Tuesday, November 09, 2010

The future babble of obesity prognostication

Lies, damn lies and statistics.

Now to be fair I've been primed to disbelieve most future predictions by being mid-way through Dan Gardner's excellent Future Babble, but really, obesity rates to hit 42% is headline news?

The headlines referred to a study published last week in PLoS Computational Biology that had some truly fancy Harvard folks hammer out a formula to predict what obesity's going to do down the road. Those fancy folks are building on a prior study of theirs that proved that obesity is socially contagious and this one takes that one to it's apparently logical mathematical landing point of 42% of everyone you know's going to be obese one day.

Of course obesity is a highly complicated condition. Yes, it's simple to describe, more energy goes in than out, but ultimately there are a great many variables at play which impact on intake and output.

This most recent study doesn't appear to me to address any of those. For instance off the top of my head I would have thought it important in a study of prognostication to provide fancy statistical ways to explain why it wouldn't matter to outcomes if portions sizes continued to grow in restaurants, advertisers continue to ramp up their targeting of children, food delivery becomes even more ubiquitous, incentive or disincentive taxation schemes were enacted, or if suddenly our governments stopped subsidizing the base ingredients that allow food manufacturers to make calories insanely cheap, but hey, I admit quite readily, I'm no mathematician.

More importantly I've got to ask, "So what?". Arguing about how high obesity rates are going to climb is about as useful to helping the problem as folks on the Titanic arguing about exactly how big that iceberg is that's looming on the horizon.

I realize that basic research is important and I admit that I've been set off more by the news coverage than by the study, but at the end of the day what I'm trying to say is that while I'm sure the intellectual exercise of guesstimating how high obesity rates can climb was personally rewarding for those researchers, I can't help but wish that instead they'd have used their massive collective brainpower to work on something that actually has even the remotest bit of clinical relevance.

Quid tum!

Hill, A., Rand, D., Nowak, M., & Christakis, N. (2010). Infectious Disease Modeling of Social Contagion in Networks PLoS Computational Biology, 6 (11) DOI: 10.1371/journal.pcbi.1000968

[thanks to Idea Sandbox for the graphic up above]

Bookmark and Share


  1. "I can't help but wish that instead they'd have used their massive collective brainpower to work on something that actually has even the remotest bit of clinical relevance"

    I know I'm probably reading too much into that last sentence, but are you suggesting that clinical research is the only research worth doing? Looking at the other papers put out by this group, I'm really not sure that their skill set lends itself to the type of work that is going to be immediately relevant to clinicians, no matter what question they try to answer.

    Similarly, do you think it's really possible to build *all* of those factors related to the food environment into their models? Of course not. It seems more than a little bit unfair to hold a paper to an impossible standard simply because it's gotten a lot of media attention.

  2. No Travis, as noted in the piece itself, I realize there's value in basic research and also as noted, I was primed to be frustrated more by the reporting than the study itself.

    Your other question though is quite pertinent. No, I don't think it'd be remotely possible to build those factors into a mathematical model, which in turn is precisely why I got my panties in a bunch in considering the impact and utility of the article.

  3. I guess my question is, if non-clinical research is important, why does it matter that this study lacks impact and utility in the clinical setting? That seems the very definition of non-clinical research. I realize that you didn't say explicitly that it lacks utility solely in the clinical setting, but it does seem likely that an estimate of future obesity rates, flawed as this estimate may be, may nonetheless be a useful thing for people in other fields.

    And if the real issues are with the media coverage, then why bother critiquing the study, rather than the media? As a researcher who has had people over-react to media coverage of my own work from time to time, it makes me grimace whenever someone gets annoyed by poor or incomplete media coverage and chooses to take it out on the research itself.

  4. Your argument's fair Travis, in that perhaps there's great utility in the mathematical modeling of the spread of medical problems in social networks.

    I didn't critique the study's ability to do that.

    I did and do suggest (and it would seem to me that you'd even agree), that as far as the modeling of future obesity rates goes, there's really no way to account for the many societal variables that may impact on those rates.

    To me at least, that's a very fair thing to critique and I think frankly, I did so quite gently.

  5. You miss the point! Clearly, this means that "What percentage of people will be obese in the future?" is the Ultimate Question of Life, the Universe, and Everything.