Objective Many studies demonstrated a decline in hospital admissions for cardiovascular diseases after the implementation of a smoking ban, but evidence for reductions in cardiovascular mortality is more limited. In Belgium, smoke-free legislation was implemented in different phases. Public places and most workplaces became smoke-free in January 2006, whereas the legislative ban on smoking in restaurants was introduced in January 2007. These successive steps in legislation provided us the opportunity to investigate possible stepwise changes in fatal acute myocardial infarction (AMI) rates.
Methods Data on all AMI deaths of 30 years of age or older in Flanders (Belgium) between 2000 and 2009 (n=38 992) were used. Age-standardised AMI death rates were analysed with segmented Poisson regression allowing for secular trends, seasonality, temperature, PM10 and influenza.
Results An immediate decrease in AMI mortality rates was observed in January 2006 (smoking ban at work). The effect was highest for women younger than 60 years of age (−33.8%; 95% CI −49.6 to −13.0), compared with an effect of −13.1% (95% CI −24.3 to −0.3) for male counterparts. Estimates for the elderly (≥60 years) were −9.0% (95% CI −14.1 to −3.7) for men and 7.9% (95% CI −13.5 to −2.0) for women. An additional effect of the smoking ban in restaurants was observed for elderly men, with an annual slope change of −3.8% (95% CI −6.5 to −1.0) after 1 January 2007.
Conclusions Smoking ban interventions are associated with reductions in the population rate of myocardial mortality, with public health gains even before and during the middle-aged period of life.
- Myocardial Ischaemia and Infarction (IHD)
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The adverse health effects of active and passive smoking are well established, in particular for coronary heart diseases.1 Growing evidence suggests that environmental tobacco exposure (ETS) increases cardiac risk through chronic pathways such as atherosclerosis, and acute mechanisms such as platelet activation and endothelial dysfunction.1 ,2 The effects of passive smoking have been found to be nearly as large (averaging 80–90%) as those of chronic active smoking.1 In order to reduce exposure to second-hand smoke, many countries have implemented regulations that prohibit smoking in public places.
Many studies have shown that the enforcement of smoking bans is followed by a rapid reduction in cardiovascular disease rates in the general population, with most of the evidence coming from hospital admission data.3–6 Most of these studies used data from hospital-based registries, whereas studies analysing population-based registries are limited.7 ,8 Moreover, only a few studies have examined the effect of smoke-free legislation on cardiovascular mortality and results are contradictory.9–14 Population-based death registry data have greater population and geographical coverage, and capture the large proportion of fatal events that occur before reaching the hospital.15
The magnitude of the effects of smoke-free legislation on cardiovascular disease rates is still uncertain, with estimates ranging from approximately 0 up to 70%.2–6 ,8 Methodological issues might play an important role in the observed differences in effect estimates. The US Institute of Medicine highlighted the importance of the adjustment for secular trend,2 and Gasparrini et al8 and Barr et al16 demonstrated the need to consider non-linear trends.
In Belgium, smoke-free legislation was implemented in different phases (http://www.tegenkanker.be/content/de-wet).17 The first phase, implemented in January 2006, required all public places and workplaces (except for bars, cafes, restaurants, night clubs and discos) to be smoke-free. The legislative ban on smoking in restaurants was introduced in January 2007, while for bars serving food, smoke-free legislation was implemented in January 2010. Finally, a comprehensive ban (including bars, discos and casinos) was introduced in July 2011.
This study investigated the impact of the first two phases of the legislation on mortality from acute myocardial infarction (AMI) in Flanders, using cause of death data from 2000 to 2009. The two later legislation phases could not yet be examined because data were not yet available. To our knowledge, only one previous study has investigated cardiovascular effects of a smoking ban that was implemented in separate phases, thereby focusing on sudden circulatory arrest (SCA).18 With 6 years of data before the first smoking ban, and 3 years of data after the second ban, the prelegislative and postlegislative periods used in this analysis are relatively long compared to most other studies.3–6 The robustness of the study findings to the model specification was evaluated by including non-linear secular trends.
The Flemish Agency for Care and Health registers data on causes of mortality in Flanders, currently coded according to the 10th revision of the International Classification of Diseases (ICD-10). Flanders is the Dutch-speaking northern part of Belgium with about six million inhabitants. This study used data on all AMI deaths (ICD-10 code I21) among Flemish residents of 30 years of age and older during the period 2000–2009 (later data were not yet available). Data were aggregated by month, gender, 10-year age groups (30–39, 40–49, 50–59, 60–69, 70–79, 80+ years), and primary (underlying) or secondary diagnosis. Flemish population figures by gender, age and calendar year were obtained from the Standardized Procedures for Mortality Analysis (SPMA) website (http://www.wiv-isp.be/epidemio/spma), which presents data from the Federal Public Service Economy–Statistics Division. Mean daily air temperature, measured at the central and representative station of Uccle (Brussels, Belgium), was provided by the Belgian Royal Meteorological Institute. Mean daily particulate matter (PM10) concentrations (spatial average for Flanders) were obtained from the Belgian Interregional Environment Agency (IRCEL). Weekly influenza rates for Belgium were provided by the National Influenza Centre.19 This centre collects information on circulating influenza viruses, whereas the Unit of Health Services Research of the Institute of Public Health collects data on consultation rates for influenza-like illnesses from a representative network of general physicians.
Annual age-standardised rates of AMI mortality were calculated by the direct method and using the European reference population.20 To compare the periods before and after the smoking ban, monthly AMI mortality rates were analysed by using an interrupted (segmented) Poisson regression, adjusting for seasonality and long-term trends. To account for changes in the age distribution during the study period, monthly age-standardised and sex-standardised incidence rates were calculated, using the population distribution in the first month of the series (January 2000) as reference. In order to model the rates of AMI directly while keeping the actual number of AMI deaths as response variable, adjusted population sizes were calculated and entered in the model as an offset. To account for the difference in the number of days between months, the logarithm of the number of days was included as a predictor in the model. The model allowed for an underlying trend throughout the study period, and seasonality was modelled by a Fourier series of sine and cosine terms.21 Because an initial exploration of the form of the long-term trend suggested a linear pattern, a linear predictor for time was used to quantify the underlying downward trend in AMI mortality rates. Mean monthly values for temperature, PM10 and influenza rates were included in the model as natural cubic splines. The degrees of freedom (4 for temperature and influenza, 3 for PM10) were chosen based on the Akaike and Bayesian Information Criteria (AIC and BIC).
The immediate effect of smoke-free legislation was modelled as a step function (level change), including a binary indicator variable which takes a value of 1 when the ban is present and 0 otherwise, while the gradual effects were studied with an interaction term between the indicator variable and time (trend or slope change). We started by examining the effect of the two legislation phases by using separate models for each phase, including the step change and the slope change into the model. Subsequently, we entered the two phases in the same model: in a first model we included only the step changes of the two phases, and in a second model we included only the slope changes. Finally, we started with a full model including the two step changes and the two slope changes and we removed non-significant factors. The performance of the above models was checked by comparing AIC and BIC values.
Because several studies reported different effects among men and women and among younger and older persons,7 ,9 ,10 ,14 ,22–25 analyses were stratified by gender and age below 60 or 60 years or older. In a secondary analysis, we used smaller age classes (30–49, 50–59, 60–69, 70–79, at least 80 years) to further investigate potential effect modification by age. Only cases with AMI as primary death cause were considered in the main analysis. In a sensitivity analysis, also deaths with AMI as secondary cause were included. In another sensitivity analysis, the assumed linearity of the underlying trend was formally tested. Because non-linear models could be attributing some of the effect of the ban to the general secular trend, potential non-linearity was investigated on prelegislation data by testing the statistical significance of a quadratic trend orthogonal to the linear trend.16 Additionally, we fitted the prelegislation trend as a natural cubic spline with 1–50 degrees of freedom and compared the fit of models using AIC.
To control for multiple testing in the final models, we used the Benjamini–Hochberg procedure.26 A false discovery rate smaller than 0.05 was considered significant. All the analyses were performed by using SAS V.9.2 (SAS Institute, Cary, North Carolina, USA).
Table 1 presents the number of AMI deaths and the age-standardised death rates by calendar year, gender and age group. A gradual reduction in AMI mortality rates was evident over the total period studied, but in all four strata, the largest (absolute) decrease in rates was observed in 2006. Whereas the CIs of the rates mostly overlap, there is no overlap in the CIs of the rates in 2005 and 2006 for men and women aged 60 years or older.
Final models were obtained by removing non-significant step and slope changes from the full model. These models also had the lowest BIC values, whereas AIC showed the tendency to select models containing non-significant smoking ban effects (p>0.13) (see online supplementary table S1). The step change in January 2006 was retained in the final models of all subpopulations. For men aged at least 60 years, the final model also included a significant slope change in January 2007. Results of the final models are reported in table 2.
The negative linear baseline trend (before 2006) in AMI mortality rates was significant in all strata except in women younger than 60 years of age. Among people below 60 years of age, the immediate decrease in the AMI death rate in January 2006 was higher for women (−33.8%; 95% CI −49.6 to −13.0) than for men (−13.1%; 95% CI −24.3 to −0.3). Estimates for elderly were lower, and were of the same magnitude for men (−9.0%; 95% CI −14.1 to −3.7) and women (−7.9%; 95% CI −13.5 to −2.0). Elderly men showed an additional annual slope change of −3.8% (95% CI −6.5 to −1.0) after the second smoking ban. Observed AMI rates and the estimated temporal trend are depicted in figure 1. Overall, from January 2006 to December 2009, the model predicts 1715 fewer AMI deaths in Flanders associated with the implementation of smoke-free legislation.
The analysis on more narrow age categories (figure 2) suggests that the immediate effect of the workplace smoking ban is highest for the youngest group (30–49 years), but estimates show an increase with age going from the middle (60–69 years) to the highest age category (80+ years). The sensitivity analysis showed that the inclusion of cases with AMI as secondary death cause produced similar results with slightly lower effect estimates. For men below 60 years of age, the step change in 2006 was no longer significant (p=0.12). In the analyses testing the potential non-linear curvature of the secular trend, no significant quadratic terms were found (p>0.21), and models with a linear prelegislation trend showed the best fit.
This study, based on nearly 40 000 AMI deaths in Flanders, showed significant reductions in AMI death rates after the introduction of national-level smoke-free legislation. Significance of estimates remained after Benjamini–Hochberg correction for multiple testing. The effects were more pronounced after the first step of the legislation (smoking ban in workplaces and public places). Our findings are compatible with a study on out-of-hospital SCA in South Limburg (a region in the Netherlands). Based on 2305 cases, the authors found a significant decrease in the incidence of SCA after the workplace smoking ban in 2004, but no further decrease after the extension of the ban to the hospitality sector in 2008.18 Whereas they found a gradual decrease in SCA incidence (annual reduction of 6.8%) after the workplace smoking ban, we observed an immediate reduction (step change). Additionally, we found some evidence for a gradual effect after the smoking ban in restaurants, but only for elderly men. However, the short time period (12 months) between the two phases of the legislation makes it difficult to separate their impact due to potential postponed effects of the first phase.
Although two reviews suggest that impacts of smoke-free legislation could increase over time,3 ,5 some other studies13 ,22 ,24 did not observe significant gradual effects additional to the immediate effects. Strong evidence that smoking ban effects are immediate rather than gradual is provided by the most recent meta-analysis of 45 studies that demonstrated that smoking ban effects on cardiovascular, cerebrovascular and respiratory diseases did not increase with a longer follow-up period.6
The magnitude of the overall reduction in AMI mortality rates associated with the 2006 legislation found in this study (−15.3% among people aged under 60 years and −8.3% among people aged 60 years or older) is comparable with the 10–17% decrease in AMI hospital admissions rate estimated by four meta-analyses.3–6 The estimated effect for women aged younger than 60 years in this study, however, is considerably higher (33.8%). A greater effect in younger individuals was also found in three Italian studies7 ,22 ,25 and is plausible because banning smoking at workplaces mainly affects the working population and appears to encourage smoking cessation particularly in younger smokers.27 The age of 60 years used in the analysis is expected to be the most realistic cut-off value to separate the working from the non-working population. Although the official retirement age in Belgium is 65 years, the median age at which people withdraw from the labour force is around 7 years below the official age.28 Studies in Scotland, England and Spain found greater effects among the elderly (above 60 or 65 years).9 ,14 ,23 ,24 Although we found much larger smoking ban effects among the youngest ages, our secondary analysis showed a higher reduction in AMI mortality rates in the oldest age categories compared with the group of 60–69 years. The elderly are generally considered to be a susceptible segment of the population. Their higher prevalence of cardiorespiratory conditions, together with age-related declines in physiological reserves or homeostatic balances, makes them more vulnerable to the effects of triggers such as air pollution and passive smoking.29
Our study suggests that the reduction in AMI mortality rates is greater among women than among men, which is compatible with findings from other studies.9 ,10 ,23 This might be explained by a stronger decrease in exposure among women, as suggested by the greater postban reduction in serum cotinine levels in non-smoking women (47%) compared with men (37%) in Scotland.23 Another reason might be the higher relative risk for AMI associated with smoking in women compared with men.5 Tobacco smoke may have an antiestrogenic effect, particularly in young, premenopausal women who would otherwise benefit from oestrogen's cardioprotective role.30 Nevertheless, the Rome, England, and Italian 4 regions studies found greater effects among men.7 ,24 ,25
Studies to date have been heterogeneous in their designs, target populations, statistical analyses, choices of control groups, and types of smoking bans investigated.16 Misspecification of the underlying long-term trend and other issues, such as changes in population size, sampling variations, different lengths of follow-up, changes in active smoking and differences in the prevalence of active and passive smoking have been raised to explain the differences in the effect estimates between studies.3 ,5 ,8 ,24 Gasparrini et al8 and Barr et al16 argue that the beneficial effect of smoking bans may be overestimated because non-linearity of trends in declining cardiovascular morbidity is not adequately taken into account in many studies. Therefore, the potential non-linear curvature of the underlying trend was formally tested in a sensitivity analysis. The validity of the linear models used in the main analysis was confirmed.
Can the reduction in AMI rates be considered a causal consequence of the smoking ban? A population intervention must be interpreted at the population level and not at the individual level. As this was an ecological study, it is possible that unmeasured confounders were responsible for the observed effects. Nevertheless, it is hard to conceive of a factor that could change the population risk of AMI mortality at the moment of the introduction of smoke-free legislation, and it is unlikely that the study findings could be explained by abrupt changes in therapeutic strategies coinciding with the smoking bans. Moreover, the plausibility of the favourable effects of smoking bans rests on well-known effects of active and passive smoking based on animal and human studies.31 Indeed, tobacco smoke consists of large amounts of particulate matter. Lung32 and systemic inflammation, prothrombotic reactions,33 modulations in heart rate variability34 and blood pressure,35 are mechanisms which explain the association between AMI and particulate air pollution29 at concentrations which are much lower than exposure to second-hand smoke.
Limitations of this study include the absence of a comparison with a nearby control population without smoking ban and the lack of data on individual smoking status neither active nor passive. The observed effects may include reduced exposure to ETS36 as well as direct health benefits from reduced tobacco consumption. Indeed, smoke-free legislation may stimulate smokers to establish total smoking prohibition in their homes37 and has resulted in increased smoking cessation.27 The Belgian Health Interview Survey suggests that the total population prevalence of active smoking in Belgium was relatively stable from 1997 to 2004, but decreased significantly from 2004 to 2008 (http://www.wiv-isp.be/epidemio/hisia/index.htm). More specifically, the percentage of female smokers in Flanders was close to 22% in 1997, 2001 and 2004, while in 2008 the percentage was only 17.9%. The prevalence of daily smoking among Flemish females reduced from 18.5% in 2004 to 15.3% in 2008, and the prevalence of heavy smoking (20 cigarettes or more per day) decreased from 7.7% to 4.9%.
Strengths of this study include the opportunity to investigate a stepwise implementation of smoke-free legislation, the large population size and the relatively long prelegislation and postlegislation periods. Several earlier studies used a ‘before and after’ design, and might have been unable to fully account for underlying trends in AMI rates. Declining time trends, as observed in this study, might be caused by the concomitant effect of other time-varying factors, like changes in the distribution of known risk factors, healthcare improvements, and development of diagnostic criteria.
This study found population changes in fatal AMI rates after the introduction of smoke-free legislation in Belgium. Given that coronary heart disease is the single most common cause of death and morbidity worldwide, even a small reduction in the incidence of AMI is of great benefit for public health. Our study indicates that 1715 AMI deaths were likely prevented in Flanders from January 2006 to December 2009. These findings add to the evidence that smoking bans improve population health and decrease health care costs and, knowing that only 16% of the world's population is currently covered by comprehensive smoke-free laws,38 provide further support for the introduction of smoke-free legislation worldwide.
What is known on this subject?
Despite the large number of studies showing decreases in hospital admissions for cardiovascular diseases after the implementation of a smoking ban, evidence for reductions in fatal acute myocardial infarction (AMI) is less abundant.
What might this study add?
The stepwise implementation of smoke-free legislation in Belgium allowed the investigation of potential successive reductions in AMI mortality rates. Significant immediate decreases in rates of fatal AMI were observed after the first step of the legislation (smoking ban in workplaces and public places). For men aged at least 60 years, an additional effect of the smoking ban in restaurants was found.
How might this impact on clinical practice?
Given that AMI is the leading cause of death and a major economic burden in most countries, our findings have important public health implications, and provide further support for the introduction and continuation of smoke-free laws worldwide.
The authors thank the Flemish Agency for Care and Health for providing death registry data.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors TSN and BC conceived and designed the study and secured funding. BC did the statistical analysis and wrote the first draft of the manuscript. All authors contributed to the discussion and interpretation of the data, the writing of the article and approved the final version of the manuscript.
Funding This study was commissioned and financed by the Vlaamse Liga tegen Kanker (VLK), Hasselt University Fund (BOF) and European Research Council (ENVIRONAGE, ERC-2012-StG 310898).
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.