New Mantra in Tobacco Control Research: Reach Your Conclusions Before Conducting the Study

As 2013 draws to a close, I reflect on the lessons of the past year on The Rest of the Story. One of the most notable phenomena that has arisen this year is what appears to be the new mantra of tobacco control research, which I would describe as follows:

The purpose of tobacco control research is to demonstrate preconceived conclusions. If the research does not support those conclusions, make up some excuse or draw those conclusions anyway.

We have seen many examples in the electronic cigarette area where researchers have conducted studies that do not support their conclusions, but drew those conclusions anyway.

Today, I close the year by revealing an example of tobacco control research where the investigators conclude what they want to conclude, not letting the research findings get in the way.

The Rest of the Story

The headlines of a September 12 news article on MedPage Today were alarming: "Third-Hand Smoke Impacts Kids' Breathing." 

According to the article: "Third-hand smoke -- residue that remains on the skin, clothes, and furniture of smokers, even if they do their smoking out of the house -- still impacts children's breathing, researchers reported here. The risk of respiratory tract infections in children from infancy to 13 years of age more than doubled (OR 2.13, 95% CI 1.04-4.36, P=0.04) in households in which parents smoked cigarettes but claimed to smoke only outside, said Edward Dompeling, MD, professor of pediatric lung diseases at Maastricht University Medical Center in the Netherlands."

The results were from a cross-sectional study that examined the prevalence of secondhand and thirdhand smoke exposure among children living in households in South Limburg (The Netherlands) and respiratory symptoms among those children at a single point in time.

The results were reported at a conference in September, but were recently published in the Journal of Allergy & Therapy.

As reported above, the study found a significant association between thirdhand smoke exposure and respiratory symptoms.

There was just one problem. The study also examined the relationship between secondhand smoke and respiratory symptoms, but found no significant relationship. In fact, for kids who were exposed to secondhand smoke in utero and currently, the adjusted odds ratio for respiratory infections was 1.0, indicating no effect at all.

So the authors were left with the odd finding that secondhand smoke is not associated with respiratory symptoms, while thirdhand smoke increases respiratory problems. This finding is not plausible, since exposure to tobacco smoke is orders of magnitude higher in secondhand smoke compared to thirdhand smoke.

So how did the researchers handle this dilemma? Very simple. They simply dismissed the fact that the paper didn't find a relationship between secondhand smoke and respiratory effects as a methodological limitation. They argued that this finding was due to the cross-sectional nature of the study, sampling bias, or reporting bias. In other words, they simply assumed that the finding must be wrong.

That's fine, but how did they handle the finding of the relationship between thirdhand smoke and respiratory effects in the same paper? Did they also dismiss that finding, due to the cross-sectional nature of the study, sampling bias, or reporting bias?

The answer is no. Instead, they reported the significant association between thirdhand smoke and respiratory effects, warned the public about the need to eliminate thirdhand smoke exposure, apparently gave the newspaper reporter the impression that they had found that thirdhand smoke affects kids' breathing, and were quoted as stating that: "We have seen much the same thing with third-hand smoke causing irritation among children. When people think about harm caused by tobacco they are generally aware of smoking and second-hand smoke. Only recently are we finding that third-hand smoke as well can cause problems."

If you read the abstract of the study, you will see that the authors dismiss the findings regarding secondhand smoke, but accept the findings regarding thirdhand smoke.

The rest of the story is that these investigators apparently reached their conclusions prior to conducting the research. They concluded that secondhand smoke and thirdhand smoke both cause respiratory impairment. When their results showed that secondhand smoke was not associated with respiratory impairment but thirdhand smoke was, they simply concluded that the first finding was wrong but the second finding was correct.

So here is a matrix for how they apparently would have concluded under each of the four possible outcomes of their study:

1. Secondhand smoke and thirdhand smoke both related to respiratory symptoms: We conclude that both secondhand smoke and thirdhand smoke are associated with respiratory symptoms.

2. Secondhand smoke associated with respiratory symptoms, but thirdhand smoke not related: We conclude that secondhand smoke is associated with respiratory symptoms. The reason thirdhand smoke was not related to respiratory symptoms was most likely due to methodological limitations, such as the cross-sectional design, sampling bias, and reporting bias.

3. Secondhand smoke not associated with respiratory symptoms, but thirdhand smoke related: We conclude that thirdhand smoke is associated with respiratory symptoms. The reason secondhand smoke was not related to respiratory symptoms was most likely due to methodological limitations, such as the cross-sectional design, sampling bias, and reporting bias.

4. Neither secondhand smoke nor thirdhand smoke related to respiratory symptoms: We believe the reason we did not find an association between secondhand smoke or thirdhand smoke and respiratory symptoms is the methodological limitations of the paper, including its cross-sectional design, sampling bias, and reporting bias.

You see how it works?