Advertisement
Research Article

Job Strain and Tobacco Smoking: An Individual-Participant Data Meta-Analysis of 166 130 Adults in 15 European Studies

  • Katriina Heikkilä mail,

    katriina.heikkila@ttl.fi

    Affiliation: Finnish Institute of Occupational Health, Helsinki, Finland

    X
  • Solja T. Nyberg,

    Affiliation: Finnish Institute of Occupational Health, Helsinki, Finland

    X
  • Eleonor I. Fransson,

    Affiliations: Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden, School of Health Sciences, Jönköping University, Jönköping, Sweden

    X
  • Lars Alfredsson,

    Affiliation: Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden

    X
  • Dirk De Bacquer,

    Affiliation: Department of Public Health, Ghent University, Ghent, Belgium

    X
  • Jakob B. Bjorner,

    Affiliation: National Research Centre for the Working Environment, Copenhagen, Denmark

    X
  • Sébastien Bonenfant,

    Affiliations: Inserm U1018, Centre for Research in Epidemiology and Population Health, Villejuif, France, Versailles-Saint Quentin University, Versailles, France

    X
  • Marianne Borritz,

    Affiliation: Department of Occupational Medicine, Bispebjerg University Hospital, Copenhagen, Denmark

    X
  • Hermann Burr,

    Affiliation: Centre for Maritime Health and Safety, Esbjerg, Denmark

    X
  • Els Clays,

    Affiliation: Department of Public Health, Ghent University, Ghent, Belgium

    X
  • Annalisa Casini,

    Affiliation: School of Public Health, Université Libre de Bruxelles, Brussels, Belgium

    X
  • Nico Dragano,

    Affiliation: Institute for Medical Informatics, Biometry, and Epidemiology, University Duisburg-Essen, Essen, Germany

    X
  • Raimund Erbel,

    Affiliation: Department of Cardiology, West-German Heart Center Essen, University Duisburg-Essen, Essen, Germany

    X
  • Goedele A. Geuskens,

    Affiliation: TNO, Hoofddorp, The Netherlands

    X
  • Marcel Goldberg,

    Affiliations: Inserm U1018, Centre for Research in Epidemiology and Population Health, Villejuif, France, Versailles-Saint Quentin University, Versailles, France

    X
  • Wendela E. Hooftman,

    Affiliation: TNO, Hoofddorp, The Netherlands

    X
  • Irene L. Houtman,

    Affiliation: TNO, Hoofddorp, The Netherlands

    X
  • Matti Joensuu,

    Affiliation: Finnish Institute of Occupational Health, Helsinki, Finland

    X
  • Karl-Heinz Jöckel,

    Affiliation: Institute for Medical Informatics, Biometry, and Epidemiology, University Duisburg-Essen, Essen, Germany

    X
  • France Kittel,

    Affiliation: School of Public Health, Université Libre de Bruxelles, Brussels, Belgium

    X
  • Anders Knutsson,

    Affiliation: Department of Health Sciences, Mid Sweden University, Sundsvall, Sweden

    X
  • Markku Koskenvuo,

    Affiliation: Department of Public Health, University of Helsinki, Helsinki, Finland

    X
  • Aki Koskinen,

    Affiliation: Finnish Institute of Occupational Health, Helsinki, Finland

    X
  • Anne Kouvonen,

    Affiliation: Warsaw School of Social Sciences and Humanities, Wroclaw Faculty, Wroclaw, Poland

    X
  • Constanze Leineweber,

    Affiliation: Stress Research Institute, Stockholm University, Stockholm, Sweden

    X
  • Thorsten Lunau,

    Affiliation: Institute for Medical Informatics, Biometry, and Epidemiology, University Duisburg-Essen, Essen, Germany

    X
  • Ida E. H. Madsen,

    Affiliation: National Research Centre for the Working Environment, Copenhagen, Denmark

    X
  • Linda L. Magnusson Hanson,

    Affiliation: Stress Research Institute, Stockholm University, Stockholm, Sweden

    X
  • Michael G. Marmot,

    Affiliation: Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • Martin L. Nielsen,

    Affiliation: Department of Occupational and Environmental Medicine, Bispebjerg University Hospital, Copenhagen, Denmark

    X
  • Maria Nordin,

    Affiliation: Department of Public Health and Clinical Medicine, Occupational and Environmental Medicine, Umeå University, Umeå, Sweden

    X
  • Jaana Pentti,

    Affiliation: Finnish Institute of Occupational Health, Turku, Finland

    X
  • Paula Salo,

    Affiliation: Finnish Institute of Occupational Health, Turku, Finland

    X
  • Reiner Rugulies,

    Affiliations: National Research Centre for the Working Environment, Copenhagen, Denmark, Department of Public Health and Department of Psychology, University of Copenhagen, Copenhagen, Denmark

    X
  • Andrew Steptoe,

    Affiliation: Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • Johannes Siegrist,

    Affiliation: Department of Medical Sociology, University of Düsseldorf, Düsseldorf, Germany

    X
  • Sakari Suominen,

    Affiliations: Folkhälsan Research Center, Helsinki, Finland, Department of Public Health, University of Turku, Turku, Finland

    X
  • Jussi Vahtera,

    Affiliations: Finnish Institute of Occupational Health, Turku, Finland, Department of Public Health, University of Turku, Turku, Finland, Turku University Hospital, Turku, Finland

    X
  • Marianna Virtanen,

    Affiliation: Finnish Institute of Occupational Health, Helsinki, Finland

    X
  • Ari Väänänen,

    Affiliation: Finnish Institute of Occupational Health, Helsinki, Finland

    X
  • Peter Westerholm,

    Affiliation: Occupational and Environmental Medicine, Uppsala University, Uppsala, Sweden

    X
  • Hugo Westerlund,

    Affiliations: Stress Research Institute, Stockholm University, Stockholm, Sweden, Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • Marie Zins,

    Affiliations: Inserm U1018, Centre for Research in Epidemiology and Population Health, Villejuif, France, Versailles-Saint Quentin University, Versailles, France

    X
  • Töres Theorell,

    Affiliation: Stress Research Institute, Stockholm University, Stockholm, Sweden

    X
  • Mark Hamer,

    Affiliation: Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • Jane E. Ferrie,

    Affiliations: Department of Epidemiology and Public Health, University College London, London, United Kingdom, School of Community and Social Medicine, University of Bristol, Bristol, United Kingdom

    X
  • Archana Singh-Manoux,

    Affiliations: Inserm U1018, Centre for Research in Epidemiology and Population Health, Villejuif, France, Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • G. David Batty,

    Affiliation: Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • Mika Kivimäki,

    Affiliations: Finnish Institute of Occupational Health, Helsinki, Finland, Department of Epidemiology and Public Health, University College London, London, United Kingdom

    X
  • for the IPD-Work Consortium
  • Published: July 06, 2012
  • DOI: 10.1371/journal.pone.0035463

Abstract

Background

Tobacco smoking is a major contributor to the public health burden and healthcare costs worldwide, but the determinants of smoking behaviours are poorly understood. We conducted a large individual-participant meta-analysis to examine the extent to which work-related stress, operationalised as job strain, is associated with tobacco smoking in working adults.

Methodology and Principal Findings

We analysed cross-sectional data from 15 European studies comprising 166 130 participants. Longitudinal data from six studies were used. Job strain and smoking were self-reported. Smoking was harmonised into three categories never, ex- and current. We modelled the cross-sectional associations using logistic regression and the results pooled in random effects meta-analyses. Mixed effects logistic regression was used to examine longitudinal associations. Of the 166 130 participants, 17% reported job strain, 42% were never smokers, 33% ex-smokers and 25% current smokers. In the analyses of the cross-sectional data, current smokers had higher odds of job strain than never-smokers (age, sex and socioeconomic position-adjusted odds ratio: 1.11, 95% confidence interval: 1.03, 1.18). Current smokers with job strain smoked, on average, three cigarettes per week more than current smokers without job strain. In the analyses of longitudinal data (1 to 9 years of follow-up), there was no clear evidence for longitudinal associations between job strain and taking up or quitting smoking.

Conclusions

Our findings show that smokers are slightly more likely than non-smokers to report work-related stress. In addition, smokers who reported work stress smoked, on average, slightly more cigarettes than stress-free smokers.

Introduction

Tobacco smoking is an important risk factor for chronic diseases, most notably lung cancer and other pulmonary diseases, and as such a major contributor to the public health burden and healthcare costs worldwide [1], [2], [3], [4]. Understanding the determinants of tobacco smoking is important because it would help healthcare professionals, policy makers and individuals to develop and utilise smoking cessation strategies, thus reducing the disease burdern associated with the habit. To date, however, the determinants of smoking behaviour are not well understood.

Two possible important determinants of smoking are stress in general and work-related stress in particular. However, recent observational studies of the relationship between work stress and tobacco smoking have produced mixed findings, with positive, negative and null-associations reported [5], [6], [7], [8], [9]. In many studies the associations of tobacco smoking and work stress have differed by study population and gender [7], [10], [11]. Important limitations of previous studies are that few studies thus far have been sufficiently well powered to detect small or moderate associations or to investigate whether any associations differ in socio-demographic or other subgroups of participants.

In order to collate and add to the existing evidence, we have undertaken individual-participant meta-analyses of the associations of tobacco smoking with work-related stress. We used a large set of data pooled from 15 independent European studies with a common measure of work stress, operationalised as job strain [12], [13]. We also examined the associations of job strain and smoking in socio-demographic subgroups.

Methods

Ethical Approval

Each constituent study in the IPD-Work consortium was approved by the relevant local or national ethics committees and all participants gave informed consent to take part. Details of the ethical approval are provided in Appendix S1.

Studies and Participants

We conducted individual-level meta-analyses using pooled data from 15 independent studies conducted between 1985 and 2008 in Belgium, Denmark, Finland, France, Germany, the Netherlands, Sweden and the UK. The “Individual-participant-data Meta-analysis of Working Populations” (IPD-Work) Consortium was established at the annual Four Centers Meeting in London, November 8, 2008. A pre-defined two-stage data acquisition protocol was used. The first stage involved the acquisition of baseline data on job strain as well as socio-demographic and lifestyle factors and the definition and harmonisation of these baseline characteristics across the studies. The second stage involves the acquisition of data on disease outcomes. Our meta-analyses were based on the first stage cross-sectional data and were thus conducted before any linkage to disease data. MOOSE checklist is provided in Appendix S2.

Details of the design and participants in the IPD-Work Consortium studies have been described previously [14] and are provided in Appendix S1. The following studies were included in our analyses: Belstress, Copenhagen Psychosocial Questionnaire version I (COPSOQ-I), Danish Work Environment Cohort Study (DWECS), Finnish Public Sector Study (FPS), Health and Social Support (HeSSup), Heinz Nixdorf Recall study (HNR), Intervention Project on Absence and Well-being (IPAW), Permanent Onderzoek Leefsituatie (POLS), Burnout, Motivation and Job Satisfaction study (Danish acronym PUMA), Swedish Longitudinal Occupational Survey of Health (SLOSH), Whitehall II and Work Lipids and Fibrinogen (WOLF) Norrland and Stockholm) Participants with complete data on job strain and smoking, as well as age, sex, and socioeconomic position were included in our analyses (Table 1).

thumbnail

Table 1. Summary of studies and participants.

doi:10.1371/journal.pone.0035463.t001

We estimated study-specific cross-sectional associations of tobacco smoking and psychosocial job strain in 15 studies, based on data from 166 130 individuals (mean age: 43.8 years). In addition, the associations of smoking and work stress in socio-demographic subgroups were investigated in studies in which we had access to individual-level data (n = 134 293, Figure S1). Longitudinal associations of smoking and job strain were examined using individual-level repeated measurements data from six studies, in which these data were available (n = 52 024; Figure S1).

Ascertainment of Tobacco Smoking and Work Stress

Tobacco smoking was ascertained from participant-completed questionnaires in all studies. Smokers were categorised into never, ex- and current smokers. Job strain was ascertained in all studies using questions from the Job Content Questionnaire (JCQ) and Demand-Control Questionnaire (DCQ) [12], [13]. A description of the questionnaires and job demand and job control scales is provided elsewhere [15]. Mean response score for job demands items and mean response score for job control items were calculated for each participant. Job strain was defined as having a high demand (>study-specific median score) and a low control score (<study-specific median score). All other combinations of job demands and job control, including the values equal to the median values, were assigned to the no strain-category. Participants with missing data on more than half of the job demands or job control items (n = 1 714, 1%) were excluded from the analyses.

Covariates

Information on sex and age was obtained from population registries or baseline interview (in COPSOQ-I, DWECS, FPS, Gazel, HNR, IPAW, PUMA, SLOSH, Still Working, WOLF Norrland and WOLF Stockholm) or from participant-completed questionnaires (in Belstress, HeSSup, POLS and Whitehall II). Sex was modelled as binary and age as a continuous variable (years). Socioeconomic position was defined based on the occupational title obtained from employers’ or other registries (in COPSOQ-I, DWECS, FPS, Gazel, IPAW, PUMA, and Still Working) or participant-completed questionnaires (in Belstress, HeSSup, HNR, POLS, SLOSH, Whitehall II, WOLF Norrland and WOLF Stockholm). In HeSSup socioeconomic position was defined based on the participant’s self-reported highest educational qualification. Socioeconomic position was categorised in all studies as low (e.g. cleaners, maintenance workers), intermediate (e.g. registered nurses, technicians) or high (e.g. teachers, physicians). Participants who were self-employed or who had missing data on socioeconomic position were included in the analyses in the “other” socioeconomic position category (n = 2 173, 1.3%).

Statistical Analyses

We used individual-level data provided by the investigators in Belstress, FPS, Gazel, HeSSup, HNR, Still Working, SLOSH, Whitehall II, WOLF Norrland and WOLF Stockholm studies. Investigators in COPSOQ-I, DWECS, IPAW, POLS and PUMA undertook the statistical analyses themselves according to our instructions and provided us with the study-specific results.

Individual-participant data meta-analyses can take a one-stage or two-stage approach. In the one-stage approach, individual-level data are pooled and analysed as clustered data, with study as the cluster; in the two-stage approach the effect estimates are calculated for each study separately and subsequently pooled using standard meta-analytical methods [16], [17], [18]. These approaches have been shown to provide similar results and the choice of approach depends on the research questions and available data.

Our main meta-analyses were done using the two-stage approach because we had access to individual-level data from 11 studies but only aggregate data from five studies (COPSOQ-I, DWECS, IPAW, POLS and PUMA). We used logistic regression models to estimate study-specific associations of smoking and job strain and pooled the resulting estimates and their standard errors using fixed effect and random effects meta-analyses [19]. We quantified heterogeneity in the pooled effect estimates using the I2 statistic, which indicates the proportion of the total variation in the estimates that is due to between-studies variation [20]. The one-stage approach was used to investigate exposure-covariate interactions and bidirectional longitudinal associations of smoking and job strain at baseline and follow-up, because this approach provides a flexible way of investigating individual-level interactions [18], [21], [22], [23]. Our one-stage analyses were conducted using mixed effects logistic regression with study as the random effect when the outcomes were rare, and using modified Poisson regression with robust standard errors and study as the cluster-variable when the outcomes were common [24]. Interactions were investigated by stratifying the mixed effects logistic models for sex, age group and socioeconomic position, and tested by including an interaction term (smoking*covariate) in the model that also contained the main effects.

All meta-analyses and statistical analyses in the pooled individual-level data were performed using Stata SE 11.0 (Stata Corporation, College Station, Texas, USA). In POLS the study-specific analyses were undertaken by the study team using SPSS 17 (SPSS Inc., Chicago, Illinois, USA) and in COPSOQ-I, DWECS, IPAW and PUMA using SAS 9 (SAS Institute Inc., Cary, North Carolina, USA).

Results

Tobacco Smoking and Job Strain

The characteristics of participants included in our analyses are shown in Table 1. Of the 166 130 participants, 26 415 (15.9%) reported job strain, 42 146 (25.4%) were current smokers, 54 029 (32.5%) ex-smokers and 69 955 (42.1%) never smokers.

A meta-analysis of smoking and job strain is shown in Figure 1. In the age, sex and socioeconomic position -adjusted analyses, compared to never-smokers, current smokers were, on average, 11% more likely to experience job strain (pooled random effects odds ratio (OR): 1.11, 95% confidence interval (CI): 1.03, 1.18). The odds of reporting job strain were similar among ex-smokers and never-smokers after the same adjustments (OR: 1.00, 95% CI: 0.93, 1.06). There was heterogeneity in the study-specific effect estimates comparing ex-smokers to never-smokers (I2: 69.3%) and current smokers to never-smokers (I2: 67.8%). The study-specific effect estimates are shown in Figure S2.

thumbnail

Figure 1. Association of tobacco smoking and job strain (adjusted for age, sex and socioeconomic status) (N = 166 130).

doi:10.1371/journal.pone.0035463.g001

Current smokers in our study population smoked, on average, 102 (standard deviation: 63) cigarettes in a typical week. Among current smokers, the difference in the mean number of cigarettes smoked during an average week between individuals who reported job strain and those who did not are shown in Table 2. Current smokers who reported job strain smoked, on average, three cigarettes more per week than current smokers who did not report job strain.

thumbnail

Table 2. Difference in mean number of cigarettes in an average week between current smokers with and without job strain.

doi:10.1371/journal.pone.0035463.t002

We investigated the availability of individual-level data in the IPD-Work Consortium as a possible source of heterogeneity by stratifying our meta-analyses by the availability of individual-level data (Figure S3). The pooled effect estimates from the studies in which we had no access to individual-level data were more extreme yet less precisely estimated than those from the studies which had these data available.

Stratified Analyses in Socio-demographic Subgroups

The findings from the meta-analysis stratified by demographic covariates (sex, age and socioeconomic position) and based on individual-level data are shown in Table S1. The associations between smoking and job strain were similar in men and women and any differences in these associations between age groups and socioeconomic positions were small.

Longitudinal Analyses of Smoking and Work Stress

Longitudinal analyses were based on pooled repeated measurements data from Belstress, FPS, HeSSup, SLOSH, Whitehall II and WOLF Norrland and the duration of follow-up varied between 1 and 9 years. The associations between job strain at baseline and taking up smoking by follow-up (among baseline never- and ex-smokers grouped together) are presented in Table 3. The same associations are presented separately for baseline never-and ex-smokers in Table S2. We found no clear evidence for an association between either job strain at baseline, or change in job strain status between baseline and follow-up, and taking up smoking. However, all these associations were imprecisely estimated. The associations between job strain at baseline and quitting smoking by follow-up (among baseline smokers) are shown in Table 4. Job strain at baseline was associated with neither taking up nor quitting smoking by follow-up. The associations between smoking at baseline and job strain at follow-up, stratified by baseline job strain, are shown in Table S3. There was no clear evidence for smoking at baseline being associated with changing from no strain to strain or vice versa between baseline and follow-up.

thumbnail

Table 3. Longitudinal associations between job strain and taking up smoking among baseline never- and ex-smokers (n = 42 049)1.

doi:10.1371/journal.pone.0035463.t003
thumbnail

Table 4. Longitudinal associations between job strain at baseline and quitting smoking among baseline smokers (n = 9 975)1.

doi:10.1371/journal.pone.0035463.t004

Discussion

In our pooled analyses of data drawn from 15 European studies, current smokers had approximately 11% higher odds of reporting job strain than never smokers. However, no difference in job strain was observed between ex-smokers and never-smokers. Job strain was associated with smoking dose in current smokers: those who reported job strain smoked, on average, three cigarettes more in an average week than current smokers who did not report job strain. No difference in job strain was observed between ex-smokers and never-smokers. Furthermore, we observed no clear evidence for longitudinal associations between smoking and job strain during follow-ups varying between 1 and 9 years.

Given that the excess of smokers in employees with job strain was relatively small, insufficient statistical power to detect modest associations is a potential contributing factor to inconsistencies in prior evidence [5], [6], [7], [8], [9]. The present analyses, however, were based on over three times as large a sample size as, to our knowledge, the largest previous study on this topic [25]. Our findings replicate the higher smoking intensity among stressed employees observed in that study [25] as well as the association between job strain and current smoking observed in other studies [5], [6], [7], [8], [9].

The direction of the association between job strain and smoking is not clear. Our cross-sectional findings indicate that they co-occur. We hypothesised that job strain could lead to taking up or being unable to quit smoking [10]; it is also possible that as a part of an unhealthy lifestyle smoking could lead to job strain. We investigated the direction of the relationship between smoking and job strain using repeat measures of both at baseline and follow-up in a subset of six studies (Belstress, Finnish Public Sector, HeSSup, SLOSH, WOLF Norrland and Whitehall II). However, our longitudinal analyses provided no clear evidence of a temporal association between smoking status and job strain, supporting the findings from the cross-sectional meta-analyses that smoking and work stress co-occur, but suggesting the two are unlikely to be causally related. It is, however, possible that the null findings in some of the longitudinal analyses reflect low statistical power to detect modest associations. Reporting job strain at baseline as well as follow-up was not associated with taking up smoking among baseline never- and ex-smokers grouped together (Table 3: OR: 1.03, 95% CI: 0.87, 1.21). When analysed separately, the majority of those who took up smoking during the study follow-up were baseline ex-smokers, with fewer than 1% of the never smokers starting to smoke (Table S2). We observed no association between job strain at baseline and taking up smoking by follow-up among baseline ex-smokers but, due to small numbers, particularly the analyses among baseline never smokers had insufficient power to allow reliable interpretation of the findings. The complementary analysis among baseline smokers suggests that smokers with job strain at baseline were 9% less likely to give up smoking than smokers without job strain (Table 4: OR: 0.91, 95% CI: 0.79, 1.04) - an effect size consistent with the cross-sectional analysis based on 166 000 participants.

Important strengths of our investigation were that our analyses were based on published and unpublished individual-level data from a large number of participants. Such individual-data meta-analysis is a strong study design, which reduces the possibility of publication bias that can hamper evidence from single studies and literature-based meta-analyses [26]. Work stress was defined in all studies using a widely accepted, harmonised measure, job strain [12]. The job strain measure is based on self-reported subjective assessment of attributes of the job, job demand and control and thus the possibility of reporting bias cannot be excluded. However, the effects of work-related stress assessed by self-report match well with those based on objective indicators [27]. Smoking was ascertained from participants’ self-report in all the studies and there is evidence that self-report is a fairly accurate measure of smoking behaviour [28], with demonstrated predictive validity [29]. However, due to the way questions on smoking were asked in some studies, it is possible that some ex-smokers who had given up smoking a considerable time before the study baseline have been misclassified as never smokers. Misclassification of the participant-reported measures may have lead to an under- or over-estimate of the study-specific effect estimates and, as a result, introduced heterogeneity in our meta-analyses. It is also possible that some of our findings have been influenced by residual confounding from unmeasured confounders, such as mood disorders, addiction or personality type [30], [31], [32], [33], or long working hours or overtime work [34]. Future research would help to understand the role of these factors in the association between smoking habits and job strain.

Conclusions

Our findings, based on individual-level data from 15 European studies, show that current smokers are more likely than non-smokers to report job strain. Work-related psychosocial stress was also associated with smoking intensity, with smokers who reported job strain smoking, on average, three cigarettes per week more than smokers not reporting job strain. These findings, which identify job strain as a factor that co-occurs with smoking, could be helpful in designing and implementing smoking cessation initiatives. Smokers who report job strain could be identified as being in slightly higher risk of smoking-associated disease due to their higher smoking intensity and in need of particular support in cutting down and quitting smoking.

Supporting Information

Figure S1.

Studies and participants included in the analyses.

doi:10.1371/journal.pone.0035463.s001

(DOC)

Figure S2.

Associations of tobacco smoking and job strain (adjusted for age, sex and socioeconomic position).

doi:10.1371/journal.pone.0035463.s002

(DOC)

Figure S3.

Associations of smoking and job strain, stratified by the availability of individual-level data (adjusted for age, sex and socioeconomic position).

doi:10.1371/journal.pone.0035463.s003

(DOC)

Table S1.

Associations of smoking and work stress in demographic subgroups1.

doi:10.1371/journal.pone.0035463.s004

(DOC)

Table S2.

Longitudinal associations between job strain and taking up smoking among baseline never- and ex-smokers.

doi:10.1371/journal.pone.0035463.s005

(DOC)

Table S3.

Longitudinal associations between smoking at baseline and job strain at follow-up, stratified by baseline job strain category1.

doi:10.1371/journal.pone.0035463.s006

(DOC)

Appendix S1.

Study design and recruitment of participants.

doi:10.1371/journal.pone.0035463.s007

(DOC)

Appendix S2.

MOOSE checklist.

doi:10.1371/journal.pone.0035463.s008

(DOC)

Author Contributions

Analyzed the data: KH STN MK. Contributed reagents/materials/analysis tools: KH STN EIF LA DDB JBB SB MB HB EC AC ND RE GAG MG WEH ILH MJ KHJ FK AK MK A. Koskinen A. Kouvonen CL TL IEHM LLMH MGM MLN MN JP PS RR AS JS SS JV MV AV PW HW MZ TT MH JEF ASM GDB MK. Wrote the paper: KH STN EIF LA DDB JBB SB MB HB EC AC ND RE GAG MG WEH ILH MJ KHJ FK AK MK A. Koskinen A. Kouvonen CL TL IEHM LLMH MGM MLN MN JP PS RR AS JS SS JV MV AV PW HW MZ TT MH JEF ASM GDB MK. Wrote the first draft: KH STN MK. Analyzed the data, had full access to anonymised data from all constituent studies, with the exception of data from COPSOQ-I, DWECS, IPAW, POLS and PUMA: KH STN MK.

References

  1. 1. Doll R (1986) Tobacco: an overview of health effects. IARC Sci Publ. pp. 11–22.
  2. 2. Forey BA, Thornton AJ, Lee PN (2011) Systematic review with meta-analysis of the epidemiological evidence relating smoking to COPD, chronic bronchitis and emphysema. BMC Pulm Med 11: 36.
  3. 3. (IARC) (2004) Tobacco Smoke and Involuntary Smoking, IARC Monographs on the Evaluation of Carcinogenic Risks to Humans: IARC.
  4. 4. Vineis P, Alavanja M, Buffler P, Fontham E, Franceschi S, et al. (2004) Tobacco and cancer: recent epidemiological evidence. J Natl Cancer Inst 96: 99–106.
  5. 5. Kouvonen A, Kivimaki M, Vaananen A, Heponiemi T, Elovainio M, et al. (2007) Job strain and adverse health behaviors: the Finnish Public Sector Study. J Occup Environ Med 49: 68–74.
  6. 6. Rugulies R, Scherzer T, Krause N (2008) Associations between psychological demands, decision latitude, and job strain with smoking in female hotel room cleaners in Las Vegas. Int J Behav Med 15: 34–43.
  7. 7. Lallukka T, Lahelma E, Rahkonen O, Roos E, Laaksonen E, et al. (2008) Associations of job strain and working overtime with adverse health behaviors and obesity: evidence from the Whitehall II Study, Helsinki Health Study, and the Japanese Civil Servants Study. Soc Sci Med 66: 1681–1698.
  8. 8. Perdikaris P, Kletsiou E, Gymnopoulou E, Matziou V (2010) The relationship between workplace, job stress and nurses’ tobacco use: a review of the literature. Int J Environ Res Public Health 7: 2362–2375.
  9. 9. Radi S, Ostry A, Lamontagne AD (2007) Job stress and other working conditions: Relationships with smoking behaviors in a representative sample of working Australians. Am J Ind Med 50: 584–596.
  10. 10. Kouvonen A, Vahtera J, Vaananen A, De Vogli R, Heponiemi T, et al. (2009) Relationship between job strain and smoking cessation: the Finnish Public Sector Study. Tob Control 18: 108–114.
  11. 11. Siegrist J, Rodel A (2006) Work stress and health risk behavior. Scand J Work Environ Health 32: 473–481.
  12. 12. Karasek R, Brisson C, Kawakami N, Houtman I, Bongers P, et al. (1998) The Job Content Questionnaire (JCQ): an instrument for internationally comparative assessments of psychosocial job characteristics. J Occup Health Psychol 3: 322–355.
  13. 13. Karasek R, Theorell T (1990) Healthy work: Stress, productivity, and the reconstruction of working life. New York: Basic Books.
  14. 14. Nyberg ST, Heikkila K, Fransson EI, Alfredsson L, De Bacquer D, et al. (2011) Job strain in relation to body mass index: pooled analysis of 160 000 adults from 13 cohort studies. J Intern Med.
  15. 15. Fransson EI (IPD-Work Consortium) Agreement between alternative versions of the job demand-control questionnaire in 70,800 participants in six European cohort studies. Manuscript in preparation.
  16. 16. Riley RD, Lambert PC, Abo-Zaid G (2010) Meta-analysis of individual participant data: rationale, conduct, and reporting. British Medical Journal 340: 221.
  17. 17. Simmonds MC, Higgins JP, Stewart LA, Tierney JF, Clarke MJ, et al. (2005) Meta-analysis of individual patient data from randomized trials: a review of methods used in practice. Clin Trials 2: 209–217.
  18. 18. Stewart LA, Tierney JF, Clarke M (2008) Reviews of individual patient data. In: Higgins JPT, Green S, editors. 5.0.1 [updated September 2008].
  19. 19. Egger M, Davey-Smith G, Altman DG (2003) Systematic Reviews in Health Care. Meta-analysis in Context: BMJ Books, Blackwell Publishing.
  20. 20. Higgins JP, Thompson SG, Deeks JJ, Altman DG (2003) Measuring inconsistency in meta-analyses. British Medical Journal 327: 557–560.
  21. 21. Riley RD (2010) Commentary: like it and lump it? Meta-analysis using individual participant data. Int J Epidemiol 39: 1359–1361.
  22. 22. Thompson S, Kaptoge S, White I, Wood A, Perry P, et al. (2010) Statistical methods for the time-to-event analysis of individual participant data from multiple epidemiological studies. Int J Epidemiol 39: 1345–1359.
  23. 23. Turner RM, Omar RZ, Yang M, Goldstein H, Thompson SG (2000) A multilevel model framework for meta-analysis of clinical trials with binary outcomes. Stat Med 19: 3417–3432.
  24. 24. Zou G (2004) A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol 159: 702–706.
  25. 25. Kouvonen A, Kivimaki M, Virtanen M, Pentti J, Vahtera J (2005) Work stress, smoking status, and smoking intensity: an observational study of 46,190 employees. J Epidemiol Community Health 59: 63–69.
  26. 26. Stewart LA, Parmar MK (1993) Meta-analysis of the literature or of individual patient data: is there a difference? Lancet 341: 418–422.
  27. 27. Kivimaki M, Vahtera J, Kawachi I, Ferrie JE, Oksanen T, et al. (2010) Psychosocial work environment as a risk factor for absence with a psychiatric diagnosis: an instrumental-variables analysis. Am J Epidemiol 172: 167–172.
  28. 28. Vartiainen E, Seppala T, Lillsunde P, Puska P (2002) Validation of self reported smoking by serum cotinine measurement in a community-based study. J Epidemiol Community Health 56: 167–170.
  29. 29. Salo P, Vahtera J, Hall M, Rod NJ, Virtanen M, et al. (2011) Using repeated measures of sleep disturbances to predict future diagnosis-specific work disability: A cohort study. Sleep Accepted Papers section of the SLEEP Web site.
  30. 30. Sinha R (2001) How does stress increase risk of drug abuse and relapse? Psychopharmacology (Berl) 158: 343–359.
  31. 31. Kassel JD, Stroud LR, Paronis CA (2003) Smoking, stress, and negative affect: correlation, causation, and context across stages of smoking. Psychol Bull 129: 270–304.
  32. 32. Paterniti S, Niedhammer I, Lang T, Consoli SM (2002) Psychosocial factors at work, personality traits and depressive symptoms. Longitudinal results from the GAZEL Study. Br J Psychiatry 181: 111–117.
  33. 33. McCrae RR (1990) Controlling neuroticism in the measurement of stress. Stress Medicine 6: 237–241.
  34. 34. Virtanen M, Ferrie JE, Singh-Manoux A, Shipley MJ, Vahtera J, et al. (2010) Overtime work and incident coronary heart disease: the Whitehall II prospective cohort study. Eur Heart J 31: 1737–1744.