Friday, March 06, 2009

Science by Press Release: Study Shows that Secondhand Smoke Exposure Doubles the Risk of Depression, But Validity of Conclusions Cannot be Evaluated

According to an article in today's USA Today, a new study presented this week at the annual meeting of the American Psychosomatic Society finds that secondhand smoke exposure may double the risk of depression.

According to the article: "Non-smokers exposed to cigarette smoke at home or work are more than twice as likely as those not exposed to have major depression, according to a report at the American Psychosomatic Society meeting in Chicago. It's believed to be the first U.S. study tying secondhand smoke to depression; another in Japan came up with a similar conclusion. Unlike the Japanese research, this study confirmed exposure to smoke by measuring cotinine — a chemical that occurs in blood after breathing in smoke. There were cotinine levels for more than 3,000 non-smoking adults in a federal health study. An additional 92,000 non-smokers only reported if they lived with or worked around smokers. Everyone also filled out questionnaires on symptoms of depression. Whether secondhand smoke was verified by the blood, those exposed to smoke were far more likely to have symptoms of serious depression, says study leader Frank Bandiera, a public health researcher at the University of Miami School of Medicine. Even working where smoking was allowed in public places more than doubled the risk of depression, he says."

The Rest of the Story

Although this conclusion has now been widely disseminated throughout the world by the media, we have no ability whatsoever to evaluate the validity of the study's conclusions.


Because this appears to be another example of science by press release. It appears that the conclusions of the study have been released to the media, but that the actual research itself is not being made publicly available. The study itself is not available, from what I can tell, on either the American Psychosomatic Association web site or the University of Miami School of Medicine web site.

Therefore, it is impossible to judge whether the conclusions of the study are valid or not. And if the conclusions turn out to be unwarranted, then it will be too late to reverse them. The media have already disseminated the conclusion widely. Any correction given down the road would have little effect.

It appears that this study has been submitted to a journal for publication. That's fine, and it doesn't preclude presentation of the results at a scientific conference. But if the results are going to be released to the media at this conference, then I believe it is imperative that the study itself be made available for public scrutiny. You can't just release the conclusions, but not the study itself.

Moreover, if the manuscript has been submitted for publication, then it is inappropriate to release the findings prior to publication. Many journals have explicit policies that preclude the authors of submitted manuscripts from releasing the results to the media until publication.

It is problematic that the study authors have apparently not released the full results and methods of the study because without the full methods and results, it is not possible for others in the field to adequately review the work and assess its validity. I personally feel that researchers should not publicize study findings prior to publication unless they are willing to make the full findings and methods available. Releasing results via press release to the media and the public should not be done until publication, or if it needs to be done before publication, then it should only be done with concomitant release of the entire study.

I have a lot of interesting data about secondhand smoke myself, but I am not going to release them publicly until after they are published. The problem is that if I release these data without the full study, then no one has an opportunity to review the study. It is very possible that upon peer review, my conclusions might be found to be invalid, or at very least, there might be some changes I need to make in my methods to correct the analysis (this is quite common). But here's the rub: if my results and conclusions have already been disseminated in newspapers throughout the country, it's too late to correct them when the corrected - or valid - results are actually published.

To me, the story here is not so much about secondhand smoke and depression (a relationship which we have no ability to evaluate), but instead is about the appropriateness of what I like to call "science by press release," by which I mean trying to communicate important scientific findings to the public via the media, but without making the full methods/results available for scrutiny and review. The problem with this approach is that should the results subsequently be found to be invalid upon peer review, it is too late: the information has already been disseminated.

With a little investigation, I was able to find what it appears is an abstract describing the basic methodology of this study.

The methods are described as follows: "Data were obtained from the 2006 Behavioral Risk Factor Surveillance System (BRFSS) and 2005-2006 National Health and Nutrition Examination Survey (NHANES). The Patient Health Questionnaire was used to measure current depression; secondhand smoke policy at work and home was determined by self-report; secondhand smoke exposure was determined by respondent's serum cotinine level categorized as high (.2-15 ng/mL), low (above the detection limit-.2 ng/mL), or undetectable (below the detection limit). Smokers and those with serum cotinine levels > 15 ng/mL were excluded from the analysis (NHANES only). Logistic regression analyses were performed with adjustment for survey design, comorbidity, age, race/ethnicity, gender, socioeconomic status, and alcohol consumption."

The results are described as follows: "Persons living in homes where smoking was allowed anywhere were more likely to be depressed, odds ratio (OR) =1.85 [95% confidence interval = 1.49-1.97] as were persons working in jobs where smoking was allowed, OR = 1.80 [1.12-2.87]. We observed a significant association between high and low/moderate cotinine exposure categories relative to those with undetectable cotinine levels OR = 3.06 [1.41-6.61] and OR =1.44 [0.66-3.10], respectively."

Importantly, this is a cross-sectional study. Levels of cotinine were found to correlate with symptoms of depression in a cross-section of people at a given point in time. This does not necessarily mean that the secondhand smoke exposure is what caused the depression. It may well be that there are confounding factors - associated both with secondhand smoke exposure and depression - that explain the relationship.

It is also important to note that the study appears to have only controlled for a few potential confounders. One very important confounder that was not (apparently) controlled for is a family history of depression, which could plausibly explain the observed association.

No comments: