Tuesday 14 February 2017

The smoking ban miracle hoax

Jacob Grier has written an excellent article for Slate about the junk science used to justify smoking bans. In particular, he looks at the ludicrous heart attack miracles than supposedly take place whenever people are prevented from smoking in pubs.

When studies sampling larger populations finally appeared, the reported declines in heart attacks began to shrink. A study of the Piedmont region of Italy found a much lower decline of 11 percent, though curiously only for residents under 60 years of age. England, which implemented a smoking ban nationwide, presented the first opportunity to study the matter on a national scale. Researchers there credited the ban with a heart attack reduction of just over 2 percent nationwide.

Critics noted that the rate of heart attacks in England had also been falling in the years prior to the ban and that the reason for the decline was still not clear. Regardless, the data there made it obvious that the miraculous reductions claimed in smaller studies were unrealistically high. Even so, despite acknowledging the wide variation in findings and the admitted methodological limitations of the studies, a 2009 meta-analysis conducted by the Institute of Medicine concluded that the impact of smoking bans on short-term heart attack rates was real and substantial: “Even a small amount of exposure to secondhand smoke… can cause a heart attack,” one member of the IOM panel informed the New York Times, urging that “smoking bans need to be put in place as quickly as possible.”
This report had, however, omitted one of the largest studies of secondhand smoke and heart attacks conducted to date. A 2008 study covering the entire country of New Zealand—a population smaller than England’s, but bigger than the American towns previously studied—found no significant effects on heart attacks or unstable angina in the year following implementation of a smoking ban; hospitalizations for the former had actually increased.

Contradictory research continued to come in. A clever study led by researchers at RAND Corp. in 2010 tested the possibility that the large reductions identified in small communities were due to chance. They assembled a massive data set that allowed them to essentially replicate studies like those in Helena, Pueblo, and Bowling Green, but on an unprecedented scale. Whereas those studies had compared just one small community to another, the RAND paper compared all possible pairings of communities affected by smoking bans to all possible controls, for a total of more than 15,000 pairings. They stratified results by age in case there were differential effects on the young, working age adults, or the elderly. And in an improvement on most other studies, they also controlled for existing trends in the rate of heart attacks.

The study found no statistically significant decrease in heart attacks among any age group. The data also suggested that fluctuations in heart attack rates were common, indicating that comparisons of small communities would frequently turn up dramatic reductions due purely to chance; large increases in heart attacks happened about as often. This explained the headline-grabbing dramatic results in places like Helena or Monroe County that eluded replication in larger jurisdictions. The conclusion of the study was blunt: “We find no evidence that legislated U.S. smoking bans were associated with short-term reductions in hospital admissions for acute myocardial infarction or other diseases in the elderly, children or working age adults.”

A 2012 study of six American states that had instituted smoking bans came to a similar conclusion. So did a 2014 study, which is notable for being co-authored by some of the same researchers who had previously published papers suggesting that the Colorado towns of Pueblo and Greeley had experienced reduced rates of heart attacks after implementing smoking bans. When Colorado enacted a statewide ban, the authors had an opportunity to see if their earlier results could be duplicated across the larger population of nearly 5 million people. No effect appeared. As an additional test, they re-examined the data excluding 11 jurisdictions that had already implemented comprehensive smoking bans: The statewide ban still showed no effect.

In the paper’s admirably honest commentary, the authors reflected on the reasons that earlier studies, including their own, had overstated the impact of smoking bans. The first is that small sample sizes allowed random variances in data to be mistaken for real effects. The second is that most previous studies failed to account for existing downward trends in the rate of heart attacks. And the third is publication bias: Since no one believes that smoking bans increase heart attacks, few would bother submitting or publishing studies that show a positive correlation or null effect. Thus the published record is likely unintentionally biased toward showing a larger effect than truly exists.


Do read it all.

No comments: