The editor of the Journal of Clinical Psychiatry needs to have his head examined

The Journal of Clinical Psychiatry recently published a paper on the relationship between smoking and post-traumatic stress disorder. This newspaper report on the article starts out okay:“Post-trauma mental health disturbances such as post-traumatic stress disorder (PTSD) are associated with increased smoking, either by starting to smoke or an increase of tobacco use,” write Dr Peter G Van der Velden, of the Institute for Psychotrauma, Zaltbommel, the Netherlands, and colleagues.

That certainly makes sense. The article continues:

However, few trauma studies have examined whether smoking is a risk factor, or “marker” for PTSD or other mental health disorders following a disaster.That seems a reasonable thing to test as well. You find a bunch of smokers and non-smokers before a traumatic event. You subject them to a traumatic event. Then you see how they fare. It won’t tell you anything about the impact that smoking has on coping with trauma, but it will tell you something about whether the kind of people who smoke might fare better or worse in response to a tragedy.

To look into that aspect, the researchers conducted surveys 18 months and 4 years after a fireworks disaster in Enschede, the Netherlands. Included were 662 adult victims and 526 residents of another Dutch city who were used as a comparison group.

Oops…by surveying victims for the first time 18 months after the disaster we cannot even answer the question about whether the kind of people who smoke respond badly to tragedy because they are classifying people as smokers based on whether they smoke 18 months after the event. If you look back up to the first passage in italics in this blog post, you will see that we already know that people start to smoke after traumatic events. Presumably, the harder the tragedy hits you, the more likely you are to take up smoking. Of course, you could ask people whether they smoked before the tragedy, but it appears that these researchers didn’t bother, or at least aren’t using the information (or maybe they do but it isn’t mentioned in this media report).

Victims who smoked at the first evaluation were more than twice as likely as those who did not smoke to have severe anxiety symptoms, nearly twice as likely to have severe hostility symptoms, and close to three times as likely to have a diagnosis of disaster-related PTSD at the 4-year evaluation.

Interesting, perhaps, but I have no idea how to interpret the result. One thing I can guarantee you I would not conclude with any confidence is what the authors conclude:

If these findings are confirmed by other studies, disaster victims who smoke may be able to reduce their risk of developing mental health disturbances if they quit smoking, Van der Velden and colleagues conclude.

What?! For starters, there is nothing in the study about quitting smoking — only about people who do smoke. Second, where do they get the idea this relationship is causal, or that the direction of causality runs from smoking to anxiety rather than vice-versa? This is not too different than advising sick people to avoid hospitals because people who visit hospitals die at a much higher rate than people who don’t. Or me thinking that if I wear a Bulls jersey with the number 23, I will win an NBA championship.

(Thanks to David Jones for pointing out this study.)


Nolan Matthias

I agree completely, but I think you might have meant the number 23.

Nolan M

billypilgrim37

The press release doesn't quite interpret the article correctly. It says:

Victims who smoked at the first evaluation were more than twice as likely as those who did not smoke to have severe blah blah blah...

What it should say is: Victims who smoked at the first evaluation were more likely to blah blah blah AT THE SECOND VISIT.

So they compared folks who were the same (controlling for a host of psychiatric symptoms) at T1 except for their smoking status, and saw how they were doing at T2. Given the natural history of symptoms associated with stress, trauma, and PTSD, clinically speaking that's a perfectly reasonable time period over which to follow people.

Your point is well taken that smoking at T=0 would have been valuable (although recall biases suggest that the accuracy of such data would be suspect itself), but assuming the matching in their cohort was reasonable and they've achieved a reasonable counterfactual (which appears to be the case), there doesn't seem to be particularly any reason why the suggested mechanism isn't very well supported by the data shown in the original article.

Read more...

kahomono

You are more unlikely to be able to evaluate the quality of a study's design from the media accounts of the study than you are to be able to diagnose, say, Britney Spears' psychiatric disorders (if any) by reading People.

Read the study before you trash it. The only reliable conclusion you can draw from the press accounts is... which issue of the actual journal you need to crack.

isaac

@Billy - It could be that downward spirals in mental health following a traumatic event and increased chances of being a smoker have the same cause.

Even if you studied people who did, in fact, quit smoking, the conclusion would still be tough - presumably the people who quit have a better ability to overcome their tobacco addiction than those who don't.

zbicyclist

1. Looks to me that Levitt read both some press release material and the actual article, but only he can clarify that.

2. There's a bias at work here. If the result agrees with a prior belief, or if it suggests something that seems reasonable/desirable, there's a tendency not to look as closely for methodological flaws. Some fields [and some journals] may be more susceptible to this than others.

As an example of this bias, look at my statement in #1: Levitt, a top researcher at a top research university, surely wouldn't write such a public criticism of a study in another social science field if he hadn't done his homework. Therefore, I both assume he did AND look for indicators in his blog posting to indicate that he had in the quotations [for example, the quoted paragraph beginning "Victims..."].

-- but now that I reread his post carefully I think kahomono (#3) might be right.

Innocent Bystander

Can I just be cynical and point out that, along with "obesity," smokers are the #1 favorite group right now that it's socially acceptible to dislike? The conclusion drawn is so absurd compared to what was actually studied that I'd be inclined to believe the purpose at the outset was to show that quitting smoking reduces psychological harm and they made up a study to try to show that. After all, MDs get so much money and attention from talking about how terrible smoking is for you, why shouldn't psychiatrists get in on the action too?

(Ex-smoker here. I'm not saying it isn't bad for you. I'm just saying there's a certain... hysteria that seems to come into play right now whenever smoking is discussed. Apparently, no matter WHAT ails you, smoking is the cause of it. Even psychological problems.)

egretman

The cumulative risk over a lifetime of smoking is 10% for lung cancer and another 5% for other cancers such as throat, mouth, etc.

Now that's cumulative. In other words, if you die from it, then it will be far into the future. So why quit?

But everytime I get a rental car or a motel room that reeks of that foul stale smelling odor from the last smoking idiot to occupy it, then I wish fervently that the bombastic Rock Apes of Southern Borneo would hunt down the culprits, string them up by their thumbs, cut off their toes, and let them bleed to death while slowing skinning them alive.

billypilgrim37

Given this statement: Of course, you could ask people whether they smoked before the tragedy, but it appears that these researchers didn't bother, or at least aren't using the information (or maybe they do but it isn't mentioned in this media report).

It certainly appears Dr Levitt did not read the actual article, because the actual article explains clearly the methodology used to assess smoking directly below one of the subheadings in the methods section.

isaac:
"It could be that downward spirals in mental health following a traumatic event and increased chances of being a smoker have the same cause."

I'm sure that's the case. Rates of smoking for almost any psych disease are off the hook compared to healthy controls.

The point is that this study actually did a reasonable job establishing a counterfactual by assessing psychiatric symptoms at T1 and finding that folks who had the same condition at T1 except for smoking status turned out differently at T2. That's a reasonable study design (not problem free-but reasonable), and certainly not worthy of having Levitt calling for their head to be examined.

If we want to talk about biases, how about an apparent antipsychiatry bias, since Dr Levitt found it prudent to tear apart a valid piece of epidemiologic research referencing only its press report? I'm sure he's very familiar with the idea that a media report of a bit of research and its actual content can widely vary, and I can't imagine that finding the article and reporting on it would have added much more than 15 minutes to writing this post.

Read more...

Bruce G Charlton

Speaking as an editor of a medical science journal, I object to the headline of this posting!

There are many reasons for publishing a paper - for instance to stimulate debate or trigger further work on a topic (perhaps work done to in order to refute the original paper). Simply because a study is flawed does not mean it should be excluded from communication in the scientific literature.

Surely it is unreasonable to assume or imply that an editor agrees with the entire content of his or her journal?

egretman

An editor of a medical journal, saying that it is OK to publish flawed studies? Are they published in the Journal of Flawed Studies?

pkimelma

It seems to me that medical, psychological, and economic research continues to be plagued by an ongoing problem of suppression of variables that are likely co-factors. We can all laugh at studies from the past that looked at birth defects from older mothers without considering smokers vs. non-smokers, or alcohol consumption and heart disease without considering diet, yet many researchers have not learned from this. The reason so many conclusions fall apart when re-examined (usually using large multi-variable studies, such as the women's health study) is that the data miners look at many variables and only remove if found not to be a factor.
In this study, they not only needed to look at whether smokers before, but also former smokers. We know smoking uptake is not random; a combination of physiological and environmental factors affect who takes up smoking and who does not. Anxiety and stress seem to be strong correlates with smokers - likely not due to being smokers but due to being the type of person who will smoke.
So, if the study had looked at status before the event (smoker, former smoker, non smoker), and the status after the event, they would likely draw far different conclusions.
That said, I have to disagree with Mr. Levitt about the jersey correlate. I agree that the problem here is that they assume smoking increases stress, and so stopping smoking would stop stress, but I think it is unreasonable to assume there is no common causal element, as in the case of jersey number.

Read more...

billypilgrim37

I agree that the problem here is that they assume smoking increases stress, and so stopping smoking would stop stress

Not quite true. They assume that smoking might modulate the biology of how the body handles chronic stress. I'm pretty sure any smoker would assure us that smoking reduces stress in the short term, so that's not what's being addressed here. The positive stress release from a short-term dopaminergic reward system (and thus modulated by cortisol in the short run) can't be compared directly to a long-term modulation of the HPA hormonal axis.

Moving ahead though, a more interesting thread would be to propose what sort of research could follow from this hypothesis. Would smoking cessation programs even be clinically effective in this population? It may be very true that cigarette smoking wreaks havoc on long term hormonal stress regulation. But it may also be true that smoking cessation programs in a post-traumatic population would be futile on the whole.

Read more...

umbrellaman

Shame on you, Steven. Really. It's clear that you and many others who commented did not read the original article. As someone whose abortion chapter was grossly misrepresented by media reports I think you should know better.

I only had a limited time to devote to looking at the original article, but it is absolutely clear that it is a lot more complicated that the Reuters article or Steven made it out to be.

Steven: "You find a bunch of smokers and non-smokers before a traumatic event. You subject them to a traumatic event. Then you see how they fare. It won't tell you anything about the impact that smoking has on coping with trauma, but it will tell you something about whether the kind of people who smoke might fare better or worse in response to a tragedy."

That's not exactly what they were they looking for. The objective of the paper was "to assess whether smoking is an independent risk factor for mental health problems among adult disaster victims and a nonexposed comparison group." They were specifically not looking for the "how the kind of people that smoked would deal with disaster." They did regression analysis trying to isolate the effects of being a smoker 18 months after the disaster on whether or not there were disturbances 4 years after the disaster. They controlled for demographic characteristics, depression, anxiety and hostility symptoms at 18 months and the presence of stressful life events.

This is how they tried to make their point
4 years after the disaster, when looking at people who experienced the disaster, the group that smoked at 18 months were 2.64x more likely to be experiencing severe anxiety symptoms than the group that did not smoke at 18 months.
- No big deal as you say, because maybe the reason that they were smoking is because they were already having problems. Steven: "Presumably, the harder the tragedy hits you, the more likely you are to take up smoking."
But that's why they performed multivariate logistic regression analysis, accounting for demographics ("the kind of people who smoke") and symptoms of depression, anxiety and hostility at 18 months ("the harder the tragedy hits you") and adjusted the odds ratio from 2.64 to 2.32.
Furthermore looking at another group of folks not exposed to the disaster, smokers at 18 months were 2.59x more likely than non-smokers to have severe anxiety symptoms 4 years after the disaster (that they did not experience) . However, when you perform the regression analysis factoring out demographics, etc, the adjusted odds ratio is 1.64, which is not statistically significant.

So this is how they conclude that smoking in the context of a disaster - independently of other factors such a demographics or symptomatology leads to an increase risk of symptoms 4 years later. This goes along with the notion that long term PTSD symptoms have to do with how the trauma is processed soon after the trauma. It may be more common to have symptoms close to the trauma, but the question is why do some people "don't get over it" 4 years later? There are other studies that suggest smoking impairs the stress response, and since their data *while controlling for demographics and symptomatology at 18 months post disaster* shows that smoking at 18 months confers an increase risk of symptoms at 4 years - it *might* be a good idea for people to stop smoking after a trauma.

Whew. I haven't had the time to nitpick this study and determine from my point of view whether this is a good study or not and whether I agree with their conclusions. But that long explanation was a way to show that 1) the press report was an extreme oversimplification of the study and that 2) Steven did not do his homework.

Notably, some of the factors people said they ought to look at were included such as whether they ever smoked and whether they smoked 2-3 weeks after the disaster.

Read more...

kahomono

@some who take me to task (along with Steven) for not having read the article:

OF COURSE I haven't read it. But it was also crystal clear to me, even at oh-dark-thirty, that Steven hadn't either, he'd only read press accounts.

I was only objecting to the unfairness of writing what he wrote from only press accounts of a peer-reviewed scientific paper.

As for the study itself, well, in my personal list of priorities, I might get to that before I die... but that's not the way to bet. ;)

abott_uhmm

They may have confused cause for effect. Smoking can serve both as a grounding mechanism, in that it involves 4 of the 5 senses, and is sometimes considered to improve cognition (its an anecdote I know, ...but there is a reason those smoke-filled rooms were where so many decisions got made.) The number of people who take up smoking after developing schizophrenia is well documented and might be considered an effect rather than a predictor, as the smoking might be considered an effect of PTSD rather than a predictor.

GamblingEconomist

Even if Levitt did not read the actual article, I'm pretty sure his view would not change from reading it. I think his framework is that if there is no instrument/natural experiment there is no proven causality. In this particular case, not only is the causality unproven, it is not even intuitive.

umbrellaman

I think we can reach common ground here. kahomono - I wasn't bashing you for not reading to article - only those who commented on what was in it. As for Steve, I can understand why he might not have read it, but I think if he didn't he ought not to have had the polemical title, or at least had a caveat that "if this news article reports this article accurately, then..."

I am a psychiatrist, and this journal is one that I read regularly. This paper was not a pleasant one to read and certainly inelegant in its methodology. At the end of the day, I think Levitt's point is still that since they did not examine people that quit smoking they ought not to reach any conclusions about whether that is a good idea or not for people exposed to trauma.

I'm not sure that this criticism is still an artifact of the Reuters article. First, there is other neurobiological work that suggest that smoking impairs the stress response. Secondly, the main point of the article is not to suggest quitting smoking. Rather, the part pulled out by the Reuters article is in the section of the discussion that speculates on further research and implications.

Let's posit that through their analysis the authors establish that smoking is an independent risk factor for developing psychopathology after a trauma. I think Levitt is saying that even so, you can't say that quitting smoking after an trauma is a good idea.. But in my reading of the article, I don't think they leap to that conclusion. If Levitt suggests that they need to do a study where they examine a large group of say, Marines that served in Iraq and have half of them try to quit smoking in order to make that conclusion, I think the authors would agree - in fact I think they are probably angling for funding to do that study right now. Rather than discounting the need for the kind of studies that Levitt would want, I think the authors are merely suggesting that their research helps justify looking at that question.

Here are the closing paragraphs of the paper:
"As such, questions about, smoking behavior among disaster victims may help to identify disaster victims who are at risk for postevent mental health disturbances. This
study and previous studies add valuable information for the early detection of victims who develop or suffer from severe mental health problems after a disaster, in addition to variables such as predisaster psychopathology28 and self-coping efficacy.29 This information is useful not only for professionals who participate in a postdisaster mental health care program, but also for general practitioners and other professionals who have survivors in their patient population,
Furthermore, results and previous research suggest that a smoking cessation program should be offered to smoking survivors as an integrated part of a mental health care program after an event/'30 However, smoking cessa¬tion may aggravate mental health disturbances, because smoking may be a coping mechanism to reduce distress. Nevertheless, in the study of McFall and colleagues,30 smoking cessation was not associated with worsening symptoms of PTSD or depression.
In addition, our results showed that survivors who smoked in the past were not more at risk for mental health problems at T2 than nonsmokers. If our findings are replicated in future studies, this may suggest that disaster victims who smoke have the opportunity to reduce the risk for postevent mental health disturbances when they quit their smoking behavior after the disaster. There are all kind of leaflets and fact sheets available for disaster victims. Perhaps the advice to stop smoking or information about the possible effects of smoking on their mental health should then be added to these leaflets. Supplying victims with this information may enhance their own efforts to cope with the adverse effects of the event and regain control over their lives."

I read a lot of caveats - *If* the findings are replicated it *may* *suggest* that quitting might be helpful. The authors are clearly suggesting that more research is in order.

Read more...

kahomono

@umbrellaman - OK, I understand your points.

I still have a problem with this whole thing. More than many people, Steven should know that press accounts of research and actual products of research are vastly different. So different that second-order accounts of those press accounts are pretty much worthless.

I find title of the post that started this thread inflammatory and well beneath what I thought to be L&D's standards.

newkon

I agree that the problem here is that they assume smoking increases stress, and so stopping smoking would stop stress

Not quite true. They assume that smoking might modulate the biology of how the body handles chronic stress. I'm pretty sure any smoker would assure us that smoking reduces stress in the short term, so that's not what's being addressed here. The positive stress release from a short-term dopaminergic reward system (and thus modulated by cortisol in the short run) can't be compared directly to a long-term modulation of the HPA hormonal axis.

Moving ahead though, a more interesting thread would be to propose what sort of research could follow from this hypothesis. Would smoking cessation programs even be clinically effective in this population? It may be very true that cigarette smoking wreaks havoc on long term hormonal stress regulation. But it may also be true that smoking cessation programs in a post-traumatic population would be futile on the whole.

Read more...

Nolan Matthias

I agree completely, but I think you might have meant the number 23.

Nolan M