ICBS Everywhere

Posted June 27, 2009

Most people do not have access to academic journals, nor do they have the training to decipher the jargon printed in scientific articles. Even if they did, most do not have the training to critically evaluate the methods used by researchers. They rely on journalistic media reports for information about new scientific findings. As Ben Goldacre of Bad Science often points out, media interpretations are usually quite far from reality. Sometimes the press releases describing findings overgeneralize or otherwise imply something other than what the researchers have concluded, but many times the reporter is simply misinterpreting the press release due to the same human errors in reasoning which plague the decision processes of most of us every day.

In search of a report on green tea and its alleged cancer fighting properties for my last blog post, I went to our local ABC station’s news website and clicked on the “your health” section link. I did not find the green tea story, but I did find a gold mine of bad reporting.

Among the “Duh!” headlines such as “Better Sleep Habits May Mean Better Grades”, I found several misleading reports about recent research. In the next four posts I will talk about the following, which are mostly errors of assuming that correlation equals cause:

• Snoring during pregnancy leads to diabetes?
• FDA cites limits of ADHD drugs study
• Breastfed kids show increase in GPA
• Job stress affects baby’s birth weight

Snoring during pregnancy leads to diabetes?

The intro line reads:

Pregnant women who snore frequently during sleep could be at a higher risk for gestational diabetes.

“At higher risk” refers to prediction. It means that A is correlated with B such that B may be predicted, with a known margin of error, using A. It does NOT mean that A leads to B. No causal connection may be drawn about this relationship because there are often confounding variables. These are variables which may explain differences in B. For example, rates of violent crimes may be predicted using the volume of ice cream sales; these variables are correlated. As ice cream sales increase, violent crime increases. We do not assume, however, that ice cream makes people violent. Explanations such excess sugar intake are premature; eliminating or controlling for other variables is necessary to determine causal paths. In this case, weather is responsible for both ice cream sales and violent crime rates. As the temperature rises, so do both of the other values.

The research is not in published article form, so the method cannot be examined closely, but a press release is available on the schools’ website.

The principle investigator is quoted as saying:

Snoring may be a sign of poor air flow and diminished oxygenation during sleep, which can cause a cascade of events in your body…. This may activate your sympathetic nervous system, so your blood pressure rises at night. This can also provoke inflammatory and metabolic changes, increasing the risk of diabetes or poor sugar tolerance.

First, this is not a statement that snoring causes diabetes, whereas “snoring leads to diabetes” IS. This statement suggests a mutual cause (much like in my ice cream/violent crime example. However, this is a complicated, convoluted explanation which requires quite a few assumptions.

Weight (which is not controlled for in this study) is positively correlated with snoring; this is a strong relationship which is well-documented. Weight is positively correlated with diabetes; this is also a strong, well-documented relationship. If poor air flow is a mediator of the relationship between weight gain and these variables, the relationship between snoring and diabetes is meaningless.

I am not making the statement that the research us meaningless or that the findings are not useful. It is obvious that research needs to continue on the relationship between quality of sleep and diabetes. However, misreporting findings and their meaning is as bad as scientists drawing poor conclusions from them. Together these things are dangerous.

So, what’s the harm? Treatment for snoring may or may not also prevent diabetes; it really depends on what that treatment is. Women who treat snoring with nasal strips may be wasting resources for a false sense of security. The snoring itself is mostly likely not the problem and eliminating the sound does not eliminate the cause. Perhaps losing weight prior to pregnancy and reducing weight gain during pregnancy are more productive pursuits.

In the next post I will discuss scaremongering about ADHD medication.

2 Comments:

Cujo359 on June 28th, 2009 at 21:51:
Thanks for pointing this out. I’ll have to make a point to come back here and read the rest of this series.
As for the subject at hand, the weight issue was one that struck me as a possible explanation. I’m not a doctor, but what you wrote about being overweight and being more vulnerable to both respiratory problems and diabetes is pretty widely known. That this didn’t occur to the person writing that article is troubling, to say the least. Supposedly, large news organizations have reporters who specialize in subjects like science and medicine.

Administrator on June 28th, 2009 at 22:50:
The press release does mention weight gain as a possible mechanism for “increased airway resistance”, but do not suggest that it is a common cause rather than part of a single causal chain (which is less parsimonious).

I haven’t been thrilled with science reporting. Even the doctors who consult on broadcast news make these kinds of suggestions often, but these online blurbs are probably written by untrained copy writers. They just restate what’s in the press release. The press releases and, to some degree the articles themselves, are written to sell something. In the case of science, it is often the scientist’s opinion.

Still, it’s the best method we have!