Smoking Bans and Heart Attacks: Pueblo Edition

Jacob Sullum | November 16, 2005

Researchers in Colorado claim to have confirmed the amazing power of smoking bans to reduce heart attacks, a phenomenon first invented discovered in Helena, Montana, in 2003. You may recall that two local physicians, aided by anti-smoking activist Stanton Glantz, reported that Helena's ban was followed by an immediate 60 percent reduction in heart attacks, a claim later downgraded to 40 percent. Now we're told that the heart attack rate in Pueblo, Colorado, fell by 27 percent within 18 months after the city banned smoking in "public places" a couple years ago.

In its press release about the unpublished study, the Pueblo City-County Health Department conflates correlation with causation, saying the data from Helena "showed restrictions on public exposure to secondhand smoke caused a sharp decline in heart attacks." If this had already been demonstrated, why do any more research? In fact, the Helena data indicated only that the ban was followed by a drop in heart attacks, not that the former caused the latter.

The press release also obscures the distinction between reducing secondhand smoke exposure (the ostensible aim of smoking bans) and pressuring smokers to quit (the real aim). It notes that "the study didn't distinguish between smokers and nonsmokers, but rather represented a combination of both smokers and those impacted by secondhand smoke." In other words, some, most, or all of the drop in heart attacks could have occurred among smokers driven to quit by the ban. Yet the press release quotes one researcher who says "this study further validates the argument that limiting exposure to deadly tobacco smoke can save lives" and another who claims "this study provides important knowledge that people can be healthier if secondhand smoke is removed from public places."

As anti-smoking activist (and smoking ban supporter) Michael Siegel points out on his tobacco policy blog, the study does nothing of the kind. Siegel notes that it's quite implausible to suggest that secondhand smoke causes three out of 10 heart attacks. (Even the American Heart Association, which sponsored the conference where the Pueblo data were presented, attributes only 5 percent of heart disease deaths to secondhand smoke.) Siegel suggests the drop in heart attacks was more likely due to a decline in the number of smokers. Yet studies of smokers who quit indicate that their heart attack risk does not fall sharply enough to account for the 27 percent drop ascribed to Pueblo's smoking ban, even if the law caused every smoker in town to quit (which it presumably did not).

In any case, it's not at all clear there's a phenomenon here that needs explaining. Hundreds of cities in the U.S. are covered by state or local smoking bans. I imagine the heart attack rate declined in more than a few after the bans took effect, went up in others, and stayed about the same in most. That's the pattern you'd expect purely by chance. Identifying cities where heart attacks declined proves nothing. If these laws have the sort of impact people like Stanton Glantz are suggesting, there should have been a noticeable post-ban drop pretty much everywhere with similar restrictions.

[Thanks to Linda Stewart for the tip.]

Comments

fyodor|11.16.05 @ 1:47PM|#

Okay, being a devil's advocate here...

Jacob, are you saying that the effects of smoking bans have been studied elsewhere and Helena and Pueblo are the only two that showed a heart attack decrease? If so, the pro-ban folks are doing some intense cherry picking. But my impression was that these were the first two towns to be studied for these efffects. If that's correct, you may still be right that nothing it proven yet as ban advocates are claiming. But even if their claims are overstated, as long as the studies' methodology are sound, a reasonable person would have to find these (ahem) correlations rather interesting.

|11.16.05 @ 1:52PM|#

Jacob, As the top notch reporter you are, maybe you could find the city where a ban increased heart attacks. And the sooner the better, there's not many cities left here in Louisiana that respect property rights.

|11.16.05 @ 1:56PM|#

IIRC, the problems with the study in MT included very small sample size and no effort to control for, well, anything. No care in how comparisons were conducted (compare same time of year, for instance, look at demographic breakdowns of heart attack patients), and no attempt to see whether heart attacks dropped most dramatically in people whose smoke exposure had dropped after the ban.

In other words, it was in the straw man category that people on H&R like to refer to when they remind us that correlations are (allegedly) useless. Correlations are only as useful as the underlying study is thorough. If I'm remembering the details (or lack thereof) in the MT study correctly, then the numbers were useless.

And yes, I know, even if they did all that, and the sample size was really large, correlations prove nothing. I know, I know. Correlations can only be suggestive, never conclusive.

keith|11.16.05 @ 1:59PM|#

I can't wait until we abandon the entire concept of hard science and let Las Vegas illusionists tell us what to do.

|11.16.05 @ 2:02PM|#

Fyodor, I'm not aware of any published studies presenting heart attack data for other cities with smoking bans. But that in itself is a bit suspicious: With an effect as dramatic as the one claimed by Glantz et al., wouldn't you expect evidence all over the place, especially in big cities such as Los Angeles, San Francisco, New York, and Boston? We have no way of knowing how many times supporters of smoking bans have looked at such data and failed to find a drop.

fyodor|11.16.05 @ 2:10PM|#

Thanks thoreau and Jacob.

Jacob, I know I should RTFA, but lacking time to do that at the moment, am I to understand that the Pueblo "study" was merely the result of anti-smoking activists "looking at" the data and then pointing out their conclusion to the health department officials who then publicized the apparent findings to the press? If this information was prompted by anti-smoking activists who had the choice of whether to say anything about it rather than reputable researchers who would have pubicized their findings either way, then your point is valid.

|11.16.05 @ 2:24PM|#

The study didn't distinguish between smokers and nonsmokers, but rather represented a combination of both smokers and those impacted by secondhand smoke.

Clear as ... smoke.

|11.16.05 @ 2:32PM|#

It's also possible that the reduced business at Pueblo bars and restaurants caused some customers to drive to another town and had their heart attacks in other locales instead of Pueblo. I'm sure the number of heart attacks in New Orleans has dropped severely over the last couple months.

|11.16.05 @ 2:34PM|#

18 months seems like an awfully short time to conclude anything about the health effects of a smoking ban, especially when it refers to heart attacks which are are events normally precipitated by many years of unhealthy behavior.

You'd be hard pressed to convinve me that 1 of every 4 heart attacks per year could be stopped during that time by removing one variable after a lifetime of unhealthy living.

|11.16.05 @ 2:43PM|#

I can't wait until we abandon the entire concept of hard science and let Las Vegas illusionists tell us what to do.

Huh...? I think I get what you're talking about, but could be be a little less vague for the folks at home?

fyodor|11.16.05 @ 3:26PM|#

Russ D,

As a Denverite I can attest that there's not a whole lot right around Pueblo. With the possible exception of the one suburb of Pueblo West, the closest town of any size is 30 to 40 miles away.

|11.16.05 @ 3:31PM|#

Fyodor, I don't know much about the backgrounds of the people who did the Pueblo study. One of them is the director of the local health department and presumably supports the smoking ban; the rest seem to be local doctors. In Helena, the two local doctors who did the study were both involved in the campaign for the smoking ban.

In any case, I'm not sure that even "reputable researchers" would go to the trouble of writing up and publishing negative findings (or that they could find a receptive journal if they tried to do so). This is one of the controversies regarding the broader secondhand smoke literature: whether publication bias accounts for some or all of the observed correlation between exposure to secondhand smoke (generally measured by the proxy of living with a smoker) and chronic diseases such as lung cancer.

|11.16.05 @ 3:41PM|#

Jacob-

You have a good point, in that "nothing to see here, move along" is hardly the best abstract for a research article. My guess is that such a paper could get published if somebody else has previously published positive findings, and the new study reaches the opposite conclusion with more rigorous methods and a larger sample size.

But that makes for less interesting news footage. It may be inevitable that the first publication on a topic will always be the one that purports to find a hazard.

fyodor|11.16.05 @ 3:46PM|#

Jacob,

Interesting.

Y'know, I often measure the value of an argument by how good I would feel about using it against someone who disagreed, and I'm not sure about this. There may not be any journal that want to print results that found no correlation? Why not, would it be too boring? Are there no smoking ban critics who would have the resources to study LA, SF, Boston or NY to show that they're different? Are these studies done in secret such that there's no one to tell the world that anti-smoking activists did a study but didn't tell anyone the results? Is publication bias a recognized problem in other research endeavors? Has anyone shown a seemingly demonstrated correlation was only the result of publication bias?

Thanks for the response, BTW.

|11.16.05 @ 3:49PM|#

This just in:

• Summer causes ice cream cones
  
• Winter causes Christmas
  
• The SuperBowl causes Valentine's Day
  

These factoids brought to you by the bumblefuck department of public resources.

Larry A|11.16.05 @ 3:58PM|#

After the law went into effect I wonder how many heart attack victims died before they made it to the hospital because they were outside smoking and no one saw them go down?

|11.16.05 @ 4:02PM|#

Not that this means much, but in went searching for heart attack data and found that heart attacks occur more often in winter.

http://in.news.yahoo.com/031126/139/29tef.html

Now, for the 2 18-month spans used in the Pueblo claims, the first one (before the ban) contained 2 winter seasons and the second period (after the ban) contained only one winter season. So you would likely find the same results even if no smoking ban were in effect.

|11.16.05 @ 4:05PM|#

Russ D-

Excellent rebuttal.

keith|11.16.05 @ 4:49PM|#

Huh...? I think I get what you're talking about, but could be be a little less vague for the folks at home?

meaning that when something so obviously relying on smoke and mirrors to deceitfully make a point is passed off and accepted as hard scientific research, we might as well abandon any vestige of hope that there will ever be honestly presented "facts," and just turn the show over to someone who will, at very least, present it to us while wearing a jaunty sequined tux.

fyodor|11.16.05 @ 5:33PM|#

Russ D,

That's very interesting! Is the winter effect large enough so that two winters instead of one could explain the 27 percent difference? (Or close enough?)

kmw,

Others have provided good rebuttals, but yours is inane. If decreases in heart attacks really did follow smoking bans as consistently as Valentine's Day follows the Super Bowl, WHO CARES if we couldn't literally prove causality? And as a matter of fact, summer, or more specifically hot weather, does kind of cause ice cream cones. If it were shown that the hotter the weather, the more ice cream cones were purchased, would you be no more likely to invest in an ice cream truck in Florida than in Maine because no "causality" was established?

In sum, if consistent correlation is demonstrated, then saying correlation doesn't prove causality is rather meaningless (to the subject at hand).

All that said, as I alluded to at the start of this post, many here have done a good job showing why these two studies do not at all demonstrate any consistent correlation.

|11.16.05 @ 5:46PM|#

fyodor-

As far as Russ's objection, I don't know if winter would give a 27% discrepancy in a large sample. But in a smaller sample, it's possible that winter plus chance could give a 27% discrepancy. And we'd also have to ask ourselves what other local variables might have changed between the two separate periods of 18 months.

On correlation: To be fair, correlation doesn't prove causality no matter how strong or persistent it is. But if a consistent correlation is found, and if it is robust against various controls and found in both time-series and cross-sectional studies, then it is worth looking deeper into the matter to see what is going on. Most research starts with either anecdote, intuition, or serendipity. Correlation is the next logical step because it is easy, and because if the hypothesis is true then a correlation should be present. After that comes controlled experiments and searches for mechanisms. To just recite the mantra "Correlation is not causation! Correlation is not causation!" is absurd.

Based on my anecdotal observations, I would say that peer-reviewed studies by people with good academic training and no conflicts of interest (personal beliefs are not a conflict of interest) tend to be much better than the sort of things that some journalists and activists do. (The guys who published the Montana study apparently did make it past peer review, but they were also activists involved in bringing about the smoking law.)

My anecdotal experience is that good peer-reviewed studies by academics include analyses from a variety of perspectives, tests of several variables, checks to see if the correlation involves a time lag, checks to see if the correlation is strongest in certain subsets of the data where it might be expected, and generally a thorough look at the data set. The methodology still suffers from inherent limits, of course, but they try to push the method as far as possible. They also employ common sense to see if there's a plausible mechanism, and propose studies that could further explore the hypothesis.

OTOH, activists and journalists are more likely to say "Hey, look, these two things follow a trend! Check it out! And please buy a paper/tune into our show/donate to our cause."

The second type of studies give rise to horror stories, and also provide justification to those who simply want to dismiss any and all studies as useless.

|11.16.05 @ 5:47PM|#

fyodor,

I was being a sarcastic smartass; it wasn't to be taken literal. I didn't want to waste my employment hours rebutting something as full of holes as this. But obviously comedy isn't my strongest attribute.

|11.16.05 @ 5:52PM|#

I wish I had some of Thoreau's comedic timing. Some of his fake names have had me laughing uncontrollably.

|11.16.05 @ 5:55PM|#

kmw-

Sorry. I'm relieving boredom while my code runs, so I proved my stupidity by writing a long rebuttal to a joke. Figures.

|11.16.05 @ 5:59PM|#

When the anti-tobacco forces run at least one ad between every radio and TV segment, there's not much chance to bring facts into the debate. There's no hope to overcome the movement, so, might as well start getting my tobacco in sandwich baggies and smoke it in the bathroom, exhaling into toilet paper.

fyodor|11.16.05 @ 6:04PM|#

kmw,

Not that you weren't funny! But as thoreau says, the "Correlation is not causation" chant is tossed about a little too casually here sometimes, and I was looking for an excuse to make that point! So my apologies for insulting your post, but thanks to that little chain of events, we got thoreau to address the matter more at length! :-)

I will only say further, in follow-up to thoreau, that even if a persistent correlation doesn't prove causality, it does prove.....a persistent correlation! Which in many cases is simply enough to be.....enough, or germane to the matter at hand.

As for these smoking bans lead to less heart attack studies, it's beginning to sound more and more like another case of liars damn liars statistics.

|11.16.05 @ 6:05PM|#

Thoreau,

No apology necessary, sorry I wasn't more clear. I suppose I should have been more over the top, to make it more obvious.

But apparently jokes aren't my natural talent.

Mike Laursen|11.16.05 @ 6:23PM|#

Fine with me as long as the Las Vegas illusionists are Penn & Teller.

s.m. koppelman|11.16.05 @ 6:24PM|#

The study didn't distinguish between smokers and nonsmokers, but rather represented a combination of both smokers and those impacted by secondhand smoke.

Or maybe three ways: between smokers, nonsmokers exposed to secondhand smoke, and smokers who lie to their dcotors. ;)

I'm skeptical about snap conclusions drawn from small, imperfect studies, too, but the vehemence with which Mr. Sullum rejects the possibility that smoking bans do lead to decreased incidence of heart attacks without credible research verifying that extreme position is a mite bit typical of his rhetoric.

For someone who ostensibly argues for the freedom to smoke, inject and ingest all sorts of substances on personal-freedom grounds, Mr. Sullum sure spends a lot of energy arguing that smoking may well not have any measurable, documentable adverse health effects at all. Even the big tobacco companies don't bother with that line of attack these days.

|11.16.05 @ 6:51PM|#

I'm reminded of this post from Alex Tabarrok of Marginal Revolution on why we should be skeptical of all published research. It can only be that much more true for unpublished studies.

Hakluyt|11.16.05 @ 7:41PM|#

thoreau,

IIRC, the problems with the study in MT included very small sample size and no effort to control for, well, anything.

By itself, those are enough reasons to ignore the study entirely. Any good epi study has got to have a large sample size. When I took epi that was pounded into our heads over and over again. Large sample sizes. Large sampkle sizes. Anything else is a waste of time.

|11.16.05 @ 9:42PM|#

The New York State ban started March, 2003. I checked the latest available statistics (those for 2002 and 2003). They show a slight decline in the death rate from major cardiovascular disease from 345.1 to 337.1 per 1,000. I would be better to compare 2002 with 2004, but we'll have to wait until December for the 2004 statistics. It's noteworthy that the death rate from lung cancer, breast cancer, and cancer in general all went up during the same period. If you look at enough statistics from a city, you'll find one that supports your view. The New York vital statistics are at http://www.nyc.gov/html/doh/html/vs/vs.shtml .

One last thought, since everyone eventually dies of something, wouldn't a drop in one type of death mean you'll eventually see a rise in another type and vice versa? The most important statistic is the life expectancy.

|11.17.05 @ 12:08AM|#

Off topic, but this might be a first. Two Colorado locales featured in thread titles on the same page! All you guys should come out to Denver and hang with us Saturday night!

|11.17.05 @ 12:47AM|#

fyodor will be there. mediageek too. It'll be groovy.

|11.17.05 @ 1:43AM|#

...hang with us Saturday night!

Is that smoking or non?

|11.17.05 @ 3:55AM|#

Since we're talking about Montana and Colorado, what are the odds that a number of patrons (in the dangerous early stages of a heart attack) went outside for some air (and a smoke) and appeared to freeze to death?

Or that they ventured out to their cars, drove around a bit to warm things up, and in the haze of smoke and frost, drove off a frozen road? I doubt the coroner of Podunk is going to look much further than 'massive hemorrhaging' under all but the most sensitive circumstances.

Statistics suck but statistics based on sloppy data gathering suck worse.

Hakluyt|11.17.05 @ 5:34AM|#

Evade is quite correct. Misidentifying the cause of death is very common.

|11.17.05 @ 10:38AM|#

koppelman,

Did you RTFA? Why get on Sullum's case when the article contains a preposterous line such as this:

"The Pueblo study affirmed that such laws can cause a dramatic improvement in public health."

Rationality was already out the window before Sullum's comments. I helped an old lady cross the street in July 2003, so I'm taking the credit for the reduced heart attack rate!

Larry A|11.17.05 @ 12:29PM|#

Pueblo City-County Health Department conflates correlation with causation, saying the data from Helena "showed restrictions on public exposure to secondhand smoke caused a sharp decline in heart attacks."

I.e. "Smoking in public places was prohibited to eliminate secondhand smoke, therefore the heart attack rate dropped 26% in 18 months."

Unscientific common sense would indicate that either:

• Inhaling minute amounts of smoke during the small percentage of the average month spent in "public places" is far more harmful than actually smoking or
• The tobacco problem would have resolved itself several generations ago when the last smoker had his heart attack.

Somewhat more scientifically, given the coverage of these two studies, the number of scientists on the anti-smoking side, and the number of cities which have similar laws, either:

• The anti-smoking scientists did what jtuf did and looked at the other cities, where they found no or negative correlations in support of this thesis or
• We would have seen them all over the media saying, "Holy cow! Lookit! Lookit!"

Frank Koza|11.17.05 @ 9:35PM|#

Mr Sullum is correct that the original Helena study was downgraded from 60% to 40% in the final paper that somehow made it through the peer review process of the BMJ. The reason the number changed so significantly was because the original methodology seriously flawed, substantial changes were made such as collecting more data and crunching the numbers differently. The methodology of the published version wasn't much better.

For the reasons someone stated in an earlier comment, things like seasonal variance may be a confounder, so the data in the published version only included AMI admissions from the June-Dec period of the years prior to and after the ban, instead of all the AMIs for the few years prior and after the period the ban was in effect. But that methodology is also a weakness of this type of "study" in that weather varies substantially from season to season. In fact there was a substantial jump of over a 30% increase in admissions recorded of 50 in the year prior to the ban compared to the average of about 37 in the years 1998-2000 and 2003. Including that fluke year pulled the average up to make the differential compared to the six months of the ban look even more pronounced.

Another confounder is they reviewed death records, but didn't include in the record any data on deaths of those heart attack victims who died on the way to the hospital and therefore didn't make it through hospital admissions process to be counted as an AMI "because of concern about the accuracy of the assigned causes of death." I haven't seen the death record data, so I have no idea how much of a factor that may be.

And for another confounder, they also didn't include people in the study group if their symptoms started outside the Helena area. Whenever a ban starts, bars and restaurants tend to report a significant drop in business as smokers stay home or don't go out for as long. So folks who might have normally have been in town when their symptoms started might have been home in the outlying 'burbs, "outside the study area", during the six months of the ban. So not only would these people who may have required a longer transport to the hospital after a cardiac event possibly be excluded if they happened to have died en route, but even if they made it to the hospital, they may have been excluded. Below is the text of that from the BMJ paper:

"We reviewed charts of patients from outside the study area to determine whether onset of symptoms occurred in or out of the study area and included them if the patient's symptoms started in the study area. Twenty six patients in the primary acute myocardial infarction group had out of area zip codes; 14 were included. Eight patients with a secondary diagnosis of acute myocardial infarction had zip codes out of the area. We included three patients with a diagnosis of primary myocardial infarction (for example, primary diagnosis of cardiogenic shock with secondary diagnosis of acute myocardial infarction) whose symptoms had started in the study area."

The above are just my opinion. Even if correct, it doesn't matter much, for any analysis today really doesn't matter. The huge 60% drop originally reported in the press releases from the American Heart Association and UCSF and picked up by the main stream press will be forever etched in the minds of the ignorant masses. Those are the people who vote for bans that anti-smokers want to enact through public referendums because they couldn't convince OSHA that their junk science was real.

Now the same from Pueblo. And the anti-smokers even drug out another bullshit "study" and blew the dust off to try more of the Chicken Little sky is falling approach to stir the complacent masses into hysteria. There's no reason why the crap below was just posted on WebMD today with absolutely no reference that it was actually done in 2001 and the anti-smokers had a field day with it in the press back then.

http://www.webmd.com/content/article/115/111726?src=RSS_PUBLIC

Anyone not watching them won't realize that it's old news. Show's you how they take old crap and try to make it look new.

|11.18.05 @ 1:07PM|#

Frank,

The dead giveaway line in the linked article is this:

"Thirty healthy male Japanese men -- 15 smokers and 15 nonsmokers "

It's impossible to make a claim that 30 minutes of exposure to SHS makes one unhealthy AT THE SAME TIME claiming that 15 smokers are healthy! The bullshit comes flying so fast it's hard to catch it all.

Frank Koza|11.18.05 @ 6:30PM|#

LOL, Russ, they sure did. The funny thing though is that a single high fat meal also causes such endothelial dysfunction as they had "proven" in the Japanese study. Only the study on the high fat meal was better constructed because instead of only taking one measurement before and immediately after exposure, the researchers kept taking measurements every hour up to four hours after after the meal to determine how long the effect lasted.

http://jama.ama-assn.org/cgi/content/abstract/278/20/1682

The junk Japanese "study" didn't do that. It appears to be cooked up by anti-smokers with one specific purpose -- to make a propaganda announcement.

Frank Koza|11.18.05 @ 6:41PM|#

I was just informed there was a "study" released from Bowling Green, Ohio, last year as well. There's not much more out there on the web than a press release with results of that "study" which purportedly compares heart attack rates in Bowling Green with Kent, Ohio. If you click on my name on this comment you'll see my write-up on that with all the pertinent links and amplifying information.

Michael J. McFadden|11.18.05 @ 7:57PM|#

I'm sorry... a friend active on this board told me that posts go up immediately. I'd assumed mine had died and resent them as a combination. After rereading the screen response about posts I realized that my earlier ones DID probably go through.

Sorry for all the extra reading/work I just put you through!

:/
Michael

|11.19.05 @ 2:08AM|#

Steve Milloy deconstructed Helena a few years ago, pointing out that, among other serious flaws in the "study," a similar dip in heart attacks had occurred in Helena in 1998, long before any bans.http://www.foxnews.com/story/0,2933,100318,00.html

Then too I'd suspect that heart attack rates have begun to dip slightly all around the country because of more aggressive preventive treatment. A good "control" of course, would be to do this kind of study (comparing this year and half to the previous year and a half) in towns withOUT bans. But where's the kick in that?

Finally, one thing neither Helena's nor Pueblo's researchers know is how many of these people actually frequented restaurants and bars. Chances are a lot of them were older folks who didn't "hang out" much, either before or after.