Wednesday, April 13, 2005

What's that I've Been Saying, Again?

The study found that having obese parents, suffering from depression, and engaging in radical dieting like forced vomiting, were all risk factors for future obesity in adulthood.

"Engaging in these radical behaviors isn't going to stop you from being obese," said psychologist Eric Stice, Ph.D., the lead author of the study. "In fact they're likely to do the opposite."


I'm curious, however, why they chose to focus only on girls.

I guess because being a fat woman is just so much grosser and more dangerous. Of course, women are more likely to engage in vomiting and anorexic behavior, leading to greater percentage of obesity? Or not? Without a comparison among teenage boys, this is sort of floating around in nowhere land.

6 comments so far. What are your thoughts?

Anonymous said...

You're of course right about the different perceptions of overweight girls and boys. I think this is a key quote though:

"The correlations between these risk factors suggest that individuals who engage in dieting and weight-control behaviors often experience elevated depression."

Or, depression leads to increased weight control behaviors? Hard to tell from correlational data alone. I'd say it's at least a reasonable supposition that a lot of the same pressures and processes which lead girls into what is diagnosed as depression, and possibly through that to weight control behavior, lead in boys to different expressions and behaviors- body dysmorphic disorder, aggression and violence, etc. The question then is, if you run a study on weight control behavior and depression taking both boys and girls into account, if you only compare those expressions are you in fact comparing unlike quantities- namely, more frequent/typical expressions of distress among girls vs. less frequent/typical expressions among boys taking into account the process of cultural norming? Darned if I have an answer, really. It's a distressingly complicated collection of mutually influencing variables to tease apart. 

Posted by Brendan

Anonymous said...

Actually, it's a testable proposition... you'd need to have data collected at at least two different points in time, and there are statistical tests that check to see if a certain variable (say, depression) is mediating another (say, extreme weight control behavior (EWCB)). You can test whether, if one holds depression constant, the other variables continue to affect EWCB, or if it cuts out the variation in EWCB, which would mean that what's really happening is: other variables -> depression -> EWCB. If one holds depression contant and there are still significant effects of the other variabls on EWCB, then it's not really mediating what's happening, at least not in such a tightly causal chain of events.

But that data is harder to collect, because of the trouble of tracking study participants over a period of time in order to collect that second set of data, and there can be high attrition rates, with can screw with the statistical analysis. Which is why so much of this stuff is correlational.  

Posted by Jenn

Anonymous said...

Hm, should have thought before I posted, because there's an addendum - this is why so much of the research out there can be problematic, because causation is hard to determine. It's nice to know there's a correlation, and it's better knowing about that correlation than guessing that it may be out there, but when the studies hit the media, it means all kinds of wacky hijinks can ensue - wives more stressed when they have a job rather than staying at home? It's because freedom is bad for women! Rather than, say, picking out that mediating variable of how much second shift work awaits her when she gets home as well... 

Posted by Jenn

Anonymous said...

Really interesting- that's regression analysis, yes? On the face of it it would seem you'd need a very large data pool to do that effectively given the very large set of variables at play, especially if you wanted to account for that attrition factor and expand the study to cover males as well (and a potentially different set of expressions). I may well be misunderstanding, but I imagine it would take a great degree of statistical manipulation to tease apart the effects of the different variables, if they were occurring in context of different paths to the same presented behavior (EWCB). Perhaps you'd need as large a sample as several thousand? I imagine locating funding for that sort of thing is, uh...difficult.

The media point is deeply true. Same thing seems to show up with studies of kids and TV violence to a degree, as another obvious case. It's in the nature of the media to suck the uncertainty out of a subject to create a falsely simple "bottom line", I suppose. It's deeply not surprising when the shortcut they choose is implicit sexism, either. 

Posted by Brendan

Anonymous said...

What I'd like to see is a study of high school wrestling teams--you know, the boys who are either starving or gorging themselves in order to make weight--sometimes for months at a tome over a period of years. What does that do to their metabolisms? Are they likely to be as fucked up metabolically as women who start on the cycle of dieting when they're teenagers? 

Posted by Natalie

Anonymous said...

Bingo Natalie, That's what I was thinking of as far as the male equivalent behavior is concerned. It's not a small problem either. Most HS wrestling teams would be intimately familiar with it. And I for one fail to see the ultimate public health reason why males were or might have been excluded from the study. While cause of death differs a bit, males die sooner at each and every age cohort than do females. It's a fact of life. They may not look very pretty while doing so, (to be honest very few consider leaving behind a 'beautiful corpse'), but epidemiology would tell you to suspect that being overweight would present much higher risks for mortality and morbidity for males rather than females.

But hey, the NIH is not exactly funding science lately. See Chris Mooney on this issue: [http://www.chriscmooney.com/blog.asp] or here:
[http://sciencepolitics.blogspot.com/]. 

Posted by VJ