Monday, November 28, 2011

Mayo Clinic Researchers Deceiving the Public by Implying that 45% Reduction in Heart Attacks in Olmsted County Is Attributable to County Smoking Ban

Researchers from the Mayo Clinic have reported finding a 45% reduction in the rate of admissions for heart attacks in Olmsted County, Minnesota from the 18 months prior to a workplace smoking ban implemented in 2002 to 18 months after a strengthened ban implemented in 2007.

These results are being presented to the media in a way that is implying that the observed 45% reduction in heart attack admissions is attributable to the county smoking bans.

For example, the headline of an ABC News article reads: "Smoking ban cuts cardiac events 45%, Mayo Clinic says."

A Procor headline reads: "Smoking ban cuts heart attacks in half."

A Thirdage.com article headline reads: "Smoking Ban Cuts Heart Attack Risk In Half."

The UPI headline about the research reads: "Smoking ban cut heart attacks risk in half."

The Minneapolis Star-Tribune headline reads: "Smoking bans cut cardiac events 45%."

The EMax Health headline is: "Smoking bans reduce heart attack rates by half, finds study."

The Business Insider headline reads: "Heart Attacks Decreased By 50% After These Workplaces Launched Smoking Bans."

These headlines appear to be consistent with the information being provided to reporters by the study investigators.

As a KAAL-TV article explains: "A new study from Mayo Clinic shows people who live in Olmsted County are nearly 50 percent less likely to suffer a heart attack than they were a decade ago. The study's main author claims the dramatic change is due entirely to the county's smoke-free laws. "We were absolutely shocked..." said Richard Hurt, M.D. Dr. Richard Hurt, the director of Mayo Clinic's Nicotine Dependence Center, is talking about the results of a recent second hand smoke study conducted in Olmsted County. In 2002, the county enacted a smoking ban in restaurants and expanded that to include work places in 2007. Since then, the study shows that heart attacks have dropped 45 percent. And sudden cardiac deaths have dropped 50 percent. Dr. Hurt says despite the numbers, all the other factors that cause heart attacks either stayed the same or got worse. 'So high cholesterol stayed the same, hypertension stayed the same, diabetes went up, obesity went up. So all the other risk factors for heart disease did not change,' he explained. 'So we don't have any explanation other than second hand smoke.'"

These findings were presented last week at the American Heart Association Scientific Sessions in Orlando, Florida. So far as I can tell, there is no actual study or report available for public review. Thus, the precise methodology used is not clear.

The Rest of the Story

Today's story illustrates the dangers of drawing conclusions from quasi-experimental studies without including a comparison group.

A quasi-experimental study is one in which investigators observe changes in an outcome variable (in this case, heart attacks) following the implementation of an intervention or " natural experiment" (in this case, the Olmsted County smoking ban). Because one never knows what would have occurred in the absence of the intervention, it is imperative in a quasi-experimental study design to include a comparison group to control for underlying secular changes taking place in the outcome variable over time in the absence of the intervention.

Unfortunately, it does not appear that the Olmsted County "study" employed any comparison group or attempted in any way to compare the 45% reduction in heart attack admissions in Olmsted County to the reduction in heart attack admissions that theoretically might have been occurring in the rest of the state of Minnesota.

Instead, the "study," the press release, and the researchers appear to have completely ignored secular changes in heart attacks occurring in Minnesota during the study period. From none of these sources is it revealed that heart attacks throughout the entire state of Minnesota also might have been declining during the study period, so that the 45% decline observed in Olmsted County might not be entirely attributable to the smoking ban.

Had the researchers examined the change in the heart attack admissions rate in Minnesota as a whole during the study period (2001 to 2008), which of the following do you think they would have found, according to the newspaper headlines:

A. No change in the heart attack admissions rate from 2001 to 2008.
B. A decrease of 3% in the heart attack admissions rate.
C. A decrease of 8% in the heart attack admissions rate.
D. A decrease of 14% in the heart attack admissions rate.
E. A decrease of 33% in the heart attack admissions rate.

According to the news articles, and according to the investigator's quote in the KAAL-TV article, one would have to assume that there was no change in the heart attack admissions rate for the state of Minnesota from 2001 to 2008 since the study and its presentation to the media and the public essentially attribute the entire 45% reduction in the heart attack admission rate to the smoking ban.


The truth, however, is that based on an analysis of data from the Healthcare Cost and Utilization Project (HCUP), the correct answer is:

E. A decrease of 33% in the heart attack admissions rate.

Therefore, even in the absence of any smoking ban, one would have expected the heart attack admission rate in Olmsted County to have declined by 33% from 2001 to 2008, simply based on the trend occurring throughout the state of Minnesota.

In light of the expectation of a 33% decline without a smoking ban, the finding of a 45% decline now doesn't seem nearly as dramatic, does it?

In fact, smoking prevalence in Olmsted County dropped sharply, by 23%, during the study period. It is conceivable that the additional 12% decline in heart attacks over the expected 33% decline was mostly attributable to the sharp decrease in smoking rates. Thus, it is entirely possible that the smoking ban - and specifically the decrease in secondhand smoke exposure - had very little impact on heart attack rates during the study period.

To answer that question would require a more rigorous analysis that adjusted not only for secular trends in heart attacks in Minnesota as a whole but also for changes in smoking prevalence that may not have been a result of the smoking ban.

Based on the only data presented in this "study" - the change in heart attack admissions rates in Olmsted County alone - one is unable to draw any conclusion about the impact of the smoking ban and secondhand smoke exposure on heart attacks, much less to be able to conclude that the observed 45% reduction in heart attacks was attributable to the smoking ban.

An honest press release would have revealed that while heart attack admissions in Olmsted County declined by 45% from 2001 to 2008, heart attack admission rates in the state of Minnesota as a whole declined by 33%.

Similarly, an honest presentation at the AHA conference would have disclosed that the 45% decline in heart attacks in Olmsted County must be interpreted in light of a 33% reduction in heart attacks in the state as a whole.

The rest of the story is that I believe the Mayo Clinic investigators have deceived the media and the public by failing to disclose the striking decrease in heart attack admission rates in Minnesota as a whole during the study period. It completely changes the public's and the media's interpretation of the meaning of the observed 45% decline in Olmsted County to know that heart attacks declined by 33% in the state of Minnesota as a whole during the same time period.

I want to make it clear to readers that I agree with the investigators' overall conclusions about smoking bans. I support these policies and I agree with their recommendation that people with heart disease avoid exposure to secondhand smoke. However, I do think that we need to support smoke-free policies with rigorous scientific evidence. Drawing causal conclusions from a quasi-experimental study without any comparison group is not, in my opinion, an example of rigorous scientific evidence, especially in a situation like this where the outcome variable is well known to be declining dramatically throughout the population.

No comments: