Wednesday, March 5, 2014

Are we really heading in the right direction with childhood obesity? (and a fun lesson in epidemiology)

I saw some very encouraging headlines last week about childhood obesity, not the least of which came in the NY Times:

"Obesity Rate for Young Children Plummets 43% in a Decade"

I've also come across graphs like this one, which support the narrative (Note this graph is for obesity only, i.e. 95th BMI percentile - More on the distinctions later).


Whenever I see headlines like this though, my first instinct is to immediately go to the actual study, which in this case was published in the Journal of the American Medical Association (or JAMA). After looking at all the data, the story is often MUCH more nuanced than gets spun in the media. This study was no different. More on this in a second.

The other thing that generally happens around these type of epidemiological studies is that speculation abounds about what CAUSED the particular change. These type of prevalence studies essentially take a static picture of a particular health concern (in this case elevated body mass index - the common marker used to assess overweight and obesity) within a given population (a nationally representative sample of children in the US ages 2-19 for this study). If the methods are consistent over time, you can then compare what today's picture looks like compared to the picture, say five years ago. But that's all you can do - compare the pictures. What often happens, of course, is commentators begin to offer different theories as to what caused the changes in the pictures, like in this Washington Post blog. It's important to remember these are just speculations. These studies were never designed to actually assess this question of causation.

We also see the unfortunate occurrence of over-generalization. This is when the findings of a particular population are extrapolated to other populations (this is a big no-no in epidemiology), like in this article that references "youth-related" obesity, but cites the decline found in 2-5 year olds (i.e. children not youth). 

Ok, let's get to the data.


The following are a series of graphs I put together using the same NHANES data published in the most recent JAMA article, but instead disaggregated by race. These charts tell a much more nuanced story, particularly as there is a significant race/ethnicity component to obesity in the US. Another modification I made was to look back to 1999-2002 to provide additional context on trends (instead of 2003, which the above graph used). The reference point for trends provides a lot of valuable information and shapes how the results are framed.

Let me show you why.

These are prevalence data for overweight and obesity (measured as a BMI of greater than or equal to the 85th percentile and 95th percentile respectively) for several different age groups disaggregated by race/ethnicity.

Overweight and Obesity (85th percentile), 2-5 years old

Obesity (95th percentile), 2-5 years old

*You'll notice from this second graph that an interesting story emerges, one that is masked when only looking at aggregated data. Though there are some significant declines among certain populations (most notably non-Hispanic whites), there is a steep upward trend for Hispanics.

Overweight and Obesity (85th percentile), 6-11 years old

Obesity (95th percentile), 6-11 years old

*Here, I think it's important to think about the reference point, especially in the first graph for overweight and obesity among 6-11 year olds. Though a little progress here and there, all categories have a higher prevalence in 2011-12 than in 1999-2002. The graph would tell a different story if the reference was 2003-04, as in the first graph in the article. We also see another age group where prevalence of overweight and obesity increased most recently among Hispanics (in addition to non-Hispanic whites).

Overweight and Obesity (85th percentile), 12-19 years old

Obesity (95th percentile), 12-19 years old

*Again, we see a similar story to the 6-11 age group. There is considerable variation when looking across race. In fact, we see an overall increase from 1999-2002 to 2011-12 when looking at all races within this age category.

Take Home Message

My take of the data is that it's a much more nuanced story than what's being portrayed. Yes, there are some positive trends. But these are within specific age groups and looking at specific markers (i.e. 85th percentile of BMI vs. 95th percentile). Looking at what happens later in childhood and into adolescence, the trends in obesity aren't as favorable. In a sense, gains early in life (2-5 years old) could be offset by what's occurring in some of the older age groups. And, just because a national-level trend is positive (all kids of a particular age), doesn't mean it's positive for all the different sub-groups (in this case broken down by ethnicity) that make up the national average.


All data in the blog were produced using the following:

Hedley AA, et al. Prevalence of Overweight and Obesity Among US Children, Adolescents, and Adults, 1999-2002. JAMA. 2004;291(23):2847-2850.

Ogden CL, et al. Prevalence of High Body Mass Index in US Children and Adolescents, 2007-2008. JAMA. 2010;303(3):242-249.

Ogden CL. Prevalence of Childhood and Adult Obesity in the United States, 2011-2012. JAMA. 2014;311(8):806-814.

Disclaimer: The views in this article are my own and don't necessarily reflect those of my employer or my employer's clients.

No comments:

Post a Comment