★ Recent publications concern reductions in heart disease following smoking bans. ★ Meta-analysis estimates of the reduction have ranged from 10% to 19%. ★ These often have major weaknesses, ignoring control data and underlying trends. ★ We describe a consistent and valid approach to estimate effects. ★ Our meta-analysis with this approach gives a 2.7% (95% CI 2.1–3.4%) reduction.
Recent studies have compared rates of acute myocardial infarction before and after introducing smoking bans. Some meta-analyses report post-ban reductions up to 19%, implausibly large considering likely changes in smoking habits and passive smoke exposure. Our literature reassessment demonstrates major weaknesses in many studies and meta-analyses, including failure to consider data from control areas or existing trends in acute myocardial infarction rates, incorrect estimation of variability, and use in some meta-analyses of results for population subsets or estimates apparently unrelated to the data reported. We report meta-analyses using a consistent approach to derive estimates of the ban effect, taking account of time trends and control data, which indicate a much smaller reduction. Preferring national to regional estimates where available, we estimate a 5% reduction (95% CI 3–8%). Omitting estimates where trend adjustment was impossible, this becomes 2.7% (2.1–3.4%), consistent with reported declines of 2–3% in large national populations (England, France, Italy, USA). We discuss some limitations of these estimates. Further evidence is needed, possibly by analyzing national mortality data. Our findings highlight the need for a valid approach when estimating the effect of bans, and demonstrate major weaknesses in many previous publications.