Thursday, May 28, 2009

New Study Concludes that Enforcement of Tobacco Sales Laws Reduces Smoking Among Adolescents: But How Good is the Science to Support its Conclusion?

A new study published in the journal BMC Public Health concludes that the enforcement of tobacco sales laws (which prohibit the sale of cigarettes to minors) leads to a reduction in youth smoking.

The study concludes as follows: "This analysis of national data covering 7 years indicates that improvements in merchant compliance that were measured in the U.S. from 1997–2003 as states complied with the Synar Amendment predicted a 20.8% reduction in daily smoking among 10th graders in 2003. ... This is the first study to document that the Synar Amendment achieved its intended effect of reducing adolescent smoking. The current study adds to a substantial body of evidence that demonstrates that improvements in merchant compliance are associated with reduced daily smoking among adolescents. This is the first study to demonstrate an effect of enforcement of tobacco sales laws on a national level. This is important because of concerns that only isolated rural communities could be successful in this regard. This U.S. national study demonstrates that this approach can be effective across an area encompassing wide geographical and demographic variety."

A press release that accompanies the study states: "Our data suggest that a 25 percent increase in compliance of laws prohibiting cigarette sales to minors has about the same deterrent effect as increasing the price by $2.00."

The Rest of the Story

The first thing that should make the reader very suspicious about the validity of the study's conclusions as well as the objectivity of the study is the fact that the paper ignores the existing published research, including a large meta-analysis of numerous studies, which fails to find any relationship whatsoever between merchant compliance rates and youth smoking. That meta-analysis is not even referenced in the article.

The meta-analysis concluded: "Based on data from 9 studies, there was no detectable relationship between the level of merchant compliance and 30-day (r = .116; n = 38 communities) or regular (r = .017) smoking prevalence. There was no evidence of a threshold effect. There was no evidence that an increase in compliance with youth access restrictions was associated with a decrease in 30-day (r = .294; n = 18 communities) or regular (r = .274) smoking prevalence. There was no significant difference in youth smoking in communities with youth access interventions compared with control communities; the pooled estimate of the effect of intervention on 30-day prevalence was -1.5% (95% confidence interval: -6.0% to +2.9%). Conclusions. Given the limited resources available for tobacco control, as well as the expense of conducting youth access programs, tobacco control advocates should abandon this strategy and devote the limited resources that are available for tobacco control toward other interventions with proven effectiveness."

It is quite odd that this meta-analysis would be completely ignored and very curious that it is not even referenced in the paper. Any article with any objectivity dealing with this issue would surely reference the meta-analysis and its findings to have any credibility whatsoever on the issue of the existing literature.

The second thing that should catch the observant reader's eye is the fact that this conclusion of a causal effect comes from a purely correlational study. The study is a cross-sectional one. It correlates youth smoking rates at a single point in time with the rates of compliance with tobacco sales laws. The study does not document that tobacco sales laws have produced any change in youth smoking rates.

This is a critical flaw in the study, and while it does not render the study useless, it does mean that the study fails to demonstrate that the tobacco sales law enforcement is the cause of the reduced smoking rates. A very plausible, and likely, alternative explanation for the observed findings is that communities with lower levels of youth smoking -- probably due to greater anti-smoking sentiment -- were more likely to vigorously enforce the tobacco sales laws.

We know that communities with stronger anti-smoking sentiment are more likely both to have lower youth smoking rates and to engage in interventions to reduce smoking, such as enforcing youth access laws. This is the classic problem of a confounding variable which is related to both the predictor and the outcome variable and which serves as an alternative explanation for the observed relationship between the predictor and outcome variables.

The study does not measure the states' anti-smoking sentiment, smoking prevalence, or any other state-level measures that might give some indication as to these factors, such as education levels or income. In addition, the study does not even measure the presence or absence of state and local restaurant smoking bans, which might have served as at least a proxy for the level of anti-smoking sentiment in these states.

Especially when viewed in light of the existing evidence, this study is non-convincing. It is clearly unable to demonstrate a causal relationship between enforcement of youth access laws and reduced smoking among adolescents. However, the study is a good example of the pitfalls of assuming that correlation equals causation.

No comments: