Abstract

Summary of findings

for the main comparison

Patient or population: Smokers and nonsmokers Settings: 21 countries including 12 European countries, Turkey, USA, Canada, Australia, New Zealand, Hong Kong, Argentina, Panama, Uruguay. Intervention: Comprehensive or partial smoking bans in public places implemented by legislation Comparison: No bans (note: observational data only)
Outcomes1	Effects of intervention	Quality of the evidence (GRADE)2	Comments
Cardiovascular health	44 studies included. 43 studies evaluated incidence of acute myocardial infarction (AMI) and acute coronary syndrome (ACS), 33 of which detected significant associations between introduction of bans and reductions in events. 6 studies evaluated stroke incidence; 5 detected significant associations between introduction of bans and reductions in events	moderate3
Respiratory health	21 studies included. Data imprecise with conflicting results. 6 of 11 studies reported significant reductions in COPD admissions. 7 of 12 reported significant reductions in asthma admissions	very low4
Perinatal health	7 studies included. Data imprecise with conflicting results; due to study designs unclear if many of observed associations due to confounding factors	very low4
Mortality	11 studies included. 8 detected significant association between introduction of bans and reduced smoking‐related mortality	low
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

Open in a separate window

¹Note, original review also included changes in environmental tobacco smoke (ETS) exposure as an outcome. Evidence was unequivocal that bans were associated with significant reductions in ETS (see Callinan 2010), and hence we did not evaluate this outcome in this update.

²As all studies are observational, starting point for GRADE rating is low. Meta‐analyses not conducted; data summarized narratively.

³Upgraded due to evidence of a dose‐response effect.

⁴Downgraded due to imprecision.

Background

Objectives

To assess the effects of legislative smoking bans on (1) morbidity and mortality from exposure to secondhand smoke, and (2) smoking prevalence and tobacco consumption.

Methods

Results

Description of studies

See: Characteristics of included studies, Characteristics of excluded studies, Figure 1.

Open in a separate window

1

• Study flow diagram.

We include 77 studies which met the eligibility criteria for this updated review. We retain 12 studies with unchanged data from the first version of the review (Cesaroni 2008; Gallus 2007; Goodman 2007; Hahn 2008; Juster 2007; Khuder 2007; Larsson 2008; Lemstra 2008; Pell 2008; Pell 2009; Sargent 2012; Seo 2007). Additional results have been reported for two previously included studies, and we have renamed them to reflect this, with original reports now listed as secondary references: Alsever 2009 (previously Bartecchi 2006), and Barone‐Adesi 2011 (previously Barone‐Adesi 2006). We have now excluded other studies previously included that reported passive smoke exposure with either self‐reported outcomes or cotinine measures. Other excluded studies are those without a six‐month follow‐up period following the ban and those that did not report smoking prevalence from national population data.

The included studies examine the effects of comprehensive or partial indoor smoke‐free legislation implemented in countries, states (regions) or at local level. We identified the effect of the implementation of national smoking bans in studies representing 21 countries. Studies with national smoking bans in countries included in this update are: Argentina (Ferrante 2012), Belgium (Cox 2013; Cox 2014), Denmark (Christensen 2014), Germany (Sargent 2012; Schmucker 2014), Hong Kong (McGhee 2014), Panama (Jan 2014), Switzerland (Bonetti 2011; Di Valentino 2015; Durham 2011; Dusemund 2015; Humair 2014; Rajkumar 2014), Turkey (Yildiz 2015) and Uruguay (Sebrié 2014).

Countries included in the earlier review are retained in the update: Canada (Gaudreau 2013; Lemstra 2008; Naiman 2010), England (Lee 2011; Liu 2013; Millett 2013; Sims 2013), France (Séguret 2014), Ireland (Cronin 2012; Goodman 2007; Kabir 2009; Kabir 2013; Kent 2012; Stallings‐Smith 2013), Italy (Barone‐Adesi 2011; Cesaroni 2008; Federico 2012; Gallus 2007; Gasparrini 2009; Gualano 2014), Netherlands (De Korte‐De Boer 2012), New Zealand (Barnett 2009), Norway (Bharadwaj 2012), Scotland (Jones 2015; Mackay 2010; Mackay 2011; Mackay 2012; Mackay 2013; Pell 2008; Pell 2009), Spain (Aguero 2013; Villalbi 2011), Sweden (Larsson 2008) and USA (Alsever 2009; Amaral 2009; Barr 2012; Basel 2014; Bruckman 2011; Bruintjes 2011; Croghan 2015; Dove 2010; Hahn 2008; Hahn 2011; Hahn 2014; Head 2012; Herman 2011; Hurt 2012; Juster 2007; Khuder 2007; Klein 2014; Landers 2014; Lippert 2012; Loomis 2012; North Carolina 2011; Page 2012; Roberts 2012; Rodu 2012; Sargent 2004; Seo 2007; Vander Weg 2012).

One study reports on the impact of national smoking bans from a number of countries including the USA, Canada, New Zealand, Scotland, Republic of Ireland, and Northern Ireland (Bajoga 2011). The majority of studies (27) are located in the USA. Other countries with multiple studies are: Scotland (7), Ireland (6), Switzerland (6), Italy (6) and England (4).

The definition used in this review for comprehensive smoking bans is prohibited smoking in work places, including restaurants and bars. We categorise legislation which permits smoking in bars and restaurants as a partial smoking ban, whether at local, state or national level. The implementation of smoking bans has varied across national jurisdictions, and exceptions for smoking rooms may be allowed within comprehensive bans. Using these definitions, we identified 18 studies reporting evidence for partial smoking bans (Aguero 2013; Amaral 2009; Bonetti 2011; Christensen 2014; Cox 2014; Di Valentino 2015; Dusemund 2015; Durham 2011; Humair 2014; Khuder 2007; Lippert 2012; Loomis 2012; McGhee 2014; Rajkumar 2014; Sargent 2004; Sargent 2012; Schmucker 2014; Villalbi 2011). We define the majority of smoking bans in place as comprehensive within this review.

The settings in this update vary considerably from the original review. For this update we identified studies reporting the impact of national smoking bans in the following settings:

• 42 studies used hospital registers for admissions or discharge data on specific population cohorts

• 20 studies used registries for national health outcomes, death rates, pregnancy and perinatal health

• 11 studies used population‐level country‐specific prevalence surveys reporting active exposure to smoking

• 4 studies are work place‐based, reporting primarily passive exposure and measured health outcomes.

We found 43 studies which reported smoking data either as a primary outcome, a descriptive variable reporting national prevalence without comparing rates before or after smoking legislation, or used as a covariate in analysis. Eleven studies (Cesaroni 2008; Christensen 2014; Cox 2014; Ferrante 2012; Head 2012; Hurt 2012; Jan 2014; Kabir 2013; Mackay 2010; Naiman 2010; Stallings‐Smith 2013) report smoking prevalence data from another data source, rather than data from their own studies. Twenty‐four studies report an impact of smoking bans on active or passive smoking (Analysis 1.1). Active smoking outcomes including prevalence, quit rate and tobacco consumption are specifically reported in 19 studies (Bajoga 2011; Bharadwaj 2012; Cesaroni 2008; Cox 2014; Federico 2012; Ferrante 2012; Gallus 2007; Gualano 2014; Hahn 2008; Hurt 2012; Jones 2015; Kabir 2009; Klein 2014; Lee 2011; Lemstra 2008; Lippert 2012; Mackay 2011; Mackay 2012; Page 2012). Combined active and passive smoking outcomes are reported in Larsson 2008. Passive smoke exposures are reported in a further four studies (Durham 2011; Goodman 2007; Pell 2008; Rajkumar 2014) with the evidence of health outcomes reported in 72 studies, including: cardiovascular outcomes (Analysis 1.1), respiratory outcomes (Analysis 2.1) and perinatal health outcomes (Analysis 3.1). Associations between indoor smoking legislation and mortality rates are reported in 11 studies included in this update (Analysis 4.1). A number of studies report multiple health outcomes or a combination of health‐related outcomes and mortality outcome data.

1.1

Analysis

Comparison 1 Cardiovascular health outcomes, Outcome 1 Cardiac outcomes.

Cardiac outcomes
Study	Location/ Intervention	Outcomes	Smoking status
ITS studies
Aguero 2013	Spain, Girona Partial 2006	All AMI events 1 January 2002 to 31 December 2008 for people aged 35 to 74 years: 3703 cases. 2142 events pre‐legislation. 3012 were admitted to hospital AMI incidence rates significantly decreased (RR 0.89, 95% CI 0.81 to 0.97); similar significant decreases observed in mortality rates, RR 0.82 (95% CI 0.71 to 0.94). Decrease observed in both genders, particularly women (RR 0.72) and in people 65 to 74 years (RR 0.74) Nonsmokers showed diminished incidence rates; passive smokers significant reductions in AMI RR 0.88, (95% CI 0.80 to 0.97) (AHA definition); RR 0.82, (95% CI 0.72 to 0.92) (WHO MONICA definition). Non‐significant in smokers; RR 0.93 (95% CI 0.82 to 1.05) (AHA definition); RR 0.91, (95% CI 0.80 to 1.04) (MONICA definition)	Smoking status reported No validation
Barnett 2009	New Zealand, Christchurch Comprehensive 2004	Poisson regression analysis pre‐ and post‐ban. Deprivation coding for socioeconomic profile. Overall RR was 0.92 (95% CI 0.86 to 0.99) between first AMI admissions pre‐ and post‐smoke‐free legislation Gender stratification identified a significant reduction for men RR 0.90 (95% CI 0.82 to 0.99) when compared to women RR 0.94 (95% CI 0.84 to 1.05) Age stratification identified significant reductions for men in admissions for first AMI event in 55 to 74 year olds RR 0.86 (95% CI 0.75 to 0.99) and 75+ age group RR 0.85 (95% CI 0.73 to 0.98) Highest RR differences in admissions were recorded for nonsmokers (aged 30 to 54 years) following smoking legislation: RR 1.71 (95% CI 1.16 to 2.52) Significant differences noted for nonsmokers in 55 to 74 year age group, compared to regular and ex‐smokers RR 0.83 (95% CI 0.69 to 1.00). Significant reductions in admissions in those aged 55 to 74 years living in quintile 2, RR 0.76 (95% CI 0.59 to 0.97) No significant differences observed for smokers	Smoking status reported No validation
Barone‐Adesi 2011	Italy 20 Italian regions Comprehensive 2005	Poisson regression analysis pre‐ and post‐ban. Mixed effects regression modelling used with fixed coefficients for national trend reporting; random coefficients reported for region‐specific deviations Overall rate ratio (RR) 0.96 (95% CI 0.95 to 0.98) for ACE admissions among people aged 70 years and younger. This was a 4% reduction in hospital admissions post‐smoke‐free legislation Men RR 0.97 (95% CI 0.95 to 0.98) Women RR 0.95 (95% CI 0.93 to 0.98) There was no effect in people aged over 70 years; RR 1.00 (95% CI 0.99 to 1.02)	No smoking status reported
Barr 2012	USA, 9 States: Illinois, Ohio, Minnesota, New York, Washington, New Jersey, Arizona, Massachusetts, Delaware. Comprehensive	Poisson regression modelling used. Adjustment for demographic and seasonal and secular trends in admission rates. State level modelling with county‐specific random effects used to estimate change in AMI admission rates Approx. 64,000 admissions for AMI per year. Statistically significant results in AMI hospital admissions post‐ban were found when strict linearity of secular trends of AMI admission rates was assumed: ‐5.4% (95% CI ‐8.2 to ‐2.5) The effect was attenuated to zero under relaxation of assumptions No significant results identified following non‐linear adjustments for secular trends.	No smoking status reported
Basel 2014	USA, Colorado Comprehensive 2006	Poisson regression analysis used to identify differences in monthly AMI admissions post legislation 63.9% of patients were men and 36.1% were women. Mean age 66.9 years SD ± 14.4 No significant reduction in AMI rates observed post‐legislation risk ratio (RR) 1.059 (95% CI 0.993 to 1.131) Results identified a steep decline in AMI rates 2000 to 2005 prior to legislation. Two smaller communities in Colorado previously enacted smoke‐free legislation and identified 27% reduction in AMI hospitalizations (Bruintjes 2011) Current study adjusted for this population of 5411 patients and adjusted population census. No significant difference post‐legislation adjusting for this group, RR 1.038 (95% CI 0.971 to 1.11) No significant impact of smoke‐free legislation demonstrated even after accounting for pre‐existing ordinances	No smoking status reported
Bruckman 2011	USA, Ohio Comprehensive 2007	Interrupted monthly time series study. Mixed linear modelling data adjusting for gender and age AMI rate reduced 1.9775 per 1000 in 2005 to 1.680 per 1000 in 2009 (1680 discharges per one million Ohio residents) For men and women the mean age adjusted discharge rate decreased over study period P < 0.0001. (men: 2.6334 vs 2.2567, P < 0.001; women: 1.432 vs 1.992, P < 0.001) Significant decrease in discharge rates before and after statewide indoor tobacco smoke ban	No smoking status reported
Christensen 2014	Denmark Partial ban (not fully enforced) 2007	Smoking prevalence decreased from 27% in 2003 to 21% in 2010 (National survey data) 109,094 admissions recorded during study period. Adjusted modelling for age, gender and type 2 diabetes No significant differences in hospital admissions for AMI identified post‐ban after adjusting for age and gender Significant differences in hospital admissions for AMI identified after adjusting for age, gender and incidence of type 2 diabetes: 1 year pre‐ban RR 0.86 (95% CI 0.79 to 0.94) 1 year post‐ban RR 0.77 (95% CI 0.71 to 0.85) 2 years post‐ban RR 0.77 (95% CI 0.70 to 0.84) Significant reduction in number of AMI admissions may be explained by incremental enactment of smoking ban activities in Denmark and implementation of nationwide ban on trans‐fatty acids in food in 2004	Smoking status not reported from AMI data Smoking prevalence reported from national surveys
Cronin 2012	Ireland Comprehensive 2004	At baseline, percentage of current smokers admitted with ACS 2003/2004 was 34%. This reduced in 2005/2006 to 31% and reduced further in 2006/2007 to 29% Pre‐legislation 205.9 ACS admissions/100,000 population. In the year following ban there was a statistically significant 12% reduction in the rate of admissions 177.9/100,000 (95% CI 164.0 to 185.1, P = 0.002) There was no change in the rate of ACS admissions in the following year. A further 13% reduction was observed in the 3rd year post‐legislation March 2006 to March 2007; 149.2 (95% CI 139.7 to 159.2) Reductions in admissions between 2003 to 2004 and 2004 to 2005 were due to smaller number of cases among men: 281.5 vs 233.5/100,000, P = 0.0011, and current smokers 408 vs 302 admissions, P < 0.0001; no significant change among women, former smokers, and never‐smokers The 2nd reduction in ACS admissions 2005 compared to 2006 to 2007 was due to a reduction among men, 235.4 vs 195.2, P = 0.0021 and in current smokers 325 vs 271, P = 0.0269, and in never‐smokers 355 vs 302, P = 0.0386 There was no significant change in total deaths for all causes during the study period and the number of deaths from circulatory causes declined 6.5% Smoking legislation was associated with early significant decrease in hospital admissions for ACS. A further reduction was noted 2 years post‐legislation	Smoking status self reported No validation
Gasparrini 2009	Italy, Tuscany Comprehensive 2005	2000 to 2004 pre‐ban 13,456 AMI cases registered. 2005 post‐legislation 2190 cases registered A decrease of 5.4% in AMI rates was observed in age group 30 to 64 years post‐legislation, RR 0.95 (95% CI 0.89 to 1.00, P = 0.07 (NS)). Adjusting for linear or non‐linear time trends (age groups in 10 year bands) or gender did not provide any statistical significant differences post‐legislation	No smoking status reported
Hahn 2011	USA, Kentucky, Lexington‐ Fayette County Comprehensive 2004	AMI hospitalization rates in age group ≥ 35 years decreased for women after law enacted; adjusted RR 0.77 (95% CI 0.62 to 0.96, P < 0.05). A decrease in rate from 334.1/100,000 to 237.3/100,000. The rate for men increased 424.6/100,000 to 438.4/100,000, RR 1.11 (95% CI 0.91 to 1.36, NS). The post‐law decline for women was maintained during the study period Gender differences observed in post‐legislation period for different workers covered by laws Pre‐ban admission age 67.3 years, post‐ban 65.5 years, t = 3.2, P = 0.001	No smoking status reported
Humair 2014	Switzerland, Geneva Partial ban with period of suspension 2008	10% trend in reduced admissions for ACS IRR 0.90 (95% CI 0.80 to 1.00, P = 0.24)	No smoking status included
Jan 2014	Panama Comprehensive 2008	Adjusted RR for AMI comparing baseline with 1st post‐smoking ban period was 0.982 (95% CI 0.967 to 0.997, P = 0.023), 1.8% decrease. The adjusted RR increased in the 2nd post‐ban period, RR 1.049 (95% CI 1.022 to 1.077, P = 0.0001) The adjusted AMI RR for women was 1.075 (95% CI 1.033 to 1.119, P = 0.0001), NS for men The adjusted RR reduced following the tax increase (final post‐ban period) RR 0.985 (95% CI 0.971 to 0.999, P = 0.041) No seasonality trends or linear trends in AMI case series tests	No smoking status reported Authors report results of reduced prevalence from other national data source
Kent 2012	Ireland Comprehensive 2004	Significant differences in admissions for ACS observed adjusted RR 0.82 (95% CI 0.70 to 0.97, P = 0.02). Reduced admissions in aged 50 to 55 years and 60 to 69 years. No changed in admissions in other age groups	No smoking status reported
Liu 2013	England, Liverpool Comprehensive 2007	Age‐adjusted CHD admissions increased in men by 8%, RR 1.08 (95% CI 1.06 to 1.11) and increased in women by 12%, RR 1.12 (95% CI 1.09 to 1.16) Age‐adjusted rates for MI admissions decreased post‐legislation by 41.6% for men, RR 0.584 (95% CI 0.542 to 0.629) and 42.6% for women, RR 0.574 (95% CI 0.520 to 0.633) Modelling identified that MI admissions reduced by 45% (95% CI 58.0 to 28.4), post‐legislation (2010 to 2011 compared to 2005/2006) in the 10 most deprived wards In comparison, the middle‐ranked wards identified 42.3% reduction in MI admissions (95% CI 56.4 to 23.6) For the 10 most affluent wards, MI admissions reduced 38.6% (95% CI 57.5 to 11.2). Absolute risk difference between least‐deprived wards for first 2 years was 69.8 MI admissions/100,000 person years compared to 2010 and 2011 data, 32 MI admissions/100,000 person years; RR 0.46 (95% CI 0.044 to 4.76) ARIMA analysis identified statistically significant effects of smoking ban for men in the most deprived wards and middle‐ranked wards Reduction in MI admissions following smoking ban was greater than secular trends. Upstream intervention	Smoking status not reported
Roberts 2012	USA, Rhode Island Comprehensive 2006/2007 2008/2009	AMI age‐adjusted admission rate pre‐ban (2003) was 35.2/10,000 population (95% CI 34.0 to 36.5) and post‐phase 11 of the ban in 2009, 23.1/10,000 population (95% CI 22.1 to 24.1) Between 2003 and 2007, following the 1st implementation of the smoking ban, the number of admissions for AMI decreased 17.1%, with a reduction in reimbursed hospital costs	No smoking status reported
Sargent 2012	Germany Federal and State bans Partial 2007 to 2008	Cohort aged 30 to 105 years, mean 56 years. 66.5% women registered. 43.5% of cohort were retired, 39.9% of members were employed. 2.2% of cohort were hospitalized for angina pectoris, and 1.1% of cohort had been hospitalized for AMI during the study period At 1 year follow‐up, smoking bans associated with 13.28% (95% CI 8.19 to 18.36) reduction in admissions for angina pectoris and an 8.58% (95% CI 4.99 to 12.17) reduction in AMI hospitalizations The percent reduction in AMI did not differ with respect to gender. Reductions in admissions for AMI higher for younger participants (30 to 68 years) compared to older group, 15.77% (95% CI 10.57 to 20.97) After the law, there was a statistically significant downward trend in admissions for angina with slope resulting in a decline of about 5 hospitalizations per month slope = −5.33 (95% CI 7.18 to 3.48). The percent reduction in angina was not significantly different for older vs younger individuals, or men vs women. Larger reductions in hospitalizations for angina were observed in older participants,15.66% (95% CI 10.9 to 20.39) Hospitalization costs reduced during study period. Overall the introduction of smoking ban was associated with prevention of 1880 hospitalizations and savings of EUR 7.7 million	No smoking status reported
Schmucker 2014	Germany, Breman Partial 2008	3545 patients admitted. Mean age 63 ± 10 years. 72% were men, 20% diabetes mellitus and 44% active smokers Smokers with STEMI were younger than nonsmokers 56 years ± 12 vs 69 ± 12, P < 0.01; men, 80% vs 66%, P < 0.01 Smokers with STEMI had significantly fewer coronary vessels diseased compared to nonsmokers, 1.76 ± 0.8 vs 1.99 ± 0.8, P < 0.01. (Nonsmokers in study included ex‐smokers in analyses) Hospitalization rates for STEMI decreased post‐smoking ban, a reduction from 65 ± 10 per month to 55 ± 9 Number of nonsmokers admitted for STEMI significantly decreased from 39 cases/month pre‐ban to 29 cases, P < 0.01. This reduction was observed in both genders and all ages in nonsmokers. Greatest reductions in nonsmokers were in those aged ≤ 65 years, 32%, P < 0.01 and in those > 65 years, P < 0.01 (after adjusting for confounders hypertension, obesity, diabetes mellitus). 16% (P < 0.01) reduction in total STEMI admissions post‐ban. Overall 26% reduction (P < 0.01) in admissions among nonsmokers. There was no significant difference in the number of smokers admitted for STEMI post‐smoking ban	Self‐reported smoking status
Sebrié 2014	Uruguay Comprehensive 2006	11,135 cases identified over study period. 65% were men (n = 7287). In 2008 there was a significant drop in AMI monthly admissions ‐35.9 ± 10.1 (SE), constant 167 ± 7, a 22% drop. A similar reduction was observed for men, women and people aged 40 to 65 years and aged 56 years and older The 2nd follow‐up analyses 2004 to 2010 identified a drop of 30.9 cases/month AMI admissions (95% CI ‐49.8 to ‐11.8, P = 0.002) The effect of the law did not increase or decrease over time The overall drop in AMI monthly admissions was 17%, IRR 0.829 (95% CI 0.743 to 0.925, P = 0.001) (to 2010) following smoke‐free legislation The results from 2010 analyses confirm the sustained impact of smoke‐free legislation on AMI admissions	No smoking status reported
Séguret 2014	France Comprehensive 1991, 2006, 2008	Adjusted for age and sex admission rates for ACS admissions observed a reduction from 269.1/100,000 2003 to 234, RR 0.87 (95% CI 0.85 to 0.89) in 2009. A reduction of 12.8% After adjusting for linear trends, reductions linked to the ban were not significant when analysed for gender or age groups (men aged ≤ 55 years or > 55 years and women ≤ 65 years or > 65 years). The study did not demonstrate a significant effect of a 2‐phase ban on ACS admissions. ACS rate was reducing in France during this 7‐year period	No smoking status reported
Controlled before‐and‐after studies
Alsever 2009	USA, Pueblo City, Colorado Control: Pueblo county outside city limits, El Paso county Comprehensive 2003	Significant drop in admissions for AMI among residents within Pueblo city limits continued in Phase 2 of the study (follow‐up 36 months) Decrease 152 per 100,000 person years, a decline of 19% since Phase 1 and a decline of 41% pre‐legislation RR 0.59 (95% CI 0.49 to 0.70) Males RR 0.67 (95% CI 0.52 to 0.82); Females RR 0.48 (95% CI 0.36 to 0.60) (pre‐legislation to Phase 2) No significant changes were observed among residents outside the city limits RR 1.03 (95% CI 0.68 to 1.39) or in El Paso County, RR 0.95 (95% CI 0.87 to 1.03) Adjusting for secular trends in pre ban period was not significant. Sustained reduction in rates of AMI admissions observed over 3‐year period	No smoking status reported
Bonetti 2011	Switzerland, Canton Graubünden Control Canton Lucerne Partial Canton Ban 2008 (National Ban up to 2010)	Adjusted for air pollution, drug prescribing and comorbidities Statistically significant differences in admissions post‐legislation identified in Graubünden (229 and 242 admissions pre‐law; 183 and 188 admissions post‐law; P < 0.05) Overall reduction in number of AMI admissions in Graubünden in the 2 years post‐ban; 21% lower than in the 2 pre‐ban years. The reduction most pronounced in nonsmokers, women and individuals with documented coronary artery disease, including those with prior AMI and prior coronary intervention or graft surgery Decrease in 2nd year of ban limited to nonsmokers 151 (2006) vs 108 (2010), P < 0.05 No decrease observed in control Lucerne No association found between magnitude of outdoor air pollution and incidence of AMI. Use of lipid‐lowering drugs increased in Graubünden and in Lucerne	Smoking status reported No validation
Bruintjes 2011	USA, Greeley, Colorado and surrounding area Smoking ordinance Greeley Control: areas outside city Comprehensive 2003	Prevalence of smoking: 482 hospitalizations analysed in Greeley with 224 in residents of surrounding area. 23.7% active smokers in Greeley; 61.4% of patients were men. (30.0% smokers in control area). A significant decrease in hospital incidence rates in Greeley observed post‐ordinance RR 0.73 (95% CI 0.59 to 0.90). NS result in comparison area. Difference between Greeley and comparison area was NS, P = 0.48 Regression analyses identified smokers experienced statistically significant reductions in hospitalizations in Greeley RR 0.44 (95% CI 0.29 to 0.65) Reduction in AMI rates in smokers in surrounding area did not differ from Greeley, P = 0.38 Significant difference observed post‐ordinance, but not in comparison with surrounding area	Smoking status reported
Di Valentino 2015	Switzerland, Canton Ticino Partial (local smoke‐free ordinance) 2007 Compared to Canton of Basel (no ban)	Mean incidence of STEMI reduced post‐legislation in Ticino 123.7/100,000 pre‐ban, to post ban 92.9 (2007 to 2008), P = 0.002; 101.6 (2008 to 2009), P = 0.024; 89.6 (2009 to 2010), P = 0.001 Post‐ban reduction in STEMI admissions observed in age group 65 years and older irrespective of gender, each year post‐ban, P = 0.0001 In the under‐65‐year age group , the mean incidence of STEMI admissions decreased in 1st year post‐ban 109.0 vs 85.3, P = 0.01 No significant differences in annual number of STEMI admissions in Basel during the study period except in age group 65 years and older 362.3 (pre‐) vs 223.6, 234.4, 199.8. Lower STEMI admissions noted in Basel compared to Ticino during study period	No smoking status reported
Ferrante 2012	Argentina, Santa Fe Comprehensive August 2006 Control: Buenos Aires City: partial October 2006	Significant reduction in in ACS admissions in Santa Fe ‐2.5 admissions/100,000, P = 0.03 and persistence change over time post‐law 0.26 fewer admissions/100,000 inhabitants per month (95% CI ‐0.39 to ‐0.13, P < 0.001). 13% reduction compared to control city, RR 0.74 (95% CI 0.63 to 0.86) In Buenos Aires City no change post‐ban, P = 0.28 or over time P = 0.89 Slight decrease (P = 0.84, NS) in smoking prevalence during study period (2005 to 2009) from national prevalence survey. More quit attempts in Sante Fe prior to ban than in control 53.2% (95% CI 42.5% to 63.6%) vs 44.4% (95% CI 34.3% to 55.0%, P = 0.045). No change in proportion of daily smokers or cigarettes consumed 100% smoke‐free law more effective in reducing and sustaining reduction in admissions for ACS in Sante Fe	No smoking status reported from data Prevalence reported from other data source
Gaudreau 2013	Canada, Prince Edward Island Comprehensive 2003 Control: New Brunswick Province	Significant reduction in mean rate of AMIs 5.92 cases/100,000 person months, P = 0.04 post‐smoking ban. The trend of admissions for angina in men reduced ‐0.44 cases/100,000 person months, P = 0.01 at 1 to 67 months post‐smoke‐free law. No significant difference when comparing age groups 35 to 64 years and 65 to 104 years No significant difference for other cardiovascular admissions in study population	No smoking status included
Head 2012	USA, Beaumont City, Texas Control: Tyler Texas and All Texas Comprehensive 2006	Texas BRFSS data estimated ethnicity of current smokers 23% black, 20% white during 2005 to 2008 Discharges for all participants (non‐Hispanic black and non‐Hispanic white) declined significantly post‐legislation in Beaumont for AMI, RR 0.74 (95% CI 0.65 to 0.85) and stroke RR 0.71 (95% CI 0.62 to 0.82)	No smoking status reported from data Reports state smoking prevalence from other data source
Herman 2011	USA, Arizona counties with bans Control: counties with no bans Comprehensive 2007	Statistically significant reduction in hospital admissions comparing ban counties with no‐ban counties, AMI 159 cases, 13% reduction in cases, P = 0.01, angina 63 cases, 33% reduction, P = 0.014	No smoking status reported
Khuder 2007	USA, Intervention city: Bowling Green, Ohio Control city: Kent, Ohio Partial ban 2002	Admission rates for CHD‐related diseases showed downward trend during study period Admission rates CHD in intervention city reduced 36/10,000 population in 2002 to 22 per 10,000 in 2003; 39% decrease (95% CI 33% to 45%) and to 19/10,000 in 2005, 47% decrease (95% CI 41% to 55%). Further ARIMA models identified a downward trend in admissions in Bowling Green, omega estimates: ω = ‐1.69, P = 0.036 compared to Kent City, ω ‐1.14, P = 0.183 No observed changes noted in Kent compared to reduced CHD admissions in Bowling Green	No smoking status reported
Loomis 2012	USA, Florida 2003, (partial) New York 1985, 2003 Comprehensive Control: Oregon (partial ban)	The effect of comprehensive smoking ban on AMI rates in aged > 35 years was significant in New York, marginally significant at 10% level in Florida The interaction of time and law is significant for Florida and New York. This indicates rates of AMI decreasing over time post‐comprehensive legislation Moderate smoke‐free laws in Oregon were associated with lower AMI rates β = 3.846, P < 0.05. The interaction with time was negative and significant β = ‐0.242, P < 0.01 Rates for AMI hospitalizations reduced 18.4% (95% CI 8.8 to 28.0) in Florida (annual decline of 5.3%) and 15.9% (95% CI 11.0 to 20.1), β = ‐1.483, P < 0.05 in New York This is equivalent to 28,649 fewer age‐adjusted admissions (95% CI 20,292 to 37,006; annual decline of 4.4%) for New York The few comprehensive smoke‐free laws in Oregon were not associated with state reduction in admissions for MI or stroke	No smoking status reported
Naiman 2010	Canada, Toronto 1999, 2001 Comprehensive 2004 13 municipalities had bans Control cities: Durham Region, Thunder Bay (no bans)	A 39% reduction in cardiovascular conditions (95% CI 38 to 40), and a 33% reduction in admissions for respiratory conditions (95% CI 32 to 34) were observed after 2001 ban A significant reduction in admissions for angina were observed after the first ban, –0.913 (95% CI ‐1.24 to ‐0.59, P < 0.001) A significant reduction in admissions for all other conditions observed after the 2nd phase of the ban was enacted (restaurants) Only a significant reduction in admissions for AMI were noted after the 3rd phase of the ban, ‐0.611 (95% CI ‐1.03 to ‐0.19, P = 0.004). Authors suggest that reduction in hospital admissions unlikely due to decreased active smoking No significant results detected for specific age group or gender reported	Smoking status reported from national Canadian survey. No smoking status data from main data set.
Sargent 2004	USA Helena, Montana, Ordinance Partial ban (then suspended) June 2002 Control: non‐residents	Reduction in monthly AMI admissions in residents Helena – 16 (95% CI ‐31.7 to ‐0. 3) post‐ordinance. No significant decrease in admissions for those living outside of Helena	No smoking status reported
Seo 2007	USA, Monroe County Comprehensive 2005 Control: Delaware County, Indiana	Admission rates for AMI. There was a significant decrease in Monroe County but not in matched control Delaware County from the period August 2001 to May 2003 to the period August 2003 to May 2005 during which the smoke‐free law was in effect for nonsmoking people. Monroe: 17 to 5 (95% CI ‐21.19 to ‐2.81) vs Delaware:18 to 16 (95% CI ‐13.43 to 9.43). There were no admissions for AMI among nonsmoking people from January 1st to May 2005 when the ban was extended to include bars and clubs. Non‐significant reduction in admissions for AMI amongst smokers in Monroe from 8 pre‐law to 7 post‐law and in Delaware from 8 pre‐law to 6 post‐law during this period There was a significant difference in AMI admissions rates from August 2003 to May 2005 between Monroe and the control area 5 vs 16, change 11 (95% CI 2.02 to 19.98)	Self‐reported smoking status
Vander Weg 2012	USA state bans 1991 to 2008 Ban varied by state Control: states with no bans	1991 to 2008 data analysed Risk‐adjusted hospital admission rates for AMI reduced 20 to 21% in the 36 months post‐implementation of smoking bans in restaurants, bars and workplaces (P < 0.001 for each ban) At baseline, counties with bans in place had higher admission rates for AMI compared to controls (and higher admissions for hip fractures) Counties with bans in 2008 had more Medicare enrollees and larger proportion of white residents At 36 months post‐legislation, counties with bans had significantly lower AMI admission rates compared to no bans: RR 0.79, (No CI reported) P < 0.001 (workplace ban in place). Significant downward trends over time as increase in bans in different settings	No smoking status reported
Uncontrolled before‐and‐after studies
Cesaroni 2008	Italy, Rome Comprehensive 2005	Prevalence: men: 34.9% pre‐law period (2002 ‐ 2003) to 30.5% post‐law period (2005); women: 20.6% pre‐law to 20.4% post‐law Significant reduction in acute coronary events in 35‐ to 64‐year‐olds from pre‐law to post‐law period, RR 0.89 (95% CI 0.85 to 0.93) and in 65‐ to 74‐year‐olds, RR 0.92 (95% CI 0.88 to 0.97) No change in 75‐ to 84‐year‐olds, RR 1.02 (95% CI 0.98 to 1.07) Data from the post‐law was compared with data in the previous year, the effect of the law was statistically significant on men but not on women and was greater for residents living in lower socioeconomic areas than those from higher socioeconomic areas Fewer acute coronary events in 35‐ to 64‐year‐olds identified (11.2%)	Self‐reported smoking status from other survey No smoking status from admissions data
Hurt 2012	USA, Minnesota, Olmsted County 2002, 2007 Comprehensive 2007	Significant differences noted pre‐ordinance 1 and post‐ordinance 2 for MI. Incidence of MI declined by 33%, P < 0.001 from 150.8 to 100.7/100,000 population adjusted (age and gender) RR 0.67 (95% CI 0.53 to 0.83, P < 0.001)	Smoking status self‐reported
Juster 2007	USA, New York Comprehensive 2003	In 2004, hospital admissions for AMI were reduced by 8% as a result of the comprehensive ban, equivalent to 3813 fewer admissions for AMI The smoking ban was associated with a reduction in admissions for AMI on average 0.32/100,000 persons per month in all counties in New York state (95% CI ‐0.47 to ‐0.16, P < 0.001)	No smoking status reported
Lemstra 2008	Canada, Saskatoon Comprehensive 2004	Age‐standardized incidence rate of AMI per 100,000 population in Saskatoon 176.1 (95% CI 165.3 to 186.8) before smoke‐free ban (1st July 2000 to 30 June 2004) to 152.4 (95% CI 135.3 to 169.3) post‐ban (1 July 2004 to 30 June 2005) Incidence rate ratio: 0.87 (95% CI 0.84 to 0.90). 13% reduction in AMI discharges in period following legislation	Smoking status reported from survey data
Lippert 2012	Country: USA, Arizona 2007* Colorado 2006 District of Columbia 2007 Hawaii 2006* Illinois 2008* Iowa 2008* Louisiana 2007 Maryland 2008* Minnesota 2007 Nevada 2006 New Hampshire 2007 New Jersey 2006* New Mexico 2007 Ohio 2006* Pennsylvania 2008 Puerto Rico 2007* Utah 2006* Clean Indoor Air Act (varied implementation) * all comprehensive bans. Remaining states: partial bans.	7 States had significant decrease in prevalence of CHD/angina post‐ban: Arizona, District of Columbia, Hawaii, New Hampshire, New Jersey, New Mexico, Pennsylvania (state N) Arizona: (311) 4.7% (95% CI 3.6 to 5.8) vs (346) 3.4% (95% CI 2.8 to 3.9, P ≤ 0.0001) District of Columbia: (141) 2.9% (95% CI 2.3 to 3.5) vs (132) 2.0% (95% CI 1.6 to 2.4, P < 0.001) Hawaii: (257) 3.4%(95% CI 2.8 to 4.0) vs (247) 2.6% (95% CI 2.2 to 3.1, P < 0.001) New Hampshire: (377) 4.5% (95% CI 4.0 to 5.0) vs (336) 3.6% (95% CI 3.1 to 4.1, P ≤ 0.001) New Jersey: (801) 4.6% (95% CI 4.2 to 5.0) vs (592) 3.6% (95% CI 3.2 to 4.0, P ≤ 0.0001) New Mexico: (340) 3.8% (95% CI 3.3 to 4.3) vs (438) 3.2% (95% CI 2.8 to 3.6, P ≤ 0.01) Pennsylvania: (891) 5.4% (95% CI 4.8 to 6.0) vs (625) 4.7% (95% CI 4.2 to 5.2, P ≤ 0.01) 2 states had increased prevalence of CHD/angina: Colorado, Louisiana 7 states/Territory had significant reductions in AMI post‐ban (state N) District of Columbia: (149) 3.3% (95% CI 2.7 to 3.9) vs (127) 1.9% (95% CI 1.5 to 2.3, P ≤ 0.0001) Hawaii: (260) 3.6% (95% CI 3.0 to 4.2) vs (263) 2.9% (95% CI 2.4 to 3.4, P ≤ 0.01) Iowa: (317) 4.7% (95% CI 4.1 to 5.3) vs (344) 4.1% (95% CI 3.6 to 4.6, P < 0.05) Minnesota: (202) 3.4% (95% CI 2.9 to 3.9) vs (271) 2.8% (95% CI 2.4 to 3.2, P < 0.05) New Hampshire: (321) 4.0% (95% CI 3.5 to 4.5) vs (296) 3.4% (95% CI 2.9 to 3.9, P < 0.05) New Jersey:(676) 3.9% (95% CI 3.5 to 4.3) vs (567) 3.5% (95% CI 3.1 to 4.0, P < 0.05) Puerto Rico: (301) 4.7% (95% CI 4.1 to 5.3) vs (268) 4.0% (95% CI 3.4 to 4.7, P < 0.05) Four states had increased prevalence of AMI post‐ban: Colorado, Louisiana, Nevada, Pennsylvania (NS) 14 States had significant decrease in prevalence of current smokers. Highest difference post‐ban observed in New Hampshire, 3% change	Self‐reported smoking status and reported health outcomes
McGhee 2014	Hong Kong Partial 2007	Study period prior to comprehensive ban (July 2009). Partial smoking bans associated with 9% decrease in admissions for ischaemic heart disease (95% CI ‐13.59 to ‐ 4.17, P < 0.05)	No smoking status reported
North Carolina 2011	USA, North Carolina Comprehensive 2010	Regression analyses identified a 21% decrease in emergency admissions for AMI 12 months following implementation of smoke‐free restaurant and bars legislation RR 0.79 (95% CI 0.75 to 0.83) Reduction in admissions: men aged 18 to 59 years 2385 vs 1916; aged ≥ 60 years 3196 vs 2885 Women aged 18 to 59 years 946 vs 778; aged ≥ 60 years 2901 vs 2421 Additional modelling including interaction variables including time, gender, age category did not improve the model Additional modelling analyses identified improved outcomes were calculated using false start dates for legislation	No smoking status reported
Pell 2008	Scotland Comprehensive March 2006	In people admitted for ACS in Scotland, there was no significant reduction in self‐reported number of cigarettes smoked in the pre‐ or post‐law periods or the geometric mean cotinine level, 152 to 147 ng/ml, P = 0.72 Never‐smokers reported decrease in SHS exposure and biochemically verified, serum cotinine mean 0.68 to 0.56 ng/ml; P < 0.001 No significant change for nonsmokers or ex‐smokers (all admitted for ACS) reporting "no exposure" to SHS from pre‐ to post‐law period in either "own home" or "other people's homes". Never‐smokers reporting "no exposure" in own home: 83% (565/677) pre‐law vs 86% (460/537) post‐law, P = 0.64. Never‐smokers reporting "no exposure" in "other people's homes": 91% (617/677) pre‐law vs 92% (495/537) post‐law, P = 0.34 14% reduction in ACS admissions among smokers, 19% reduction among ex‐smokers and 21% reduction in never‐smokers. Greater reduction in admissions current smokers: women 19% (95% CI 15% to 23%) compared to men 11% (95% CI 9% to 13%) Reductions highest in women nonsmokers 23% (95% CI 20% to 26%) compared to men nonsmokers 18% (95% CI 16% to 20%) Greater reduction in admissions detected in male smokers aged ≤ 55 years and in women ≤ 65 years 9% (95% CI 6% to 12%) when compared to older people 8% (95% CI 15% to 21%) Similar results obtained for nonsmokers 8% (95% CI 4 to 12) vs 22% (95% CI 20 to 24).	Smoking status validated
Rajkumar 2014	Switzerland, Basel City, Basel County and Zurich Partial 2010	Pulse wave velocity and heart rate variability parameters significantly changed (dose‐dependent) for the 55 nonsmoking hospitality employees. A 1 cpd decrease was associated with a 2.3% (95% CI 0.2 to 4.4; P < 0.031) higher root mean square of successive differences, a 5.7% (95% CI 0. to 10.2; P < 0.02) higher high‐frequency component and a 0.72% (95% CI 0.4 to 1.05; P < 0.001) lower pulse wave velocity The measures significantly improved after introducing smoke‐free legislation and identify a decreased cardio vascular risk	SHS validated measure Self‐reported smoking status
Yildiz 2015	Turkey, Kocaeli City Comprehensive 2009	Admissions for diagnoses of COPD and MI were unchanged (NS differences) post‐legislation	No smoking status reported

Open in a separate window

2.1

Analysis

Comparison 2 Respiratory health outcomes, Outcome 1 COPD.

COPD
Study	Location/ Intervention	Outcomes	Smoking status
ITS studies
Croghan 2015	USA, Minnesota, Olmstead County Comprehensive 2007	In relation to COPD, the implementation of smoke‐free legislation was not associated with a downward step change in ED visits P = 0.158 or change in trend, P = 0.313.	No smoking status reported
Humair 2014	Switzerland, Geneva Partial ban (with period of suspension) 2008	Hospitalizations for COPD significantly decreased over 4 periods of time, IRR 0.54 (95% CI 0.42 to 0.68)	No smoking status reported
Kent 2012	Ireland Comprehensive 2004	Admissions for pulmonary illness 439/100,000 population per annum to 396/100,000, 1 year post‐ban unadjusted RR 0.91 (95% CI 0.83 to 0.99, P = 0.048) and adjusted for confounders RR 0.85 (95% CI 0.72 to 0.99, P = 0.04) Significant differences observed for asthma and pneumonia, but not for COPD in any age group	No smoking status reported
Controlled before‐and‐after studies
Dusemund 2015	Switzerland, Canton of Graubünden Local ordinance: Partial 2008 Control: Rest of Switzerland (not including Graubünden or Ticino)	22.4% reduction in incidence of AECOPD admissions, IRR 0.78 (95% CI 0.68 to 0.88, P < 0.001). Rest of Switzerland, reduction 7%, IRR 0.93 (95% CI 0.91 to 0.95, P < 0.001) Greater reduction in admissions observed in Intervention Canton, P = 0.008 compared to control	No smoking status reported
Gaudreau 2013	Canada, Prince Edward Island Comprehensive 2003 Control: New Brunswick Province	No significant differences reported for respiratory admissions	No smoking status reported
Hahn 2014	USA, Kentucky Comprehensive 2004, 2008 to 2011 Control: counties with smoking policy < 12 months or no ban	Adjusting for all characteristics, population and seasonal trend factors, risk ratio of COPD hospitalizations in communities with comprehensive smoking bans was 0.781 compared to communities with a weak or no policy Chi² = 6.65, P = 0.01; 95% CI 0.647 to 0.942 The risk ratio of hospitalizations for COPD in communities with established laws was 0.789 compared to communities with new or no laws Chi² = 9.91, P = 0.02; 95% CI 0.680 to 0.914 Protective factors for reduced COPD admissions were being male, aged 45 to 64 years and living in county with higher post‐secondary education Overall the study identified those living in counties with comprehensive smoke‐free laws were 22% less likely to be hospitalized for COPD compared to those living in counties with weak or no laws. Counties that had smoking bans in place for > 12 months were 21% less likely to be hospitalized for COPD compared to communities with laws < 12 months or no laws The study found that smoke‐free policies can improve health outcomes and can negate risk factors including lower socioeconomic status and living in rural tobacco‐growing communities	No smoking status reported
Head 2012	USA, Beaumont City, Texas Comprehensive 2006 Control: Tyler Texas and All Texas	COPD discharges for non‐Hispanic black residents RR 1.04 (95% CI 0.85 to 1.27 (NS)) and non‐Hispanic white residents RR 0.64 (95% CI 0.54 to 0.75) in Beaumont. NS in control areas	No smoking status reported
Naiman 2010	Canada, Toronto Comprehensive 1999, 2001, 2004 13 municipalities had bans Control cities: Durham Region, Thunder Bay (no bans)	33% reduction in admissions for respiratory conditions, (95% CI 32 to 34) observed after 2001 ban	Smoking status reported from national Canadian survey. No smoking status reported from main data set
Vander Weg 2012	USA State bans 1991 to 2008 Control: States with no bans	36 months post‐legislation, states with bans had significantly lower COPD admission rates compared to no bans, 11% to 17%, P < 0.001 with significant decreasing trends over time as increase in bans in different settings	No smoking status reported
Uncontrolled before‐and‐after studies
McGhee 2014	Hong Kong Partial 2007	Respiratory admissions and admission for lung cancer increased	No smoking status reported
Yildiz 2015	Turkey, Kocaeli City Comprehensive 2009	Bronchitis admissions reduced 39.8%, 44,141 to 26,558 post‐ban Admissions for LRTI decreased (7048 to 6738, P < 0.01) post‐legislation. Peak admission levels noted May 2010 Admissions for diagnoses of COPD and MI were unchanged (NS differences) post‐legislation Admissions for allergic rhinitis: NS trend analysis observed. Admissions for asthma showed NS increase (6805 vs 7895)	Principal diagnostic codes used No smoking status reported

Open in a separate window

3.1

Analysis

Comparison 3 Perinatal health outcomes, Outcome 1 Effect on perinatal health.

Effect on perinatal health
Study	Location/ Intervention	Outcomes
ITS studies
Amaral 2009	USA, California Local smoke free ordinances 1988 to 1994. State workplace ban Partial 1995	44181 births during study period Local workplace ordinances decreased the fraction of very low birth weight births in cities with local ordinances by 0.04 percentage points The implementation of local smoking ordinances was associated with a decrease in birth weight of 1.83 grams and increased gestation by 0.03 days The statewide ordinance was associated with a reduction in birth weight of 6.58 grams, P < 0.001 reducing to non‐significant changes of ‐2.45 grams and ‐3.12 grams after adjusting for different cities and ban trajectories Subgroup analyses identified that white mothers had an increase in gestation of 0.19 days, P < 0.001 after local ordinances and a significant decrease in very low birth weights by 0.06 percentage points, P < 0.001. Education level of mothers was not associated with significant differences in birth outcomes if local ordinance was in place. The statewide ordinance was significantly associated with lower birth weight and decreased gestation for lower‐educated mothers. Mothers with high school degree education were significantly associated with increased birth weight by 10 grams and decreased fraction of very low birth weight by 0.2 percentage points The statewide smoking ordinance, after adjusting for race and ethnicity, was associated with a significant reduction in birth weight of 7.2 grams, P < 0.05 for Hispanic mothers Results suggest that state work place smoking bans had a statistically significant but small negative effect on birth weight. Local ordinances did not have a similar effect
Cox 2013	Belgium Comprehensive 2010	606,877 singleton births delivered at 24 to 44 weeks gestation 448,520 births spontaneous deliveries Reductions in risk of preterm births reduced at each phase of smoking ban legislation After 2010 comprehensive ban, there was step change in the risk of spontaneous preterm delivery; slope change ‐2.65% (95% CI ‐5.11 to ‐0.13; P = 0.04) Similar reductions noted for all births, change ‐3.5% (95% CI ‐6.35 to ‐0.57; P = 0.02) No significant effect of smoking ban on risk of low birth weight or small‐for‐gestational‐age in population or on average birth weight (adjusted modelling)
Kabir 2013	Ireland Comprehensive 2004	Maternal smoking rates from 2000 to 2008 were higher in mothers who had SGA or vSGA. Data available from 1 maternity hospital 2000 to 2008 data. Not linked to national registry data Reduced monthly rates of SGA and vSGA reductions were observed post‐legislation (adjusted modelling); 4.7% to 4.3% (vSGA) and 6.9% to 6.6% (SGA). Effects continued in the post‐ban period: vSGA ‐0.6%, P < 0.0001 and SGA ‐0.02%, P < 0.0001 Significant reduction in low birth weights observed indicates evidence of smoke‐free legislation
Mackay 2012	Scotland Comprehensive 2006	Post‐legislation there was a significant reduction in current smoking rates, 25.4% to 18.8%, P < 0.001; and an increase in never‐smokers 57.3% to 58.4%, P < 0.001 Univariate modelling identified decrease 11.07% 95% CI 6.79 to 15.15, P < 0.001) in overall preterm deliveries and a decrease 10.26% (95% CI 4.04 to 16.07, P < 0.002) in spontaneous preterm labour. Significant for current and never‐smokers (model used date 1st January 2006, not 26th March) Prior to legislation multivariate analyses observed significant decreases (after adjusting for confounders) in SGA ‐4.52% (95% CI ‐8.28 to ‐0.60, P = 0.024); vSGA ‐7.95 (95% CI ‐15.87 to ‐7.35, P = 0.048), overall preterm delivery ‐11.72% (95% CI ‐15.87 to ‐7.35, P < 0.001), and for spontaneous preterm labour ‐11.35% (95% CI ‐17.20 to ‐5.09, P = 0.001). Significant reductions for current and nonsmokers Analyses using later start date identified increase in preterm delivery rates 3.83 (95% CI 1.42 to 6.30, P = 0.002), following adjustment for pre‐eclampsia data
Controlled before‐and‐after studies
Bharadwaj 2012	Norway Intervention: Mothers who work in bars and restaurants Control: All other mothers on register Comprehensive 2004	Post‐legislation mothers in the treatment group significantly reduced their risk of < 1500 grams birth by 1.9 percentage points (P < 0.05) and < 2000 grams birth by 2.5 percentage points (P < 0.05) and a significant reduction of 2.5 percentage points in being born preterm. There was no effect on < 1000g, APGAR score or if birth defect or male birth Approximately 20% of mothers in treatment group reported smoking at start of pregnancy; 64% were not smoking at start of pregnancy. No details reported for remainder. Following the smoking ban, mothers in the treatment group were 15.4% more likely to quit smoking during pregnancy (P < 0.05). The impact of quitting smoking at start of pregnancy increased birth weights on average by 162.5 grams, P < 0.05 There was no effect on birth weight for mothers who were nonsmokers at start of pregnancy. Mothers with missing data for smoking status also had increased birth weights of 105.5 grams and may suggest underreporting of smoking status Further analyses did not identity changes in birth weight associated with self‐reported income Occupational status during pregnancy changed for the treatment group. A number of mothers changed employment from bars and restaurants. Analyses of these changes did not identify significant differences to the results The impact of fathers' smoking status on birth weight identified a decrease of 77.09 grams in the treatment group (significant at 10% level) Further analyses on the impact of birth weight on later life success predicted that at age 28 years, a 100 gram increase in birth weight could increase adult income by 1.8%. For the sample in the study, their birth weight increase of 164 grams would translate into a 2.7% increase in salary This study identified that mothers working in bars and restaurants after smoke‐free legislation was introduced were 15% more like to quit smoking and this impacted on increased birth weights and on lower incidences of preterm births
Page 2012	USA, Colorado Intervention: Pueblo Control: El Paso Comprehensive 2003	Significant differences observed at baseline between the intervention city and the comparison in relation to mother's mean age. race, ethnicity, education, alcohol consumption, marital status and anaemia Significant differences existed in relation to previous pregnancy and medical history. Mothers from Pueblo were more likely to be Hispanic, have lower education and report previous pregnancy complications Results identified significantly more mothers were smoking in the control City 8.66% pre‐ban compared to 11.89% post‐ban, P < 0.0001 The percentage of smokers in Pueblo was 16.64% at baseline and 15.07% post‐ban, P < 0.0786, NS No significant differences were noted post‐ban in intervention city in relation to LBW. In control city, there was an increase in births < 3000 grams, 29.78% to 32.02%, P < 0.0001 Unadjusted rates of preterm babies did not change over time in Pueblo but increased in the control city, 7.93% to 9.23%, P < 0.001 Multivariable logistic regression modelling, adjusted for medical conditions, and birth characteristics found no significant association among location, ban and LBW Unadjusted models for preterm births identified a 21% (23% adjusted) reduction in odds of preterm birth associated with smoking ban, P < 0.05, in Pueblo When compared to control city, the smoking ban in Pueblo was associated with a 38% reduction in odds of maternal smoking, OR 0.620 (95% CI 0.529 to 0.727, P < 0.05)
Uncontrolled before‐and‐after studies
Kabir 2009	Ireland Comprehensive 2004	1 year post‐smoking legislation, a 25% decrease in risk of preterm births was observed; OR 0.75 (95% CI 0.59 to 0.96) There was a 43% increased risk of LBW; OR 1.43 (95% CI 1.10 to 1.85) after adjusting for all potential confounders A 12% reduction in maternal smoking rates (23.4% to 20.6%) was observed post‐ban There was an increase in smoking cessation prior to pregnancy in 2005, P = 0.047 Significant decline in preterm births and maternal smoking. Increase in LBW birth risks may reflect secular trend

Open in a separate window

4.1

Analysis

Comparison 4 Mortality outcomes, Outcome 1 Effect on mortality rates.

Effect on mortality rates
Study	Location and Ban	Study Design/ Outcomes
ITS studies
Aguero 2013	Spain, Girona Partial 2006	AMI admissions and mortality. AMI case fatality n = 891 Post‐ban decrease observed in AMI mortality rates, RR 0.82 (95% CI 0.71 to 0.94, P < 0.05) AMI mortality age < 65 years NS. ≥ 65 years RR 0.82 (95% CI 0.74 to 0.91, P < 0.05) (AHA/ESC definition) Subgroup analysis: women AMI mortality rates, RR 0.72 (95% CI 0.52 to 0.97, P < 0.05) men: AMI mortality rates, RR 0.85 (95% CI 0.72 to 0.99, P < 0.05) (AHA/ESC definition)
Cox 2014	Belgium, Flanders Partial 2007	Flemish Agency for Care and Health registry data on AMI deaths for people aged ≥ 30 years during 2000 to 2009. 38,992 AMI deaths recorded Decreased AMI mortality rates January 2006. Highest in women ≤ 60 years, ‐33.8% (95% CI ‐49.6 to ‐13.0) compared with effect for men ‐13.1% (95% CI ‐24.3 to ‐0.3) Estimates for aged ≥ 60 years ‐9.0% (95% CI ‐14.1 to – 3.7) for men, and ‐7.9% (95% CI ‐13.5 to ‐2.0) for women. Additional effect post‐2007 legislation for men aged ≥ 60 years with annual slope change ‐3.8% (95% CI ‐6.5 to ‐1.0) From January 2006 to December 2009, the model predicts 1715 fewer AMI deaths with smoke‐free legislation. Step change in mortality after 1st ban.
De Korte‐De Boer 2012	Netherlands, Limberg General work place ban 2004 Included hospitality sector 2008 Comprehensive 2008	Weekly incidence data on sudden cardiac arrest from ambulance registry South Limberg. 2305 sudden cardiac arrest cases recorded during study period (2002 to 2010), mean incidence 5.3 (SD 2.3) Adjusted Poisson model identified small increase in sudden cardiac death pre‐ban and reduced post‐ban 2004 ‐0.24% cases/week, P = 0.043. Equivalent to 6.8% reduction 1 year post‐ban, 22 cases. No further decrease noted after 2nd ban. This may be due to poor enforcement of 2008 legislation
Jan 2014	Panama Comprehensive 2008	Mortality regression models (January 2001 to April 2008) on changes in deaths from MI identified 0.5% annual percentage change, P < 0.05. The trend was 0.47% up to June 2010, with a trend change of ‐0.3% July 2010 to December 2012. The change was not statistically significant
Stallings‐Smith 2013	Ireland Comprehensive 2004	Impact on mortality rates. During study period 215,878 non‐trauma deaths recorded in population ≥ 35 years (2000 to 2007) Following smoke‐free legislation, there was a 13% immediate decrease in all‐cause mortality, RR 0.87 (95% CI 0.76 to 0.99) There was a 26% reduction in deaths from ischaemic heart disease, RR 0.74 (95% CI 0.63 to 0.88), a 32% reduction in deaths from stroke, RR 0.68 (95% CI 0.54 to 0.85), and a 38% reduction in COPD deaths, RR 0.62 (95% CI 0.46 to 0.83) after smoke‐free legislation Post‐ban reductions for IHD, stroke and COPD were observed in ages ≥ 65 years COPD mortality was reduced in women, RR 0.47 (95% CI 0.32 to 0.70) 15% decrease in non‐smoking‐related mortality, RR 0.85 (95% CI 0.75 to 0.97). There was a 5% increase in mortality each post‐ban year. No post‐ban annual trend reductions were detected for any smoking‐related causes of death Unadjusted estimates of 3726 smoking‐related deaths (95% CI 2305 to 4629) were probably prevented as a result of smoke‐free legislation, primarily due to reduced passive smoke exposure Follow‐up paper mortality rates and socioeconomic status (2000 to 2010) Stallings‐Smith 2014 identified smoking ban reduced inequalities in smoking‐related mortality. 2 factors emerged explaining 81% of the variance: Structural factors were characterised with high loadings on education, occupation, foreign nationality and family composition Material aspects loaded in the 2nd factor included: unemployment, housing tenure and car access No post‐ban annual trend effects were detected for any cause of death in the period 2000 to 2007 Post‐ban mortality effects of structural socioeconomic indicators identified a reduction in smoking‐related inequalities For IHD and COPD mortality rates, reductions were strongest in the most deprived tertile; decreases in stroke mortality were observed across all socioeconomic groups (Stallings‐Smith 2014)
Controlled before‐and‐after studies
Dove 2010	USA, Massachusetts Control 290 cities and towns with no bans Comprehensive 2004	AMI deaths recorded on national registry Post‐legislation statistically significant reduction in AMI mortality rates 7.4% (95% CI 3.3 to 11.4, P < 0.001); 270 fewer deaths Significant reduction in AMI mortality rates aged ≥ 75 years compared to younger, ‐9.1% (95% CI ‐13.9 to ‐4.1, P < 0.001). Higher reduction detected in women compared to men, ‐9.7% (95% CI ‐15.1 to ‐3.9, P < 0.001) No significant results in control groups State ban reduction ‐1.6% in 1st year and increased to ‐18.6%, P < 0.001, in 2nd year following legislation
Rodu 2012	USA, state bans California 1 January 1995 Utah 1 January 1995 South Dakota 1 July 2002 Delaware* 27 November 2002 Florida 1 July 2003 New York * 24 July 2003 * Comprehensive bans Remaining states no bans	Secondary analysis of AMI mortality rates aged > 45 years California: The AMI mortality rate declined pre‐ban 1992 to 1993 from 225/100,000 to 204/100,000, annual reduction of 3%. Post‐ban the AMI rate declined 2%, P = 0.16 Utah: 3 years pre‐ban, the AMI mortality rate decreased from 200 to 180/100,000; 3.3% annual reduction. In 1995, post‐ban, the rate declined 7.7%, P = 0.43 Between 1991 and 1994, no significant difference was noted in other 48 States without smoking bans at that time. South Dakota: In the 3 years pre‐ban, AMI mortality rates dropped 253 to 198/100,000, 7.2% annual reduction. In the year post‐ban, the rate increase 8.9% to 216/100,000, P = 0.007 Delaware: Pre‐ban the AMI mortality rate decreased 199 to 160/100,000, 6.6% annual decline. Post‐ban the rate decreased 8.1%, P = 0.89 Between 1999 and 2002 the AMI rate declined for the other 46 States without a ban. In 2003 the rate of AMI decline was 7.2%, significantly greater than expected, P < 0.0002. Florida: Pre‐ban the AMI mortality rate declined 169 to 132/100,000, 6.4% annual decline. Post‐ban the rate significantly reduced 8.8%, P = 0.04 New York: Pre‐ban AMI mortality rates reduced 187 to 160/100,000, 4.9% annual reduction. Post‐ban the rate significantly declined 12%, P < 0.0002 Statewide smoking bans had little or no immediate effect on AMI death rates
Uncontrolled before‐and‐after studies
Hurt 2012	USA, Minnesota, Olmsted County 2002, 2007 Comprehensive 2007	Incidence of sudden cardiac death (SCD) declined pre‐ordinance 1 and post‐ordinance 2 by 17%, P = 0.13, 109.1 to 92.0/100,000 population; RR 0.83 (95% CI 0.65 to 1.06) NS
McGhee 2014	Hong Kong Partial 2007	Hospital admission and mortality rates: Ischaemic heart disease, acute myocardial infarction, cerebrovascular disease, cardiovascular disease, respiratory disease, lung cancer, all natural causes, injury poisonings and external causes, cancer excluding lung cancer. Mortality rates for lung cancer diagnosis significantly reduced 5.65% (95% CI ‐9.73 to ‐1.39, P < 0.05) The authors suggest this is not attributable to the smoking ban, but to improved treatment and other factors as follow‐up post‐legislation is 12 months
Pell 2009	Scotland Comprehensive March 2006	Cohort study. Mortality rates in ACS admissions amongst nonsmokers All‐cause mortality increased from 10 in those with mean cotinine ≤ 0.1 ng.ml to 22 in those with cotinine > 0.9 ng/ml, P < 0.001 All‐cause mortality (after adjusting for age and gender) associated with cotinine > 0.9 ng/ml, OR 4.80 (95% CI 1.95 to 11.83, P = 0.003 Current smokers excluded from the primary analyses (n = 1831), 53 (3%) died and 78 (4%) died or were readmitted for myocardial infarction within 30 days of the index admission. The early risk of death in smokers was comparable to that among never‐smokers; however, the difference was no longer statistically significant when adjusted for differences in age
Villalbi 2011	Spain Partial 2005/2006	Secondary analysis of AMI mortality rates. 2004 to 2007 study period Reduction in AMI deaths observed 2004: Rate 119.99/100,000 population (95% CI 117.98 to 122.01) vs 2007: 102.28 (95% CI 100.49 to 104.07) Adjusted AMI mortality rates in 2004 and 2005 are similar, but in 2006 there is a 9% decline for men and 8.7% decline for women, especially aged > 64 years. In 2007 there is a statistically significant decline for men (‐4.8%), but not for women Post‐ban the annual age‐standardized AMI mortality risk was significantly reduced in the years after legislation compared to 2003/2004 rates Men: 2006: RR 0.90 (95% CI 0.88 to 0.93, P < 0.001). 2007: RR 0.86 (95% CI 0.83 to 0.88, P < 0.001) Women: 2006: RR 0.90 (95% CI 0.87 to 0.92, P < 0.001). 2007: RR 0.86 (95% CI 0.84 to 0.89, P < 0.001) The smoking ban was associated with a reduction in AMI mortality

Open in a separate window

Study Design

We did not identify any randomized controlled trials, due to a lack of feasibility in using this methodology in population‐level studies measuring the effect of national legislative smoking bans. Of the 77 studies included in this update, 36 used an interrupted time series design measuring the impact of smoking bans using data from national registries, episodes of monthly hospital admissions or discharges, or reporting multiple prevalence surveys from population health surveys (Aguero 2013; Amaral 2009; Bajoga 2011; Barnett 2009; Barr 2012; Barone‐Adesi 2011; Basel 2014; Bruckman 2011; Christensen 2014; Cox 2013; Cox 2014; Croghan 2015; Cronin 2012; De Korte‐De Boer 2012; Federico 2012; Gasparrini 2009; Gualano 2014; Hahn 2011; Humair 2014; Jan 2014; Kabir 2013; Kent 2012; Klein 2014; Liu 2013; Mackay 2010; Mackay 2011; Mackay 2012; Mackay 2013; Millett 2013; Roberts 2012; Sargent 2012; Schmucker 2014; Sebrié 2014; Séguret 2014; Sims 2013; Stallings‐Smith 2013).

Twenty‐three studies use a quasi‐experimental (controlled before‐and‐after) study EPOC 2013 design (Alsever 2009; Bharadwaj 2012; Bonetti 2011; Bruintjes 2011; Di Valentino 2015; Dove 2010; Dusemund 2015; Ferrante 2012; Gaudreau 2013; Hahn 2008; Hahn 2014; Head 2012; Herman 2011; Jones 2015; Khuder 2007; Landers 2014; Loomis 2012; Naiman 2010; Page 2012; Rodu 2012; Sargent 2004; Seo 2007; Vander Weg 2012). Three of these studies reported using a matched control area for comparison (Hahn 2014; Khuder 2007; Seo 2007). The remaining 18 studies used before‐and‐after methods with no control group (Cesaroni 2008; Durham 2011; Gallus 2007; Goodman 2007; Hurt 2012; Juster 2007; Kabir 2009; Larsson 2008; Lee 2011; Lemstra 2008; Lippert 2012; McGhee 2014; North Carolina 2011; Pell 2008; Pell 2009; Rajkumar 2014; Villalbi 2011; Yildiz 2015). Six of these studies used a cohort design (Durham 2011; Goodman 2007; Larsson 2008; Pell 2008; Pell 2009; Rajkumar 2014).

Excluded studies

For this update, we exclude 36 studies included in the first version, as they did not meet the revised inclusion criteria for this update (Abrams 2006; Akhtar 2007; Alcouffe 1997; Allwright 2005; Biener 2007; Bondy 2009; Braverman 2008; Brownson 1995; CDC 2007; Eagan 2006; Eisner 1998; Ellingsen 2006; Farrelly 2005; Fernandez 2009; Fernando 2007; Fichtenberg 2000; Fong 2006; Fowkes 2008; Galán 2007; Gilpin 2002; Gotz 2008; Hahn 2006; Haw 2007; Helakorpi 2008; Heloma 2003; Hyland 2009; Jiménez‐Ruiz 2008; Menzies 2006; Mulcahy 2005; Mullally 2009; Palmersheim 2006; Pearson 2009; Semple 2007; Vasselli 2008; Verdonk‐Kleinjan 2009; Waa 2006). We now included two further studies as secondary references in this update (Barone‐Adesi 2006; Bartecchi 2006).

In this update, we exclude uncontrolled before‐and‐after studies reporting unverified health outcomes or those which only reported cotinine biomarkers and no other additional health outcome data, as the focus for this update is on including studies reporting reduced passive exposure that also measured health outcomes. The evidence from the first version clearly established that reduced passive smoke exposure results in reduced cotinine measures. We exclude from this update studies reporting the impact of smoking bans on smoking prevalence, tobacco cessation or quit rates which are not representative population‐level measures. See Characteristics of excluded studies for specific details.

Risk of bias in included studies

We made explicit judgements of bias according to the criteria in the Cochrane Handbook for Systematic Reviews of Interventions (Cochrane Handbook, Higgins 2011). We provide a summary of the assessments in Figure 2. The study designs used in this review for evaluating a policy‐level health promotion outcome do not fulfil the criteria used to confirm a low risk of bias, and as such we consider the evidence to be at high risk of bias for many of the studies included. However, we acknowledge that the majority of study designs included in this update used data from large hospital and national data registries, and for 23 studies include a control reference area.

Open in a separate window

2

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Sequence generation and allocation concealment

The non‐randomized studies used in this review did not facilitate random sequence generation, allocation concealment or blinding of participants, as smoking is a visible and active process. A number of studies used large representative population surveys which employed stratified or random sampling nationally (Bajoga 2011; Federico 2012; Gualano 2014; Jones 2015; Lee 2011; Lippert 2012; Liu 2013; Mackay 2011). Volunteer samples were reported in four studies (Durham 2011; Goodman 2007; Larsson 2008; Rajkumar 2014).

Blinding

It was not possible to blind participants in the studies included in this review, as the intervention was a national public policy and smoking is visible. The use of large data sets also negated blinding. However the large data sets obtained from hospitals used the Internation Classification of Diseases (ICD) coding to confirm principal diagnoses. Studies reporting mortality data similarly used data sets from large national registries.

Incomplete outcome data

A number of studies did not report total sample sizes. Durham 2011 and Larsson 2008 reported high attrition rates, with consequent reporting bias for outcomes. Two studies reported the use of imputed scores (Aguero 2013; Hurt 2012). Klein 2014 reported that records were excluded from the data set when smoking status or other key descriptive variables including gestational age or data on duration of pregnancy were missing. This led to the exclusion of 6.3% of cases, amounting to over 30,000 records.

Selective reporting

Within this review a large number of studies used existing data sets, and individual‐level data were not available. Whilst the outcomes associated were reported, the data sets were pre‐existing and may have given rise to bias associated with misclassification of data, i.e. residual confounding. Prevalence studies used different data sets for each survey and this can introduce bias when combining data (Bajoga 2011; Federico 2012; Gallus 2007; Gualano 2014; Jones 2015; Lee 2011; Lippert 2012; Mackay 2011). There is a reliance on self‐reported, unverified smoking status in studies included in this update. Verified smoking status (confirming either smoker or nonsmoker status) was reported by Goodman 2007; Larsson 2008; Pell 2008; Pell 2009. Pell 2009 primarily analysed data for nonsmoker outcomes, but provided a comparison for current smokers, with limited data reported.

Other bias

Other bias identified in the included studies is the lack of adjusting for confounders, as data were not available within the accessed data sets. Smoking status was self‐reported for the majority of studies covering active and passive smoke exposure. Cesaroni 2008; Christensen 2014; Cox 2014; Ferrante 2012; Head 2012; Hurt 2012; Jan 2014; Kabir 2013; Mackay 2010; Naiman 2010; Stallings‐Smith 2013 report smoking prevalence data from other data or from national surveys, and not from their main data sources. A number of these studies only provided a single smoking prevalence result, and we have not included this information in further statistical analyses (Christensen 2014; Head 2012; Kabir 2013; Mackay 2010; Naiman 2010; Stallings‐Smith 2013). Kabir 2013 included maternal smoking prevalence data for analyses reported from an earlier paper (Kabir 2009). Verified smoking status was measured in four studies (Goodman 2007; Larsson 2008; Pell 2008; Pell 2009).

A number of the studies using data from large hospital or population registries did not provide information on individual smoking status or other individual confounders. However, these data sets used statistical modelling (both linear and non‐linear) and adjustments to account for confounding of included variables. A number of studies adjusted for air quality, pollution, influenza rates and seasonality, using national data sets in an effort to reduce confounding and influence on health outcomes.

Other factors that could have led to bias include: changed prescribing practices for statins during the period of data collection (Cesaroni 2008; Christensen 2014); legislation banning trans‐fatty acids in foods, resulting in dietary changes which could influence cardiovascular outcomes (Christensen 2014). Legislative changes during the period of data collection, including an increase in the price of cigarettes, was reported by Federico 2012, Jan 2014 and Klein 2014. This may have influenced their study outcomes. Bharadwaj 2012 reported changed occupational status for pregnant women during the period of the study, and this was identified as a factor which reduced the power of the study. Page 2012 reported significant differences in demographic data between the control area and the intervention area at baseline, and the influence of this on their outcomes. Larsson 2008 reported that the study was predominantly in women, as only 30% of the study participants were men. Schmucker 2014 included ex‐smokers in the group of nonsmokers, due to a small sample size of less than 6%, and inconsistent documentation. Di Valentino 2015 detected a significant reduction in the control area which did not have a ban in place. Other new legislation, including laws banning advertising and sales of cigarettes to minors, may have influenced these outcomes.

Sample size

Two studies reported power calculations (Bajoga 2011; Lee 2011). Aguero 2013 did not analyse the impact of legislative changes on mortality, due to the small sample size reported. Fifteen studies did not report a sample size (Alsever 2009; Bajoga 2011; Bharadwaj 2012; Bruckman 2011; Gaudreau 2013; Gualano 2014; Head 2012; Herman 2011; Khuder 2007; Landers 2014; Loomis 2012; McGhee 2014; Mackay 2011; Naiman 2010; Seo 2007), although a number of these studies reported that large data sets were used with samples in excess of 1000 and up to 26,000 participants during annual data collections. Seo 2007 does not include an overall sample size, although the totals included in tables reported in the paper are suggestive of small numbers. Naiman 2010 reported population statistics and analyses based on rates per 10,000 population.

Follow‐up

The minimum period required for follow‐up was six months. The period for follow‐up extended from nine months post‐legislative bans (Kabir 2009) up to 81 months (Stallings‐Smith 2013; secondary reference Stallings‐Smith 2014). Gualano 2014 reported an eight‐year follow‐up period post‐legislation. A number of studies reported phased implementation of national smoking bans in a variety of settings. Cox 2013, De Korte‐De Boer 2012, Gaudreau 2013, ,Hahn 2014, Naiman 2010, Roberts 2012, Sebrié 2014 and Séguret 2014 report phased implementation of smoking bans in Belgium, Netherland, France, USA, Canada and Uruguay. Landers 2014 detected the impact of county‐level and state‐level bans on child and adult asthma discharge rates across multiple US states; Amaral 2009 compared the impact of local and statewide ordinances on perinatal health outcomes in California over a period of six years.

Biochemical verification

Smoking status was not reported in the majority of studies included in this update. Biochemical verification of smoking status was measured through analysis of cotinine in saliva or urine for four studies (Goodman 2007; Larsson 2008; Pell 2008; Pell 2009). Health outcomes data were verified by primary diagnosis using International Classification of Diseases (ICD) codes. Definitions of current, ex‐ or nonsmoker status in prevalence surveys were reported using WHO guidelines; cotinine measures (when present) for nonsmoking status were confirmed as those less than 15 ng/ml (See Characteristics of included studies).

Adverse events

Four included studies identified adverse events which may have influenced their study populations and reported outcomes. Humair 2014 and Sargent 2004 reported suspension of smoking bans in each of their studies during the periods of data collection. Gualano 2014 reported that 2007 was the peak year in the Italian recession and that this may have influenced smoking rates. Head 2012 reported the influence of hurricanes Katrina and Rita, which may have affected population levels during their study period.

Assessment of heterogeneity

As in the original version of the review, due to the heterogeneity in clinical variation and study designs reporting primary and secondary outcomes, we did not attempt a meta‐analysis. We offer a qualitative narrative analysis to report the outcomes in this updated review.

Discussion

Legislation restricting or prohibiting smoking in work places and public places is a public health measure at the population level. There were no randomized controlled trials where the intervention was a smoking ban. The predominant study designs evaluating the effectiveness of smoking bans were interrupted time series studies, quasi‐experimental before‐and‐after studies with a control area for comparison, and before‐and‐after studies with no control area for reference. Three studies used matched areas for comparison of controls (Hahn 2014; Khuder 2007; Seo 2007). While the before‐and‐after studies with no controls were often unable to control for possible confounders and changes in secular trends over time, the interrupted time series studies used statistical modelling in an attempt to adjust for these effects in analyses. However, because of uncertainty about the underlying trends, some study authors noted that their results were sensitive to the choice of model.

The evidence supports a temporal association between the introduction of national smoke‐free bans and subsequent reductions in smoking‐related morbidity and mortality. Evidence for smoking bans in improving cardiovascular, respiratory and perinatal health outcomes for both smokers and nonsmokers is persuasive. The evidence in this update identified a dose‐response association, with sustained and improved health outcomes over time, specifically cardiovascular. As the period since bans were enacted has lengthened, improvements in health outcomes have increased or have been maintained. Evidence in this review identified improved health outcomes for nonsmokers in relation to cardiovascular and asthma health outcomes and to reduced mortality rates. Evidence of a biologically plausible effect emerged in studies examining STEMI admissions. Schmucker 2014 detected reduced STEMI rates for nonsmokers compared to smokers, with identified divergent trends in the incidence of disease observed. Di Valentino 2015 also suggests a biological plausibility, with reduced STEMI admissions in those aged 65 or older; however, smoking status was not reported in this study.

Perinatal outcomes provide evidence of reduced maternal smoking and acknowledged impact on foetal health. Inconsistent evidence emerged for other outcomes, including birth weights. The benefits identified in some studies are consistent with those reported in Been 2014, Jones 2014 and Kelleher 2014; however, the studies in this review do not provide compelling evidence of a clear association between smoke‐free legislation and improved perinatal outcomes; we need more evidence to confirm or refute such associations.

Consistent evidence of reduced mortality is reported, with an observed temporal dose‐response effect. Statistically significant reductions and downward trends were noted for cardiovascular and respiratory illnesses. Evidence of a reduction in mortality in lower socioeconomic groups is persuasive, especially in Stallings‐Smith 2014, given the duration of the study period (81 months).

As in the previous version of this review, inconsistent evidence emerged of the impact of smoking bans on reducing smoking prevalence rates and tobacco consumption.

The studies in this review are heterogeneous in their design, populations and interventions, and we were unable to perform statistical comparisons or meta‐analyses. Despite the different study designs, this update provides more methodologically robust studies than those reported in the first version, incorporating large data sets facilitating modelling and regression analyses and adjusting for non‐linear trends and confounders. The majority of studies have evaluated comprehensive smoking bans; only 18 studies investigated partial bans. Significant improvements in health outcomes were reported in countries where comprehensive bans were in place and compared to areas with either no ban or partial bans. Since the first version of this review (2010), there has been an increase in countries worldwide implementing national smoke‐free bans. The FCTC (WHO 2014) identified an 84% increase in countries implementing smoking policies, and a 61% increase in countries implementing complete smoking bans.

The 2008 MPOWER evidence‐based measures include protection from tobacco smoke to reduce tobacco‐related morbidity and mortality (WHO 2009; WHO 2013). The results from the original review indicated that introducing legislative smoking bans leads to a reduction in exposure to passive smoke. Key population groups benefiting from the enactment of legislative smoking bans reported in this review include pregnant women and their babies, children and nonsmokers. There is also evidence of improved cardiovascular outcomes for smokers in three studies (Bruintjes 2011; Cronin 2012; Pell 2008).

Socioeconomic gradients indicate that men in lower socioeconomic groups are benefiting from the effect of smoke‐free legislation. In the original version of the review, the evidence of the impact for active smoking was unclear but indicated a downward trend. The studies included in this update provide some evidence of reductions in smoking prevalence. However, a number of studies did not detect evidence of a change in prevalence, or change in rate of decline in prevalence, associated with the introduction of bans, irrespective of the population studies. Four studies (Bharadwaj 2012; Kabir 2009; Klein 2014; Mackay 2012) identified declining smoking rates in pregnant women, but this was not borne out for all studies.

Limitations in studies included in this review are the absence of randomised trials. The inevitable reliance on observational data means that we can only identify correlations between the introduction of smoking restrictions, and the health and behavioural outcomes of interest. The studies using national population surveys employed random sampling or stratified sampling techniques. The data sets used in many studies were relatively large and allowed for statistical modelling and adjustment for possible confounders. Small sample sizes are reported in a number of the studies which used volunteer samples recruited within the hospitality sector. A number of studies did not report sample sizes, and individual‐level data were not available within large registry data sets, which limited analyses for confounders, e.g. smoking status and comorbidities. Other confounders included increased pricing of cigarettes during study periods, removal of trans‐fatty acids in foods, and suspension of bans. These and other factors may have led to changes in health outcomes over the study periods which could not be controlled for in analyses. These may have influenced the reported results. It is possible that some studies that did not detect changes in health outcomes have not been published and are unavailable for inclusion in this review. However, this update includes some studies that did not identify a positive impact of smoking bans. We excluded from this review studies reporting only cotinine biomarkers; studies reporting passive smoke exposure had to include a measured health outcome. This provided a wider body of literature, but there are few studies which verified smoker status. Smoker status was reported in 24 studies in this review, and verified in only four.

From a public health perspective the impact of smoking legislation is to reduce passive smoke exposure and to reduce active smoking. Since first publication of this review in 2010, the evidence is mounting and the concentration of studies clearly identifies reduced passive smoke exposure with associated reductions in morbidity and mortality post‐smoking bans. Smoking policies usually comprise multicomponent efforts to tackle smoking cessation as well as the public health objective of reducing exposure to environmental tobacco smoke. Populations exposed to smoking restrictions are likely to be exposed to other interventions. The implementation of comprehensive legislation on smoking will necessitate other tobacco control measures to prepare for its successful implementation, such as increased media awareness, telephone smoking cessation helplines, and smoking cessation support services to ensure awareness, comprehension and support for those affected by it (Callinan 2010). The effectiveness of legislative efforts will also depend on successful enforcement of smoking bans and compliance with the legislation. Other tobacco control measures, such as taxation on tobacco products, limits on advertising and sponsorship, and limits on the sale of tobacco products, may vary between jurisdictions. A comprehensive approach to tobacco control will utilize both individual and population‐based intervention strategies, causing difficulties in evaluating the effect of a single intervention such as the smoking ban legislation.

Authors' conclusions

What's new

Acknowledgements

Appendices

Notes

New search for studies and content updated (no change to conclusions)

Data and analyses

Characteristics of studies