Wednesday, November 30, 2005

New Study Casts Doubt on Claim that Smoking Bans Substantially Reduce Heart Attack Admissions

A new study released yesterday on the Smoker's Club, Inc. web site, questions the claim that smoking bans cause a 40% (as observed in Helena, Montana) or 27% (as observed in Pueblo, Colorado) drop in hospital admissions for myocardial infarctions (heart attacks).

The authors, David W. Kuneman and Michael J. McFadden, analyzed data on hospital admissions for acute myocardial infarction from the HCUP project (Healthcare Cost and Utilization Project), which is "a family of health care databases and related software tools and products developed through a Federal-State-Industry partnership and sponsored by the Agency for Healthcare Research and Quality. HCUP is based on statewide data collected by individual data organizations across the United States and provided to AHRQ through the HCUP partnership. ... HCUP data are used for research on hospital utilization, access, charges, quality and outcomes. ... Researchers and policymakers use HCUP data to identify, track, analyze and compare hospital statistics at the national, regional and State levels."

Specifically, the authors examined the total number of hospital admissions for acute myocardial infarction in the year prior to and after a smoking ban in each of four states which enacted some form of smoking ban in restaurants and/or bars during the period for which data from HCUP are available (1997-2003): California, New York, Florida, and Oregon.

For California, a complete ban on smoking in bars was implemented in January 1998. Heart attack admissions increased from 40,608 in the year preceding the bar smoking ban (1997) to 43,044 during the year following the smoking ban (1998), an increase of 6.0%.

For New York, a complete ban on smoking in bars and restaurants was implemented in July 2003. Heart attack admissions increased from 31,728 in the year preceding the bar smoking ban (2002) to 31,888 during the year in which the smoking ban was implemented (2003), an increase of 0.4%.

For Florida, a ban on smoking in restaurants (free-standing bars excluded) was implemented in July 2003. Heart attack admissions decreased from 40,077 in the year preceding the bar smoking ban (2002) to 39,783 during the year in which the smoking ban was implemented (2003), a decrease of 0.7%.

For Oregon, a ban on smoking in restaurants which allow children was implemented in July 2001. Heart attack admissions increased from 4,957 in the year preceding the ban (2000) to 5,125 in the year following the ban (2002), an increase of 0.4%, and there was almost no change in heart attack admissions during 2001 -- the year in which the ban was implemented (4,927, 0.1% decrease from 2000).

The authors point out that none of these findings provides any suggestion that the statewide smoking bans had any immediate and substantial effect on heart attack admissions.

The authors point out that while the total number of heart attacks studied in Helena and Pueblo totaled 315, the total number of heart attacks in this study was over 315,000, or 1,000 times higher. They suggest that this larger sample size as well as the examination of state-wide data rather than just data in isolated cities makes the conclusions from this study more stable than from the existing studies on this topic.

The paper concludes: "Statistically this larger population base makes for a far more stable statistical environment and the data from this population would provide a far sounder scientific basis for decisions about smoking bans that will affect the lives and livelihoods of millions of people."

The Rest of the Story

In addition to confirming Kuneman and McFadden's findings, I extended their analysis by:
  • examining trends going back in time as far as 1997, the earliest available online data (in order to have a more stable baseline period to establish secular patterns); and
  • examining trends in heart attack admissions in all the other states in the online database without smoking bans that included data for the entire study period 1997-2003 (a total of 8 states - New Jersey, South Carolina, Utah, Washington, Arizona, Colorado, Hawaii, and Iowa; Massachusetts was not included because of the extensive local smoking bans) (in order to have a comparison group).
For California, I compared heart attack admission trends during the period 1997-2002 for California versus the 8 non-ban states in the HCUP online database and versus the nation as a whole. Trend lines were essentially parallel throughout the period. From 1997 to 1998, heart attack admissions in California increased by 6.0%, compared to a 3.8% increase in the comparison states and a 6.2% increase in the remainder of the nation. From 1997 to 1999, heart attack admissions increased by 9.9% in California, compared to 4.8% in the comparison states and 4.3% in the remainder of the nation.

For New York, overall trends were similar to those in the comparison states and to the nation as a whole, except that New York did not experience the slight decline in heart attack admissions during 2003 that was observed elsewhere. In New York, admissions for heart attacks increased by 0.4% from 2002 to 2003, while heart attacks decreased by 3.1% in the comparison states and by 2.8% nationally during the same time period.

For Florida, heart attack admissions increased slightly faster than in the comparison states between 1997 and 2000, but the patterns were similar from 2000-2003. There was a slight decrease in heart attacks between 2002 and 2003 in Florida (0.7%), the comparison states (3.1%), and the nation as a whole (2.8%).

For Oregon, there was a 0.4% increase in heart attack admissions from 2000 to 2002, while admissions in the comparison states dropped by 0.7% during the same period, and admissions nationally increased by 4.3%.

Commentary and Conclusions

I think Kuneman and McFadden are to be congratulated for having made an important contribution to the analysis of this research question. I think that their analysis, along with my extension of that analysis, provides compelling evidence that brings into question the conclusion that smoking bans have an immediate and drastic effect on heart attack incidence.

In fact, these analyses demonstrate that on a state-wide level, there is no suggestion of any large-scale effect on heart attack admissions associated with the implementation of statewide bans on smoking in child-friendly restaurants, all restaurants, bars, or bars and restaurants.

If there were a true 27% or 40% decrease in heart attack admissions due to smoking bans that occurred almost immediately (within six months, as claimed), one would have expected to see a demonstrable decline in such admissions in states that implemented such bans.

An effect of such smoking bans can certainly not be ruled out, especially because the 2004 data for New York and Florida are not yet available (so only the first six months post-ban could be examined). However, it does seem quite clear that if there is an effect, it is not nearly as immediate or as dramatic as suggested in press releases. (see also Pueblo release and Bowling Green press release and Greeley news article)

My honest appraisal of what is going on here is that anti-smoking groups have been too quick to go to the media with definitive claims of a drastic and immediate effect of smoking bans on heart attacks when the scientific evidence is simply not sufficient to support such claims. What is happening, I believe, is that the anti-smoking agenda is driving the interpretation of the science. As I stated before, it is an agenda which, in this case, I wholeheartedly support (I have been lobbying for workplace smoking bans, especially those in bars and restaurants for 21 years). However, I don't think the importance of the ultimate objective justifies the use of shoddy science to support that objective.

At this point, I must make 3 critical points:

First, I am not suggesting that there was anything wrong with the studies that were done in Helena and Pueblo or that the authors did anything wrong in stating their conclusions within the Helena paper. What I am suggesting, instead, is that drawing definitive, generalized conclusions based on these two small studies, and sending out press releases to the media with these definitive conclusions (before the Pueblo study has even been published) is irresponsible and undermines the scientific credibility of the tobacco control movement.

Within themselves, it may be that the Helena and Pueblo studies are quite solid (I have argued not with respect to the Pueblo study, but there is room for differing interpretations of the evidence); however, that doesn't mean that the evidence is sufficient to support a general conclusion that smoking bans reduce heart attacks by 27-40%. The fact that population-wide studies with much larger sample sizes do not seem to bear out these findings is exactly the reason why one has to be careful in concluding an effect with a small and select sample (and especially, in the face of huge random variations in secular trends in a small geographical area).

Second, I am not suggesting that this is a reason not to support smoke-free restaurant and bar laws. In fact, one of the things that I think tobacco control groups have been doing wrong is using data such as this to support such ordinances. I think the reason for these laws is that secondhand smoke is a substantial workplace hazard for bar and restaurant workers. That's it. Whether the laws end up reducing heart attacks (probably by virtue of smokers quitting or cutting down) or not is not relevant in my mind to the issue of whether we should protect workers from a substantial and preventable occupational hazard.

I think by harping on these data, anti-smoking groups have set themselves up for failure, and therefore done a disservice to the overall effort to protect workers from secondhand smoke. Because now that valid scientific doubt is being cast on this exaggerated claim, it may well hurt the effort to protect these workers.

This is what I meant when I suggested that the credibility of the movement is being threatened by the tactics being used. If the focus of the debate shifts to the validity of the heart attack reduction claim rather than the need to protect workers from a severe and preventable occupational hazard, then we may well lose the debate. I fear this is now going to happen now that the "cat is out of the bag."

Third, and finally, I am not concluding here that smoking bans do not reduce heart attacks. I am not even concluding that smoking bans did not reduce heart attacks in Helena or Pueblo. What I am concluding is that the overall evidence does not support the conclusion that the observed declines in heart attack admissions in Helena or Pueblo (or Bowling Green or Greeley) are in fact: (1) real, rather than simply chance variations; (2) attributable to the smoking ban, rather than some other factor; and (3) widely generalizable to other communities.

It is possible, for example, that local smoking bans may have an effect that state-wide smoking bans do not have. Perhaps all the local media attention focuses public attention on the matter and results in publicity that ends up changing smoking behavior. And perhaps that doesn't happen as effectively at a state level. But I think a lot more research is needed before we can conclude that the reason why we don't observe a substantial reduction in heart attack admissions associated with smoking bans on a state-level is that the effect only holds at a local level.

Moreover, I would point out that in my analysis of trends in heart attacks in Massachusetts, where there was a huge proliferation of smoke-free bar and restaurant regulations between 2000 and 2003, I found that heart attack admissions increased in Massachusetts by 31.8% during this time period, compared to a 2.4% decline in the comparison states, and a 1.5% increase nationally.

In short, what I am concluding is that it is far too premature to conclude that smoking bans reduce heart attacks substantially and in a short period of time. And that as much as anti-smoking groups are doing important work by promoting smoke-free bar and restaurant laws, it simply is not responsible (nor wise strategically, I think) to be using shoddy science to support this cause. In the long run, it is our credibility (and ultimately therefore, the health of the public) that is going to lose out.

No comments: