Authors of a widely reported article on the health impacts of a local ban on smoking in restaurants claim their research shows significant health benefits, but the data examined in the article show no health benefits at all.
Conflicts of Interest
Richard D. Hurt, M.D., and colleagues published an article in Archives of Internal Medicine on the health effects of a smoking ban in restaurants (2002) and bars (2007) in Olmsted County, Minnesota, home of the Mayo Clinic in Rochester. The research was presented in 2011 at a meeting of the American Heart Association and published in Archives of Internal Medicine.
The authors claimed no conflicts of interest, but their study was financed by a smoking ban advocacy group ClearWay of Minnesota, and government smoking ban entities, the National Heart, Lung, and Blood Institute; and the National Institute on Aging. It is hard to imagine the authors receiving future government money to conduct studies if they didn’t report health benefits for the bans already enacted.
Methods and Results
The authors claimed their study showed a benefit from a 2002 smoking ban for restaurants and a 2007 smoking ban for bars. They studied rates of myocardial infarction (also known as MI’s or heart attacks) for 18 months before and after the bans. They also studied rates of what they called sudden cardiac death, using a “soft” definition, obtained from death certificates without autopsy in the majority of cases.
The Patient Profiles from Table 1 in the article show:
MI’s occurred in a patient population mean age 67, 67 percent hypertensive, 22 percent diabetic, and 25 percent smokers.
The sudden death group were mean age 77, 73 percent hypertensive, 24 percent diabetic, and 15 percent smokers.
Table 2 is quite telling and includes their results on rates of MI and sudden death.
Table 2. Incidence Rates and Relative Risks of MI and SCD 18 Months Before and After Implementation of Smoke-Free Laws
Characteristic No. Rate per 100 000 (95% CI) a No. Rate per 100 000 (95% CI) a Adjusted RR, (95% CI) a P ValueMI Ordinance 1 187 150.8 (129.0-172.6) 185 144.6 (123.6-165.5) 0.96 (0.78-1.18) .71Ordinance 2 206 152.3 (131.4-173.3) 139 100.7 (83.8-117.5) 0.66 (0.53-0.82) <.001Before Ordinance 1 vs after Ordinance 2 187 150.8 (129.0-172.6) 139 100.7 (83.8-117.5) 0.67 (0.53-0.83) <.001SCD Ordinance 1 143 109.1 (91.0-127.2) 148 112.7 (94.3-131.0) 1.01 (0.80-1.27) .96Ordinance 2 111 78.8 (64.0-93.5) 133 b 92.0 (75.7-108.3) 1.17 (0.91-1.51) .22Before Ordinance 1 vs after Ordinance 2 143 109.1 (91.0-127.2) 133 92.0 (75.7-108.3) 0.83 (0.65-1.06) .13Abbreviations: MI, myocardial infarction; RR, relative risk; SCD, sudden cardiac death.a Adjusted for age and sex.b Cause of death was missing for 3.7% of out-of-hospital deaths. The number reported herein represents the estimated number of SCDs obtained via multiple imputation for missing data.
Data ‘Clearly Unreliable’
There are important factors that jump out in Table 2 and the paper as a whole:
1. The Hurt paper is an analysis of data, not a toxicology study. The authors have no information other than hospital records and death certificates, so they cannot describe the exposures to cigarette smoke pre- and post-ban for the individuals who died or the individuals who had a myocardial infraction.
2. The basic data are clearly unreliable. The data have adjusted relative risks (RRs) less than 1 for all the acute myocardial infarction results. The data have a 1.01 RR post restaurant ban on sudden death, which has a p value (statistical probability that an event occurred by chance alone) of .96 that makes the result unreliable. The bar ban produced an association with sudden death of 1.17, but again the p value, 0.22, shows the result as unreliable. In all of the results, the Confidence Interval included 1, so the results all had a range of error or accuracy that included RR of 1 which means no effect is in the range or error or confidence. This study should have never been published. It proves nothing at all.
3. Despite the authors’ claims, there are no studies that demonstrate credible and plausible explanations for an assertion that inhaled secondhand smoke causes death acutely or MI. The studies cited by the authors all include the same methodological and statistical deceptions that taint this study, including small associations rather than evidence of causation, and in many cases containing confidence intervals including 1. This systematic deceit is the product of a very well-financed and powerful crusade in which anti-smoking activists justify the use of flawed analyses by the asserted noble result of restricting smoking.
4. The authors propose the smoking ban reduced deaths and myocardial infarctions because the ban made smoking inconvenient and less frequent, although the authors present no evidence of this effect. The authors cannot know who was smoking or not smoking at home or in the car, and what effect a smoking ban might have on smoking habits. In addition, the myocardial infarctions and sudden deaths tended to occur in people past age 60, and the smoking rates were 25 percent and 15 percent, respectively. This invites the question, what should we blame for the nonsmoker deaths and MIs, and isn’t that potentially a predominant factor and a confounder when a study attempts to attribute rate changes to smoking bans?
Better Evidence Elsewhere
A more comprehensive and reliable study on smoking bans, national in scope for 15 years, was published in 2011 by the National Bureau of Economic Research and the Rand Corporation.