Thursday, August 18, 2011

Why are there less "fit and fat" folks than there used to be?

That was the question that leaped to mind after looking at one of the appendices of the recent Edmonton Obesity Staging System paper in the CMAJ.

Looking at the graphic up above (if you click it, it'll get larger), it would appear that the earlier NHANES III cohort (1988-1994) included a significantly higher percentage of so-called, "fit and fat" folks, as compared with the later NHANES 1999-2004 cohort.

In the later cohort, the percentage of the study population with an EOSS score of zero (meaning folks with overweight and obesity and no medical or life related co-morbidity) was pretty much zero, whereas in the earlier cohort it would appear that over 15% of overweight folks had an EOSS score of zero, as did just under 10% of folks whose BMIs ranged between 30 and 35 and just over 5% of folks whose BMIs were greater than 35.

Now I'm far from a statistician, but certainly were than any stats savvy folks reading this post, I'd be very curious to know if that differences up above were statistically significant. Given the size of the cohorts and the dramatic differences seen, I'd be surprised if they weren't.

In terms of what's going on?

According to Obesity Panacea's Travis Saunders', one possibility is that the distribution of our weight is changing, and that where we're carrying it is the problem, with increasing abdominal distribution weight-independently increasing our risks of developing a constellation of different chronic diseases.

Travis forwarded me a paper written by Ian Janssen, Margot Shields, Cora L. Craig and Mark S. Tremblay that looked at differences in waist circumferences and 5 skinfold thicknesses for given weights between 1981 and 2009. What they found was that for any given weight, waist circumferences were higher, as were skinfold thickness values.

Their conclusion in the paper seems downright prescient when applied to that graph up above,

"These findings suggest that even in the absence of a change in population obesity prevalence as determined by BMI, the population health consequences of obesity seem likely to increase more than anticipated"
Given the impact of exercise on distribution of body weight, this definitely lends ammunition to ongoing exercise promotion efforts, and combining that impact along with the results of the second EOSS paper, the one that suggests lifestyle dramatically attenuates EOSS staged risk, and together they would suggest that exercise promotion should take obesity treatment out of the rationale for exercise, and instead focus on exercise/health at any size.

To suss things out further, I'd love to see which weight related co-morbidities increased in the later years cohort - certainly knowing which co-morbidities are on the rise, would help in understanding what exactly's going on.

So whether it's increased abdominal distribution of weight or not, it would certainly seem that there's something else out there, something other than absolute weight, that's increasing morbidity in the population with overweight and obesity. My recommendation is that we redouble our efforts to figure out what that something else is and work on it, as I'm guessing we'll have more luck treating it, than we do treating weight.

Padwal, R., Pajewski, N., Allison, D., & Sharma, A. (2011). Using the Edmonton obesity staging system to predict mortality in a population-representative cohort of people with overweight and obesity Canadian Medical Association Journal DOI: 10.1503/cmaj.110387

Janssen, I., Shields, M., Craig, C., & Tremblay, M. (2011). Changes in the Obesity Phenotype Within Canadian Children and Adults, 1981 to 2007–2009 Obesity DOI: 10.1038/oby.2011.122

Jennifer L. Kuk, Chris I. Ardern, Timothy S. Church, Arya M. Sharma, Raj Padwal, Xuemei Sui, & Steven Blair (2011). Edmonton Obesity Staging System: association with weight history and mortality risk Appl. Physiol. Nutr. Metab., 36, 570-576 : 10.1139/H11-058

Bookmark and Share

7 comments:

  1. Anonymous8:36 am

    I'm no scientist but I suspect that "something else out there" has to do with the fact that we used to get fat eating the food mom made us; now we get fat eating unatural chemical based products whose side effects we do not fully understand.

    ReplyDelete
  2. I like your questioning.

    One possible way to see things:
    Imagine we lived in an environment that both promoted optimal health (not unchecked corporate profit) AND did not have a tremendous amount of fat hatred.
    Perhaps those who were most genetically inclined to be fatter would be fatter, but few others would be, and also the entire population would be healthier, and there wouldn't be a dieting industry pushing weight upward.

    I appreciate the work you do to fight against fat hatred. I think that's a huge thing that can help people to be and stay in the "fit and fat" group.

    ReplyDelete
  3. When I look at these charts, I think about systems dynamic modeling. If each of those classifications are "buckets" -- the question is -- where did the people in the first set of buckets go?
    We're looking at two population snapshot, not following specific individuals over time. Analysis of a cohort followed over time would give us a better idea of what's really going on.
    But, given these two charts, people who were in the zero classification on the first time could have 1) lost weight, and not ended up in one of the classifications on the next chart; 2) died, and not ended up in one of classifications on the next chart; or 3) moved to one of the other classifications on the next chart.

    I also notice fewer people in the EOSS 3 group in the second timeframe. The same 3 scenarios could have happened to those people. In the population as a whole, there was a movement to the middle two classifications.

    So, the overall health may have both improved AND declined, depending on who you are looking at in the population.

    ReplyDelete
  4. Anonymous12:48 pm

    I am also wondering if perhaps some methodological issues or bias creeps into the NHANES data. A significant amount of the NHANES data is gathered from surveys; the increased attention paid to health & the potential connections between obesity and health outcomes may have resulted in recall or measurement bias.

    ReplyDelete
  5. Anonymous12:49 pm

    What would happen if you included not just overweight and obese I II & III people, but also normal weight people and underweight people and very underweight people.

    Maybe in all categories of weight (anorexic to very obese) the overall health of everybody declined.

    Do EOSS with people of all weights, plot that, see if the same pattern emerges.

    ReplyDelete
  6. Roman Korol9:01 am

    Anonymous might have put his/her finger on an important question, to wit, does the study encompass a sufficiently significant range of population to be able to draw conclusions.

    For instance, the consumption of HFCS has increased dramatically in the past 20 years. But fructose is very hard on the liver, intensive consumption can lead to cirrhosis (see the Lustig online video), and one of the symptoms of cirrhosis of the liver is weight loss. So in the span of time encompassed by the NHANES studies, conceivably some of the "fat and fit" folks" might have overindulged in HFCS and thereby classified themselves straight out of the parameters of these studies. End result: less fat and fit folks in the sample but, maybe, a spike in instances of cirrhosis of the liver caused by alcohol and by HFCS. Plus, over and above that, an increase in instances of the higher-category obese.

    ReplyDelete
  7. I ran your post by the statistician in our house and his comments were: the NHANES cohort studies adjusted for age and sex. There are many other possible population differences (income, education, ethnicity...) - were these adjusted for as well?
    The waist circumference difference amounted to less than 1/4 inch. The doctor measures my waist with his finger between me and the tape, the trainer by crossing the tape and subtracting the overlap, and my waist measurement varies with the time of day. Sampling and testing errors are very possible.

    ReplyDelete