The Effect of Smoking Prevalence at Worksites on Individual Cessation Behavior

Transcription

1 J Occup Health 2009; 51: Journal of Occupational Health The Effect of Smoking Prevalence at Worksites on Individual Cessation Behavior Chihiro NISHIURA 1, Rie NARAI 2, Takayuki OHGURI 2, Atsushi FUNAHASHI 2, Keiichirou YARITA 2 and Hideki HASHIMOTO 3 1 Department of Safety and Health, Tokyo Gas Co., Ltd., 2 Mazda Motor Corporation and 3 Department of Health Economics and Epidemiology Research, The University of Tokyo School of Public Health, Japan Abstract: The Effect of Smoking Prevalence at Worksites on Individual Cessation Behavior: Chihiro NISHIURA, et al. Department of Safety and Health, Tokyo Gas Co., Ltd. Objectives: To identify the effect of worksite smoking prevalence on individual cessation behavior. Methods: We conducted a cohort study at a Japanese worksite without a total ban on smoking, but with designated smoking areas. Baseline data were obtained in 2005 through a self-administered questionnaire from 15,229 workers in 322 work units as part of annual health checkups, and followed up in Data on smoking status, time to first cigarette after waking up, desire to quit, number of cigarettes smoked per day, and respiratory symptoms were obtained. Details like workers demographics, blue/ white-collar workers, night shift duties, administrative position, and work unit codes were obtained from the firm s administrative records. Smoking prevalence in work units was calculated as a ratio, standardized by age, and categorized into quartiles. Multiple logistic regression was used to predict cessation by smoking prevalence, adjusting for individual level variables. Results: In the lowest smoking prevalence quartile compared with the highest, odds ratios (95% CI) of cessation among those who smoked their first cigarette more than 30 min after waking up were 2.32 (1.06, 5.09) in white-collar units and 1.86 (0.98, 3.55) in bluecollar units, and that among those with a moderate desire to quit was 2.05 (0.94, 4.49) in white-collar units. Conclusion: Worksite smoking prevalence affects the likelihood of successful cessation, especially among those with less nicotine dependence and who are in the early stage of behavioral change. This suggests that serious consideration should be placed not only Received May 19, 2008; Accepted Oct 11, 2008 Published online in J-STAGE Dec 5, 2008 Correspondence to: C. Nishiura, Department of Safety and Health, Tokyo Gas Co., Ltd., Kaigan , Minato-ku, Tokyo , Japan ( chihiro.n@tokyo-gas.co.jp) on individual behavior modification but also on modification of worksite conditions. (J Occup Health 2009; 51: 48 56) Key words: Smoking prevalence, Cessation, Worksite, Social environment Smoking is one of the leading preventable causes of death and disability for both smokers 1) and nonsmokers 2). The health loss incurred by smoking negatively affects productivity and contributes to absenteeism 3). Therefore, smoking is regarded as one of the major targets of occupational health programs at worksites, to enhance the well being of both workers and organizations. The first guideline for smoking control at worksites was issued by the then Ministry of Labour in According to this guideline, individual education and a designated smoking area were recommended as a protocol. At present, a number of pharmaceutical and behavioral programs are available for encouraging smoking cessation. An alpha4beta2 nicotinic acetylcholine receptor partial agonist was reported to result in a significantly high likelihood of success in cessation, compared to a placebo 4, 5). Behavioral programs based on the transtheoretical model 6) and other behavioral theories have also proved fruitful 7). However, the one year success rate of cessation is approximately 30%, and the issue of relapse remains a major concern 5, 8). In addition to these individual level interventions, environmental level interventions are expected to be effective at inducing cessation behaviors 9, 10). Although, so far, there has been no evidence to support the effectiveness of designated smoking areas, a number of studies have shown the effectiveness of a total smoking ban at worksites, in the encouragement of individual cessation 11). A ban is effective not only with regard to smoking cessation at worksites, but also in off-duty situations 12), suggesting that environmental change such

2 Chihiro NISHIURA, et al.: The Effect of Smoking Prevalence on Cessation Behavior 49 as a no smoking policy would affect individual psychological intentions to stop smoking. The mechanism and conditions as to when and how environmental conditions affect individuals behaviors are debatable. Previous studies on the efficacy of a smoking ban had several limitations. Some studies failed to control individual level confounders such as psychological readiness to quit and physiological nicotine dependence 12 14). A mere comparison of smoking cessation rates across companies with/without a smoking ban might be biased by the difference in the socioeconomic status of workers, since companies that have a no smoking policy may be larger and more affluent than those that do not 13, 14). Furthermore, relying on selfreports of smoking status under a total smoking ban may be susceptible to socially desirable responses, which threaten the reliability of the variable. Using a cohort at a Japanese worksite, this study attempted to identify the effect of environmental conditions on individual smoking cessation behavior. We regarded the smoking prevalence in a work unit as the social environment affecting individual smoking cessation. With concurrent measure of individual level confounders, the cohort of a single company with designated smoking areas enables the mitigation of the selection bias due to the socioeconomic difference among workers and misclassification owing to socially desirable responses. We also tested whether workers with certain characteristics would be more susceptible to the environmental effect. Methods Data source The data were derived from annual health checkups that were conducted in a large automobile company in Japan. The number of the entire population of this worksite was 19,000 in 2005, and all workers underwent a health checkup. Each worker underwent a health checkup in his/her birth month and was followed up after a year. Our data were limited 80% of the workers (15,229/ 19,000) for whom health checkup data in a limited time span were available. Since birth months of workers are evenly distributed, we do not think this would lead to sampling bias. The worksite employed a wide variety of white- and blue-collar workers spread over 322 work units. At the time of the study, the company had not introduced a total smoking ban, however, smoking was only allowed in designated closed-off smoking rooms. In addition, individual-based and/or group education had been provided by occupational health staff for years. There was no major organizational reconstruction, and the majority of workers continued working in their initial work units throughout the study period. We further restricted our study sample (15,229 workers, 322 units) to workers who (1) continued to work in their Fig. 1. The process of sample selection. initial work units (12,921 workers), (2) underwent a follow-up checkup in 2006 and completed interview sheets (12,689 workers), (3) were aged between yr (12,437 workers), and (4) belonged to work units that had more than 10 workers (12,124 workers, 230 units), as shown in Fig. 1. We included the first and fourth conditions because our main concern was the impact of the social environment in each work unit on individual smoking behavior. Consequently, 12,124 workers (79.6% of 15,229) in 230 work units were available for the following analysis. The use of health checkup data in this study was officially approved by the company, and oral consent for generic use of data for evaluation and research was obtained from all participant employees at the time of health checkup. All data were provided anonymously. Study measures Dependent variables. The outcome variable was the change in smoking status. Respondents were classified as quitters if they reported having smoked in the first year and not having smoked in the follow-up year.

3 50 J Occup Health, Vol. 51, 2009 Independent variables. A self-administered questionnaire was distributed at the time of the health checkup to obtain smoking status (current smoker, nonsmoker), time to first cigarette after waking up (less than 30 min or later), desire to quit smoking (a lot, someday, not at all), the number of cigarettes smoked per day, and respiratory symptoms (such as cough, phlegm, and breathlessness). Data pertaining to demographics, worker type (white-collar, blue-collar), night shift duties, administrative position (whether administrative or not), and work unit codes to which the workers belonged were derived from the company s administrative records. Our primary explanatory variable of interest was the prevalence of smoking in each work unit. The prevalence was obtained as the standardized smoking ratio (SSR) by age for each work unit as of We calculated the SSR as follows: S SSR= Σ (Wi Pi) where S denotes the number of smokers in the work unit; Wi, the number of workers in age category i in the work unit; and Pi, the smoking prevalence in age category i in the entire population. Since there were very few female workers in this sample and smoking prevalence was generally low among women, we only considered male workers to calculate the SSR. We defined a work unit in which more than half of workers were white-collar as a white-collar dominant work unit (WW), and a unit in which more than half of workers were blue-collar as a blue-collar dominant work unit (BW). We obtained the SSR separately for WWs and BWs, as smoking prevalence was distinct between these two work unit types. We classified SSR categories according to quartiles of work units and classified SSR categories into levels 1 4, by arranging the SSR categories in ascending order; the SSR categories of the WWs were 0.84, , , and 1.33+, and for the BWs, they were 0.92, , , and Analysis Chi-square tests were used to examine the associations between baseline smoking prevalence and the behavioral characteristics of smokers. Univariate logistic regression analysis was performed to calculate the odds ratio of being a quitter. Multivariate logistic regression analyses were performed to control for individual level confounders. Finally, the interaction terms between the SSR and individual factors were tested. When the interaction was found to be significant, stratified analysis by the identified individual factor was conducted with the remaining variables in the model. A trend test was also performed on each logistic model in which SSR was entered as an interval variable. We chose to conduct the above analyses through stratification by work unit type because the prevalence of smokers and the association of smoking status with workers characteristics were distinctive according to work unit type, as is shown later in the results. We also conducted a similar analysis using a generalized estimation equation to reflect the hierarchical structure of the data. Since the contribution of work unit level variance was not statistically significant, we simply reported the results of logistic regression. These analyses were performed with SPSS version 15.0J, using 5% as a conventional significance level. Results Table 1 presents the baseline characteristics and smoking prevalence of respondents by work unit type. Workers in the BWs showed an overall smoking prevalence of 55.2%, compared to 31.3% in the WWs. Smoking prevalence was lower among females and higher among night shift workers. Otherwise, the prevalence proportion and distribution of characteristics were distinctive by work unit type. Older workers had the highest smoking prevalence in the WWs, while the same age groups showed the lowest in the BWs. Workers in administrative positions had lower prevalence in the BWs, although this did not hold true for the WWs. Table 2 presents the behavioral characteristics of smokers at baseline by work unit type. The majority of smokers had their first smoke less than 30 min after waking up, expressed the desire to quit someday, and smoked cigarettes per day. Smokers in the BWs had their first cigarette earlier after waking up and smoked more cigarettes per day than those in the WWs. Table 3 presents the results of univariate logistic regression analyses on quitting by work unit type. The proportion of quitters was higher among workers in the WWs than among those in the BWs. Fewer cigarettes per day, taking 30 min or more to smoke the first cigarette after waking up, and the desire to quit were common significant predictors of successful quitting in both work unit types. Younger age was also a significant predictor of successful quitting in the WWs. The point estimate of the odds ratio of quitting was the highest in the lowest category of SSR, however, the relationship did not reach conventional statistical significance (p-value for trend test =0.16 in the WWs, 0.22 in the BWs). Table 4 exhibits the results of multivariate logistic regression analyses of the SSR on quitting after controlling for gender, age, night shift duties, administrative position, number of cigarettes per day, time to first cigarette after waking up, desire to quit, and respiratory symptoms. Compared to the results of the univariate analysis, the lowest SSR persistently showed the largest odds ratio of quitting, and the estimated odds

4 Chihiro NISHIURA, et al.: The Effect of Smoking Prevalence on Cessation Behavior 51 Table 1. Baseline characteristics and smoking prevalence by work unit type White-collar dominant Work units (WWs) a Blue-collar dominant Work units (BWs) a No. Prevalence p b No. Prevalence p b Total Smoker (%) Total Smoker (%) Enrollments 3,984 1, ,140 4, Gender Male 3,503 1, < ,058 4, <0.001 Female Age , < , ,581 1, , , ,397 1, Night shift duties None 3,938 1, , <0.001 Had ,249 3, Position Not administrative 2, ,038 4, <0.001 Administrative 1, a We defined a work unit in which more than half of workers were white-collar as a white-collar dominant work unit (WW), and a unit in which more than half of workers were blue-collar as a blue-collar dominant work unit (BW). b Chi-square tests were performed by worksite types. Table 2. Behavioral characteristics of smokers at baseline by work unit type WWs a (N=1,248) BWs a (N=4,490) No. of Smokers % No. of Smokers % p b Time to first cigarette after waking up Less than 30 min , < min or later , Missing data Desire to quit A lot Someday , Not at all , Missing data Cigarettes per day < , Missing data Respiratory symptoms c None 1, , <0.001 One or more a WWs: White-collar dominant Work units; BWs: Blue-collar dominant Work units. b Chi-square tests were performed between worksite types. c Respiratory symptoms: cough, phlegm, and breathlessness.

5 52 J Occup Health, Vol. 51, 2009 Table 3. Results of univariate logistic regression analyses on quitting by work unit type WWs a (N=1,248) BWs a (N=4,490) No. of Quit Odds Ratios No. of Quit Odds Ratios Respondents Rate (%) (95%CI) Respondents Rate (%) (95%CI) Gender Male 1, (0.24, 1.14) 4, (0.10, 1.17) Female Age (1.51, 5.64) (0.87, 1.76) (1.24, 4.23) 1, (0.79, 1.52) (0.90, 3.29) (0.54, 1.23) , Night shift duties None 1, (0.27, 15.53) (0.82, 1.50) Had , Position Not administrative (0.70, 1.52) 4, (0.19, 2.06) Administrative Cigarettes per day (3.95, 24.90) (2.03, 12.27) (1.87, 8.76) (1.59, 3.93) (0.65, 3.03) 2, (0.84, 1.94) Time to first cigarette after waking up Less than 30 min (0.20, 0.44) 3, (0.43, 0.72) 30 min or later , Desire to quit A lot (1.84, 5.72) (2.11, 4.44) Someday (1.33, 3.42) 2, (1.43, 2.75) Not at all , Respiratory symptoms None 1, (0.59, 2.17) 4, (0.62, 1.70) One or more SSR b Level (0.80, 2.48) (0.89, 1.99) Level (0.71, 1.87) 1, (0.75, 1.46) Level (0.45, 1.18) 1, (0.69, 1.34) Level , a WWs: White-collar dominant Work units; BWs: Blue-collar dominant Work units. b SSR: Standardized Smoking Ratio by age; SSR was categorized in quartiles of the number of work units. All listed factors were in the first checkup. ratios were almost similar before and after the adjustment for individual factors, suggesting that the effect of the work unit SSR was somewhat independent of individual level predictors. However, it still did not reach statistical significance (p-value for trend test=0.10 in the WWs, 0.18 in the BWs). As we found in the subsequent analyses, the interaction between the SSR and desire to quit (log likelihood ratio test: p<0.001) and that between the SSR and time to first cigarette after waking up (log likelihood ratio test: p<0.001) were significant for WW subjects. We conducted stratified analysis, using these variables. As shown in Table 5, there was a significant odds ratio of quitting in the lowest quartile SSR among the subsample who smoked their first cigarette more than 30 min after waking up in the WWs (2.32; 95% CI: ). The same odds ratio was marginally significant in the BWs (1.86; 95% CI: ). The magnitude of the odds ratios and SSR suggests a linear trend in those subsamples of smokers who smoked their first cigarette more than

6 Chihiro NISHIURA, et al.: The Effect of Smoking Prevalence on Cessation Behavior 53 Table 4. Results of multivariate logistic regression analyses a on quitting by work unit type WWs b (N=1,223) BWs b (N=4,365) No. of Quit Odds Ratios No. of Quit Odds Ratios Respondents Rate (%) (95%CI) Respondents Rate (%) (95%CI) SSR c Level (0.84, 2.81) (0.88, 2.05) Level (0.73, 2.09) 1, (0.75, 1.49) Level (0.45, 1.29) 1, (0.63, 1.25) Level , a Odds ratios were adjusted for gender, age, night shift duties, position, cigarettes per day, time to first cigarette after waking up, desire to quit, and respiratory symptoms in the first checkup. b WWs: White-collar dominant Work units; BWs: Blue-collar dominant Work units. c SSR: Standardized Smoking Ratio by age; SSR was categorized in quartiles of the number of work units. Table 5. Results of multivariate analyses on quitting with interaction of smoking prevalence by work unit type WWs a (N=1,223) BWs a (N=4,365) No. of Quit Odds Ratios No. of Quit Odds Ratios Respondents Rate (%) (95%CI) Respondents Rate (%) (95%CI) Desire to quit b SSR d A lot Level (0.38, 6.71) (0.55, 3.07) Level (0.72, 8.23) (0.57, 2.23) Level (0.06, 1.18) (0.54, 2.00) Level Someday Level (0.94, 4.49) (0.67, 2.12) Level (0.63, 2.60) (0.59, 1.47) Level (0.46, 1.89) (0.45, 1.18) Level Not at all Level (0.13, 3.29) (0.77, 4.86) Level (0.17, 2.21) (0.64, 2.97) Level (0.33, 2.55) (0.51, 2.45) Level Time to first cigarette c Less than 30 min Level (0.28, 2.25) (0.64, 2.01) Level (0.37, 1.96) (0.72, 1.73) Level (0.10, 0.74) (0.60, 1.47) Level min or later Level (1.06, 5.09) (0.98, 3.55) Level (0.83, 3.39) (0.62, 1.83) Level (0.66, 2.58) (0.49, 1.49) Level a WWs: White-collar dominant Work units; BWs: Blue-collar dominant Work units. b Odds ratios were adjusted for gender, age, night shift duties, position, cigarettes per day, respiratory symptoms, and time to first cigarette after waking up in the first checkup. c Odds ratios were adjusted for gender, age, night shift duties, position, cigarettes per day, respiratory symptoms, and desire to quit in the first checkup. d SSR: Standardized Smoking Ratio by age; SSR was categorized in quartiles of the number of work units. 30 min after waking up (p-value for trend test=0.03 in the WWs, 0.10 in the BWs). Similarly, a linear and marginally significant increase in the odds ratio were also seen among those with a desire to quit someday in the WWs (2.05; 95% CI: , p-value for trend test=0.07). Discussion We found that there was a distinctive relationship between smoking prevalence in work units and the

7 54 J Occup Health, Vol. 51, 2009 individual likelihood of quitting, and the degree of this relationship varied according to individual characteristics and work unit types. These findings suggest that facets of the social environment, such as smoking prevalence, could be influential factors in inducing individual behavioral change in worksite smoking cessation programs, at least for a certain segment of workers. A previous study by van den Putte et al. also reported the effect of smoking prevalence on the individual intention of smoking cessation among 2,895 Dutch smokers through an internet survey 15). In this study, the measurement of smoking prevalence was based on a subjective report of smoking prevalence in a group, scaled on the basis of qualitative categories. To the best of our knowledge, our study is the first to report the relationship between objectively defined prevalence in work unit groups and individual change in smoking behavior. Although not conventionally statistically significant, we observed a dose-response negative trend in the odds ratios of SSR on quitting, independent of individual level behavioral factors. Smoking prevalence is presumed to reflect the degree of group norms for or against smoking. In a work unit with lower smoking prevalence, smoking would be regarded as unacceptable behavior since smoking during work hours would mean a temporal excuse from work. Social pressure against smoking would lead to a stronger motivation to quit and a higher likelihood of quitting. A previous study also revealed that social norms on the acceptability of smoking and subsequent pressure had an effect on individual intention of cessation 15), which, in turn, is a powerful predictor of quitting behavior 16, 17). The above-mentioned relationship was especially remarkable and significant among those who smoked their first cigarette more than 30 min after waking up (Table 5) and those in WWs. Time to first cigarette after waking up is an item that is included in the Fagerstrome Test for Nicotine Dependence 18) and is known to be the most powerful predictor of cessation 19) and nicotine dependence 20). Our finding may suggest that smokers with less physiological nicotine dependence are more susceptible to the surrounding social environment. Goto et al. in their internet survey of 616 Japanese smokers revealed that health risk information is more likely to induce cessation among those with less nicotine dependence 21). As such, those with less physiological addiction may be more open to external information to evaluate and reflect the relevance of smoking behavior. We also found that workers in WWs were more susceptible to smoking prevalence than those in BWs, a finding that needs some explanation. Workers in these two work units differed in their degree of job discretion. Those in BWs were less likely to exercise job discretion because of the nature of their work. Workers in BWs could only take scheduled short breaks, and the breaks were too short for them to get to designated smoking rooms to smoke. Therefore, workers in BWs smoked during lunch time and off-work hours, which tended to be outside the influence of the work unit to which they belonged to. In contrast, workers in WWs enjoyed more job discretion during duty to take a short break to smoke. As such, smoking would be regarded as something based on personal discretion in this type of work unit, and it was more likely to be blamed as being a deviation from social norms in the group. Other studies also reported that blue-collar workers are likely to experience less pressure from their coworkers to quit smoking 22), which is consistent with our findings. A linear trend between smoking prevalence and quitting was also observed among those with a moderate level of quitting desire (someday) in the WWs. According to the transtheoretical model 6), the psychological readiness to quit marks various behavioral stages. In the contemplation stage, where a subject takes a behavioral change into consideration without any action, perceived social environments and norms against smoking are influential factors affecting the individual likelihood of stage shift from contemplation to subsequent behavioral change. It would be reasonable to regard those who have the desire to quit someday as being in the stage of contemplation. Prochaska et al. reported that 40% of current smokers are in the contemplation stage; this figure is roughly consistent with that of the quit someday in our study sample 6). Our findings may bring to light an important implication for planning behavioral programs at worksites for encouraging smoking behavioral change. Those with less nicotine dependence and in the early stage of behavioral change may be affected not only by individual education but also by environmental factors. Thus, workers with the same likelihood of behavioral change with regard to their psychological readiness may still have a different likelihood of success with regard to quitting if they belong to a worksite that has a higher smoking prevalence. In such cases, education should be individually tailored and reinforced by taking the social environment into account. Besides, work units with higher smoking prevalence should be targeted with intensive interventions to change the norms concerning smoking behaviors, for example, by introducing a smoking ban 23). Limitations This study has several limitations. First, we relied on workers self-reports to obtain data on smoking behavior and nicotine dependence. Although our sample workers were not subject to a total smoking ban at the site, the social desirability response bias may have nevertheless affected the accuracy of the self-report on individual smoking behavior. However, there would be no valid

8 Chihiro NISHIURA, et al.: The Effect of Smoking Prevalence on Cessation Behavior 55 reason to believe that such a response would vary according to worksite smoking prevalence. The lesser accuracy of outcome measurement, if present, would have lead to less statistical power. Second, the work unit classifications that we used in this study might have been too large to accurately reflect the group norms against smoking, especially among the BWs. This might have attenuated the SSR association with quitting in this subsample. If we had used narrower work unit classifications, it might have been possible to obtain a clearer association. Third, the estimation of SSR might be highly variable in very small units. To avoid this issue, we excluded units with less than 10 workers from our SSR calculation. We also conducted a similar analysis with exclusion of units with less than 15 workers (11,829 workers, 206 units), and obtained similar results. However, the effect of smoking prevalence in a very small unit with less than 10 workers remains a challenging question because SSR would be more susceptible to measurement error, and individual smoking behaviors would be influenced by more intense and complicated interaction among unit members. This issue should be addressed in future studies. Most importantly, the degree of smoking prevalence and work unit type (i.e., blue/white dominance) was strongly correlated, and we were unable to determine which was more influential. We recalculated the SSR by combining the two work unit types and conducted a similar analysis. The odds ratio (95% CI) of quitting was 2.61 (1.45, 4.69), 1.93 (1.14, 3.27), and 1.27 (0.84, 1.92), in ascending order of the SSR quartiles (data not shown in the tables). As such, the likelihood of quitting among those with lower SSRs, namely those belonging to the WWs, was more strongly affected by the SSR. Further studies are needed to clarify whether the type of work unit would affect individual cessation behavior independent of work unit smoking prevalence. Conclusion Our study showed that worksite smoking prevalence affected the likelihood of successful cessation, especially among those with less nicotine dependence and who were in the early stage of behavioral change. The result suggests that serious consideration should be placed on not only individual behavior modification but also on the worksite itself to reduce smoking prevalence, as this would potentially impact approximately half the number of smokers. Acknowledgments: We would like to extend our special thanks to Toshiko Sunakoda for her invaluable advice and to Michie Kusamoto and Kiyoko Hosomoto for data collection. References 1) Edwards R. The problem of tobacco smoking. BMJ 2004; 24: ) Barnoya J, Glantz SA. Cardiovascular effects of secondhand smoke: nearly as large as smoking. Circulation 2005; 24: ) Halpern MT, Shikiar R, Rentz AM, Khan ZM. Impact of smoking status on workplace absenteeism and productivity. Tob Control 2001; 10: ) Jorenby DE, Hays JT, Rigotti NA, et al. Varenicline Phase 3 Study Group. Efficacy of varenicline, an alpha4beta2 nicotinic acetylcholine receptor partial agonist, vs placebo or sustained-release bupropion for smoking cessation: a randomized controlled trial. JAMA 2006; 5: ) Hays JT, Ebbert JO, Sood A. Efficacy and safety of varenicline for smoking cessation. Am J Med 2008; 121: S ) Prochaska JO, Velicer WF. The transtheoretical model of health behavior change. Am J Health Promot 1997; 12: ) Barth J, Critchley J, Bengel J. Psychosocial interventions for smoking cessation in patients with coronary heart disease. Cochrane Database Syst Rev 2008; 23. 8) Lancaster T, Hajek P, Stead LF, West R, Jarvis MJ. Prevention of relapse after quitting smoking: a systematic review of trials. Arch Intern Med 2006; 24: ) Sorensen G. Worksite tobacco control programs: the role of occupational health. Respir Physiol 2001; 128: ) Tanaka H, Yamato H, Tanaka T, et al. Effectiveness of a low-intensity intra-worksite intervention on smoking cessation in Japanese employees: a three-year intervention trial. J Occup Health 2006; 48: ) Fichtenberg CM, Glantz SA. Effect of smoke-free workplaces on smoking behaviour: systematic review. BMJ 2002; ) Borland R, Chapman S, Owen N, Hill D. Effects of workplace smoking bans on cigarette consumption. Am J Public Health 1990; 80: ) Moskowitz JM, Lin Z, Hudes ES. The impact of workplace smoking ordinances in California on smoking cessation. Am J Public Health 2000; 90: ) Bauer JE, Hyland A, Li Q, Steger C, Cummings KM. A longitudinal assessment of the impact of smoke-free worksite policies on tobacco use. Am J Public Health 2005; 95: ) van den Putte B, Yzer MC, Brunsting S. Social influences on smoking cessation: a comparison of the effect of six social influence variables. Prev Med 2005; 41: ) Godin G, Kok G. The theory of planned behavior: a review of its applications to health-related behaviors. Am J Health Promot 1996; 11: ) Bledsoe LK. Smoking cessation: an application of theory of planned behavior to understanding progress

9 56 J Occup Health, Vol. 51, 2009 through stages of change. Addict Behav 2006; 31: ) Heatherton TF, Kozlowski LT, Frecker RC, Fagerstrom KO. The Fagerstrom Test for Nicotine Dependence: a revision of the Fagerstrom Tolerance Questionnaire. Br J Addict 1991; 86: ) Heatherton TF, Kozlowski LT, Frecker RC, Rickert W, Robinson J. Measuring the heaviness of smoking: using self-reported time to the first cigarette of the day and number of cigarettes per day. Br J Addict 1989; 84: ) Haberstick BC, Timberlake D, Ehringer MA, Lessem JM, Hopfer CJ, Smolen A, Hewitt JK. Genes, time to first cigarette and nicotine dependence in a general population sample of young adults. Addiction 2007; 102: ) Goto R, Nishimura S, Ida T. Discrete choice experiment of smoking cessation behavior in Japan. Tobacco Control 2007; 16: ) Morris WR, Conrad KM, Marcantonio RJ, Marks BA, Ribisl KM. Do blue-collar workers perceive the worksite health climate differently than white-collar workers? Am J Health Promot 1999; 13: ) Brownson RC, Hopkins DP, Wakefield MA. Effects of smoking restrictions in the workplace. Annu Rev Public Health 2002; 23: