Something in the air

Obese mouse. Courtesy of Wikipedia.

Here’s a fun headline from the Mail Online: "’CO2 in the atmosphere is making us all fatter’: Researcher says we are increasing in size as gas levels go up.

The premise of the research is this. Over the course of 22 years, participants have been taking part in a Danish study looking at trends in cardio-vascular disease. One researcher, Lars-Georg Hersoug, noticed that the participants in this study were putting on weight; moreover, it didn’t matter whether the participant was fat or thin, they were subject to the same proportional increase in size. Something else has been increasing over the past 22 years, too: atmospheric levels of CO2. The conclusion? CO2 is making us all fat.

In the original report in Science Nordic, three factors are cited as supporting evidence for this theory. Firstly, a correlation between CO2 concentrations and obesity levels on the US east coast. Secondly, animal models showing that dogs, cats, rats, mice and monkeys all put on more weight when they were exposed to higher levels of CO2. Finally, a 2007 study which shows that increased CO2 results in elevated blood acidity, which affects neurons in the hypothalamus which are responsible for appetite regulation.

This culminated in Hersoug and colleagues running a pilot study to test the effects of increased exposure to CO2 on energy consumption in humans. 6 males were randomly exposed to air and CO2, and energy consumption was subsequently measured. The researchers found that consumption increased by 6.1% when exposed to CO2; suggesting that CO2 might make you more likely to get fat.

On correlation and causation

I’m reminded of a fun example of the issues of correlation and causation that I was told back in my undergraduate days. Suppose I’m interested in the causes of skin cancer. I go out and collect some data on various consumption metrics from people in, say, Weston-Super-Mare, and subsequent incidences of skin cancer being reported over the course of a year. I don’t find much going on with most metrics, but one thing in particular produces a particularly striking correlation:

So, from this, I can conclude that eating more ice cream increases your risk of skin cancer. Right?

Well, not quite. There is, of course, another factor that I’ve not taken into account; one that affects both the amount of ice cream sold/eaten, and the incidence of skin cancer. That confounding variable is sun exposure – when it’s hot and sunny outside, people are more likely to go out to the beach and eat ice cream, but greater exposure to UV light increases the risk of getting burnt, which will increase your risk of skin cancer.

So, what about the correlation between CO2 and weight gain? Well, it might be a bit more complex than including a single extra confounding variable. One factor might be age; as we get older, we might be more inclined (although not necessarily!) to lead more sedentary lifestyles (for example, getting a better job might mean you can afford a car, so you do less exercise per day), or our metabolisms might slow down. So it’s a bit simplistic to highlight a single correlation as a cornerstone of a theory, especially when something as complex as weight gain is involved. If you want to know more about issues with correlation, check out Bob O’Hara’s recent excellent blog post.

Pilot studies are not massively useful on their own

The pilot study that I mentioned used 6 participants, and found a 6.1% increase in energy consumption when they were exposed to CO2. However, If you look at the individual participant results in that paper, the results were being driven by two of the participants; 3 of those involved actually showed reduced energy consumption after CO2 exposure. Moreover, it’s not clear from the methods whether all 6 participants received both CO2 and air, or whether participants were assigned to different conditions. The authors themselves concede that the results are not statistically significant, mainly due to the extreme lack of power that testing only 6 people will give you. So really, without a fully-fledged study, we don’t know what the effects of CO2 would be in these circumstances. In all fairness, the authors emphasise that these are proposed effects, and that more testing is needed. That’s not necessarily how it comes across in the news, though.

So, next time you see a scientific study reported in the news that relies on correlations, ask yourself what might also be affecting the factors that are being correlated. If you’re not convinced that what’s being said is correct, try searching for the original paper (the pilot study mentioned here is free to access), and comparing the author’s conclusions to the newspaper headlines. It might be that something’s been lost in translation along the way, or it might be that something’s not quite right with the original study. Either way, it’s always satisfying to come to your own conclusions about the evidence behind the story.