Tuesday, January 29, 2013

New Study Reports Decline in Heart Attacks After Smoking Ban, But There Was No Decline in Heart Attacks After Smoking Ban

A new study published last week in PLoS ONE reports that a partial smoking ban enacted in Spain in 2006 resulted in an 11% decline in heart attack incidence and an 18% decline in heart attack mortality rates in six counties in northern Spain.

(See: Aguero F, et al. Impact of a partial smoke-free legislation on myocardial infarction incidence, mortality and case fatality in a population-based registry: the REGICOR study. PLoS ONE 2013; 8(1): e53722. doi:10.1371/journal.pone.0053722.)

The study compared the incidence and mortality rates for acute myocardial infarction for the four years prior to the partial smoking ban (2002-2005) with those during the first three years after implementation of the law (2006-2008). The law banned smoking in workplaces, but exempted hospitality establishments including bars and restaurants. The data were collected from a population-based heart attack registry, covering six counties in northeast Spain served by a single tertiary care hospital with a coronary care unit.

The paper used a negative binomial regression to model the annual rates of heart attack incidence or mortality, with an indicator variable corresponding to whether the data were from before or after the partial smoking ban.

The results were reported as follows: "In the post-ban period, AMI incidence and mortality rates significantly decreased (relative risk [RR] = 0.89; 95% confidence interval [CI] = 0.81–0.97 and RR = 0.82; 95% CI = 0.71–0.94, respectively)."

The conclusion was: "The 2006 Spanish partial smoke-free legislation was associated with a decrease in population AMI incidence and mortality, particularly in women, in people aged 65–74 years, and in passive smokers. These results clarify the association between AMI mortality and the enactment of a partial smoke-free legislation and reinforce the effectiveness of smoking regulations in preventing CHD."

The results were disseminated widely through the media. The press release boasted that: "Implementation of smoke-free legislation reduces the number of acute myocardial infarctions by 11 percent."

The Rest of the Story

There's just one problem with this study concluding that the partial smoking ban implemented in 2006 reduced the number of heart attacks by 11% (and the number of heart attack deaths by 18%):

The data show that there was no decrease in heart attacks or heart attack deaths following the smoking ban.

One can easily see this in Figure 1 of the paper. There was indeed a decline in heart attack incidence and mortality in northeast Spain during the study period, but the decline occurred prior to the smoking ban, not after it. There was in fact little or no change in heart attack incidence or mortality after implementation of the smoking ban.

The data speak for themselves:

Here are the heart attack incidence rates for 2005 (immediately before the ban) and 2008 (after the ban):

Before: 173.35
After: 175.15

Here are the heart attack hospitalization rates for 2005 (immediately before the ban) and 2008 (after the ban):

Before: 144.90
After: 142.71

Here are the heart attack mortality rates for 2005 (immediately before the ban) and 2008 (after the ban):

Before: 37.58
After: 39.57

As you can easily see from the data reported in the paper, heart attack incidence and mortality rates were almost the same after the partial smoking ban as they were immediately prior to the ban.

What's quite clear is that the heart attack rates were dropping significantly during the baseline period, from 2002-2005. From 2005-2008, they remained relatively steady. Therefore, a negative binomial regression is going to "pick up" an "effect" of the smoking ban, that is not at all due to the smoking ban. Instead, this detection of an 11% "decline" in heart attacks is simply a reflection of the secular decline that occurred during the baseline period. If you just use data from 2005 for the pre-ban period, this effect will completely disappear!

This is hardly surprising, since the ban was hardly a "ban." It simply eliminated smoking in office-type workplaces, but it didn't get rid of smoking in bars or restaurants. Since most people were already working in smoke-free offices in 2005 anyway, and because the law didn't apply to bars and restaurants, one would not have expected to observe any decline in heart attacks to begin with. I find it surprising that the study authors actually expected to see such an effect.

But the most significant aspect of this story is that the paper is reporting a conclusion based on data which demonstrate the opposite. The paper reports that heart attacks declined after the smoking ban, but the data show that heart attacks did not decline after the smoking ban.

This is yet another study on the acute cardiovascular effects of smoking bans that reports exactly the opposite of what the study data demonstrate. It appears that anti-smoking researchers are so determined to find an effect of smoking bans on heart attacks that they will manufacture conclusions that simply are not consistent with the reported data.

In most of the public's eyes, a change from 173 to 175 is not an 11% decline, but the minds of anti-smoking researchers, it is. After all, we're working for a good purpose, and so it's perfectly acceptable to manipulate the data with statistical models so as to find an effect that doesn't really exist.

No comments: