ≥18 years of age; being admitted to the cardiac ward for less than 96 h and being hospitalised because of a coronary heart disease (acute coronary syndrome, stable angina, and other forms of chronic and acute heart diseases) following the standards of the ICD-10 [68]. Exclusion criteria were language limitations that would impede completion of the self-administered questionnaire, a medically unstable cardiac situation or cognitive impairments.

As shown in Figure 1, 245 cardiac patients were included in the usual care condition, 223 in the experimental condition of TC, and 157 cardiac patients in the experimental condition of FC. With inclusion rates varying from 42 to 121 patients per cardiac ward, rates varied considerably per cardiac ward (see Table 2).

As presented in Table 3, patients were predominantly male with a mean age of 57 years. Education levels were relatively equally distributed over low, intermediate and high education. Most of the patients had an admission diagnosis of acute coronary syndrome (ACS) or had been diagnosed with ACS previously and were admitted to the hospital for treatment with a percutaneous coronary intervention (PCI) or a coronary artery bypass graft (CABG). During the six months prior to admission a few patients had been admitted to the hospital owing to a cardiac event. Nicotine dependence was moderate, and patients reported smoking 21 cigarettes on average per day before hospital admission. Although the majority of the patients had not made any quit attempt over the past 12 months, one-third reported on admission that they had not smoked over the past seven days. There were no significant differences between the three groups in terms of demographic characteristics. Nicotine dependence was significantly higher, however, in the control group than in the experimental groups.

Six months after discharge (T1) patients were interviewed by telephone by members of a professional call agency to obtain follow-up data (see Figure 1). The 12-month follow-up measurements are still in progress (T2). The smoking cessation outcomes are assessed by means of validated reliable questions [86‐88]. Telephone interviews are conducted in order to decrease the likelihood of attrition [10, 89]. One postal reminder letter is sent 10 days before each of these telephone interviews. All patients who complete the 12-months interview are also invited to the hospital for biomedical tests of blood pressure, cholesterol and saliva.

Cardiac nurses are responsible for conducting the blood pressure tests, the cholesterol tests and the cotinine tests at the cardiac nursing ward. Cardiac nurses ask patients while conducting the blood pressure test if they have been smoking over the past seven days to assess point-prevalence abstinence. If patients claim not to have smoked any cigarettes, their saliva is tested on cotinine residuals by means of the NicAlert Cotinine Test Strip to test whether the self-reported smoking behavior is accurate. Results can be either positive (no cotinine detected) or negative (cotinine detected). Cardiac nurses register the results of the blood pressure, cholesterol and cotinine tests and send the findings to the research team.

Data is analyzed with SPSS. Descriptive statistics are used to describe the sample, and to compare the groups at baseline, ANOVA analyses and chi-square analyses are applied. In order to determine possible selective loss at follow-up, attrition analyses are conducted. The effects of the interventions on smoking cessation outcomes and possible interaction effects (condition * SES/condition * intention to quit) are tested by means of logistic regression analyses. The hospital group size of eight is inadequate for reliably testing the existence of variation by a multi-level model [102]. Therefore, a fixed-effect regression approach is likely to be used to eliminate possible hospital-level effects. Baseline variables are included as covariates to correct for possible baseline differences, to reduce unexplained variance and to increase the power of the tests. To analyse effects on rehospitalisation, new cardiac events, number of visits to the cardiologist and cardiovascular-related death, multiple regression analyses are used. Effects of the interventions on blood pressure, TC/HDL cholesterol, and intentions to quit are tested by means of repeated measures analyses of covariance. Logistic regression analyses are also used to examine predictors of smoking cessation overall and within subgroups. Patients in the experimental groups are compared by paired sample t-tests for level of appreciation of the intervention. Economic evaluation analyses are performed by applying specific cost-effectiveness calculation formulas. Moreover, a health-economic Markov Model is built to estimate long-term costs and health effects.

This study is a multi-centre trial carried out in eight hospital cardiac wards from different geographical regions of the Netherlands. Most randomised controlled trials on preventative interventions for smoking cardiac patients are conducted in only one or a few hospitals, so the information gathered from eight hospital wards improves generalisation of the study results. Cross-over randomisation was conducted at the ward level by sequentially implementing the conditions, which was done to overcome problems related to traditional randomised controlled trials. This design meant contamination between patients was avoided, as well as learning effects on ward nurses. The chosen design also circumvented nurses’ mistakes in performing the interventions. Moreover, the experimental design allowed us to control for confounding effects such as seasonal influences or other extraneous events. Last but not least, a cost-effectiveness study was carried out alongside the standard clinical trial, which allows us to estimate the additional costs and benefits of the interventions compared with a control group receiving usual care.

Although we made efforts to include all eligible smoking cardiac patients within the a time frame of one year and to register data on those eligible patients not enrolled in the study, the baseline results suggested that neither task was performed as intended. Nurses were asked to register the refusals of patients not enrolled in the study but this was not well done. There are indications that in some hospitals there were more refusals than in others. The inclusion period was expanded to 18 months as the inclusion of smoking cardiac patients in hospital cardiac wards took longer than expected. Low inclusion rates were related to ward factors such as insufficient patient admissions, few patients who met the inclusion criteria of the research, and limited admission capacities of cardiac wards. Other factors hindering patient inclusion in the study were high workloads on cardiac wards, non-collaborative nursing teams, and little interest in smoking cessation programmes at the general hospital. These factors have previously been shown to impede implementation of smoking cessation interventions in cardiac wards [103‐105]. The inclusion rate in the FC group was relatively low which might be related to its time-consuming treatment form, resulting in a skewed distribution of patients with 157 patients in the FC group, 223 in the TC group and 245 in the UC group. Moreover, inclusion rates varied considerably between cardiac wards and this suggests the likelihood of non-representative outcomes of the effectiveness study for those cardiac wards that only included a few patients per group. We also note that there were differences in nicotine dependence between the experimental groups and the usual care group, as nicotine dependence was significantly higher in the experimental groups. Altogether these considerations pose the limitation of a selective sample of cardiac patients, implying the potentially limited generalisability of the study results.