Tuesday, March 22, 2011

Be careful of what you study, how you study it, and how you write it up: the case of climate change and obesity

I happened upon an article looking at the association between obesity and climate change that was published in 2009 by the International Journal of Epidemiology (I know I am a little behind the times, but was intrigued by the title). It generated a lot of criticism from the North American public and science blogging community (examples here, here, and here). This has been mostly for good reason I think, but not always well articulated as to why. I don't believe that the authors, P. Edwards and I. Roberts of the London School of Hygiene and Tropical Medicine, set out to blame obese people for climate change, but it certainly came out that way.

The study used computer modeling to compare two hypothetical populations of 1 billion people, with BMI distributions similar to the figure I quickly sketched together below.  They dubbed the one on the left the 'normal' population and the one on the right the 'overweight' population. The labeling is unfortunate because both populations do have normal weight, overweight, obese individuals, and even underweight people, just not in the same proportions. Both distributions were assumed to have the same mean height for men (1.75 m) and women (1.6 m). 

Writers that have critiqued this paper have argued that computer modeling is not science (and are even skeptical of the term 'Population Health'). While I disagree with that sentiment (well both actually), I don't believe that this is truly an epidemiological study either, and am somewhat surprised that IJE published this. Nonetheless, computer modeling can be made more credible with plausible assumptions and statistical tests. Unfortunately, both of these things were lacking in this study. 

The main aim of the paper was to estimate, rather crudely, differences in greenhouse gas (GHG) emissions from the food industry, and from car and aviation travel, between the 'normal' and 'overweight' populations.

To estimate GHG emissions from the food industry, they first estimated the aggregate energy requirements for each population (to maintain weight). Basal metabolic rate was estimated using the Schofield  equations for men and women, and the activities of daily living were considered to be the same for both populations and added to BMR requirements (a simple energy in=energy out equation, which we know is not that simple).  They estimated that the overweight population required 19% more energy than the normal population. Then, based on aggregate GHG data from 2000, they estimated that this would translate into an increase of 0.27 Giga tonnes (GT) of GHG per year from the food industry alone.

Okay so far? It's rather crude and for this to be somewhat credible I am assuming that the two populations are exactly the same in terms of distributions in age, height, and behaviours (which are usually not normally distributed - often depending on socioeconomic status and other factors). This 'exact sameness' between populations, other than their distribution in BMI, was not explicitly stated as far as I can tell (only an age range and mean height were given).

Next, was estimation of GHG from car travel. The assumption was that extra fuel would be needed to transport a heavier population. Makes sense, but the equation for how mass affects mileage was based on a personal communication with Leonard Evans and not a published reference. If Leonard Evans is well-known in his field, great, but readers not familiar with the field (and I am just guessing Traffic Safety based on a quick Google search) need to know that this is a reputable reference.

I get that if a population, on average, is heavier that fuel consumption would likely be higher for the heavier population. I don't get why two different types of cars were used to calculate the GHG emissions from driving (remember we were assuming that behaviours were the same). The normal population was assumed to have a smaller car (because they are thinner?) such as the Ford Fiesta, even though there are overweight and obese people in the normal weight population (remember that it's a distribution). The overweight population was assumed to have a larger car such as a Ford Galaxy (because they are bigger?), even though there are normal weight people in the overweight population (again it's a distribution). Larger cars emit more GHG - the Ford Galaxy emits more CO2 then the Ford Fiesta. I'm pretty sure that in North America, car choice is not totally dependent on body size (i.e. thinner, rich people still buy gas-guzzling SUVs), although I have no reference for that.

The authors then go on to assume that the overweight population would walk less and drive more. Again behaviours are dependent on a lot of other things besides weight. Plus, there are overweight people in the 'normal' population, so why wouldn't they chose to drive more too? With all of these assumptions taken together, and using aggregate GHG emissions for the transportation sector as a whole, the overweight population was obviously estimated to be responsible for more GHG emissions due to car travel; 0.154 GT/y more based on 30 min/d of driving and then an added 0.019 GT/y because of deciding to take the car more often instead of walking.  

Photo credit: Nika Vee
Based on the assumptions of car travel, I'm surprised that the overweight population was not assumed to have chosen a larger jet for their air travel needs compared to the normal population. Alas, no, the same fuel requirements were calculated for each population with just average weight varying between the two. The authors assumed that 5% of each population would take one short-haul flight per year. This resulted in 2.038 more GT/y of GHG emissions for the overweight population. I could also turn the tables here with my own 'out-of-left-field' assumptions and say that because some airlines charge an extra seat for overweight passengers who do not fit comfortably in one seat, overweight people may be less likely to fly and therefore should not be counted in the 5%, or held responsible for extra GHG emissions from flying.  Even if overweight people are just as likely to fly, having to buy an extra seat could offset their extra weight anyway.  

On top of the sometimes absurd assumptions, and simplistic calculations, the authors still failed to adequately explain their findings in context. Even if this study had been perfectly conducted, obesity and climate change are now deeply ingrained socio-political issues, where adept writing is needed. Writing articles that are interpreted as blaming, even if based on epidemiological or experimental studies (which this study was not), are potentially harmful. Judging by the comments to news articles and hate mail that the authors received, I would assume that many people interpreted the article as blaming, as did I. I am not even certain that the differences reported in the paper are actually important in the grand scheme of things...Maybe someone can enlighten me.

Some quotes from the paper:

"We argue that increased population adiposity, because of its contribution to climate change from additional food and transport GHG emissions, should be recognized as an environmental problem."

"Neverless, the assertion that increasing population adiposity will result in an increase in GHG emissions is justifiable and provides further evidence of the link between human health and climate change mitigation."

There was no consideration for time ('before-after') in this study, which is necessary to say that A causes B.  Although a now heavier population may exacerbate climate change (this study can't even provide that level of evidence) it is more than likely not the cause, as both climate change and population shifts in BMI have taken many years to occur. In fact, human induced climate change has been noted as far back as the late 1900s, well before the obesity epidemic. Maybe climate change caused obesity? 

The underlying processes leading to both climate change and to obesity are more important to focus on than simply obesity alone (for example, our dysfunctional food system, culture of convenience and technology, and the way we have designed our communities, to name a few). Given the complexity of society, we as scientists, using the best available evidence, cannot say for certain what causes what, but we are starting to develop a pretty good idea. Focusing on obesity alone will not mitigate climate change, as noted by a climate scientist. Based on the quality of the study as well as this lack of discussion of potential underlying influences, I do not believe that the authors had grounds to make such strong statements, such as the ones quoted above. Other authors share a similar view, and perhaps explain more eloquently than I. Unfortunately the letter is not open access (twitter me if you would like to read it @PhDPopHealth).

It's up to us as researchers to think about how all faucets of our research can have an impact on our specific scientific fields, the larger research community, as well as the public (this includes knowledge translation).  It is our responsibility to ensure that our research is worthwhile, well conducted, and well communicated and thus, for the betterment of society (whether results are positive, negative, or null). Bad research is useless at best and harmful at worst. Research construed as blaming the obese for climate change will do nothing to help obesity or climate change. 


Edwards P, & Roberts I (2009). Population adiposity and climate change. International journal of epidemiology, 38 (4), 1137-40 PMID: 19377099

1 comment:

  1. Great post Megan, I think I had browsed their paper a while back and didn't give it much weight, for several of the reasons you mentioned. It is absolutely true that authors have the responsibility to set their data in context and really think out its transferability to the general public. I can see why IJE might have published it (afterall it is a "hot" topic), but I'm surprised a reviewer didn't pick up on your points.