Three Reasons Why The 40% Drop in Infant Obesity Means Less Than Some May ThinkMonday, March 3, 2014
Even the casual reader will not have escaped the recent exuberant media reports that, when it comes to childhood obesity rates in the US, the worst may be over.
There was certainly also no shortage of folks claiming credit for this “success”.
Interestingly enough, this enthusiasm was not shared by Cynthia Ogden and colleagues in their JAMA paper on the Prevalence of Childhood and Adult Obesity in the United States, 2011-2012, came to the following conclusion:
“Overall, there have been no significant changes in obesity prevalence in youth or adults between 2003-2004 and 2011-2012. Obesity prevalence remains high and thus it is important to continue surveillance.”
Indeed, the overhyped media report was based on an isolated subgroup analysis in infants and toddlers (2 to 5-year olds), where obesity rates dropped from from 13.9% to 8.4% a difference that barely scraped statistical significance (P = .03).
No changes or increases were seen in every other subgroup with rates as high as 16.9% in 2- to 19-year-olds and 34.9% in all adults 20 years or older. In women aged 60 years and older, obesity rates actually increased from 31.5% to 38.1% (P = .006).
So here are my three reasons why these results of the putative drop in toddler obesity mean almost nothing.
1) This is a subgroup analysis of a subgroup – anyone with any knowledge of statistics should treat any such results with extreme caution (as clearly the authors themselves do).
2) We are talking infants and toddlers – this is not the group where obesity powerfully tracks into adulthood (or even into adolescence).
3) Infants and toddler don’t generally make “independent” behavioural decisions (although parents of young kids will agree that this is not at all what life with toddlers feels like). Thus, it is entirely unclear what “behaviours” of these toddlers will carry over as they get older and become more independent.
If the findings are at all true and representative, then at best, they suggest that parents of very young kids may be “getting the message”, though exactly what that “message” is and what exactly the parents are doing about it is quite unclear to me.
On a happier note, overall obesity rates appear to have stabilised at least in the US. Perhaps we have now reached that point in the epidemic, where everyone who can be obese is obese.
Ogden CL, Carroll MD, Kit BK, & Flegal KM (2014). Prevalence of childhood and adult obesity in the United States, 2011-2012. JAMA : the journal of the American Medical Association, 311 (8), 806-14 PMID: 24570244
Monday, March 3, 2014
I have never seen a definition of childhood obestiy that made sense. Most definitions talk about the 95th percentile of some “fatness” measure. For instance, a kid will be classified as obese if he is above the 95th percentile of BMI for his age.
That means, by the very definition of percentiles, that exactly 5% of children are obese. In any circumstances, exactly 5% will be above the 95th percentile. That is what percentiles mean!
I am left to guess that the references for BMI-for-age come from old statistics, which might or might not get updated, and might or might not be about the population studied. For all I know, the percentage of obese children changed because they updated their reference tables (using less outdated statistics, though still not representing the current population they are studying.)
How can they claim obesity rates different from 5% when using the 95th percentile as a definition?
Tuesday, March 4, 2014
Valerie – The answer to your question is complicated, but one key part of the answer is that yes, the percentiles are not true percentiles — some data are excluded because of concerns that otherwise the upper percentiles would be too high and would not falsely identify overweight/obese kids as having a healthy weight. (I am just explaining, not agreeing or disagreeing with this approach). They used CDC curves for the Ogden study. The CDC curves were published in 2000. Weight data for kids over 6yo collected after 1980 were excluded from the data used to make those percentiles.
The other main percentiles in use now are the WHO percentiles. They use a different method for data exclusions but also exclude data in order to lower the upper percentiles.
Even without data exclusions, percentiles are always made with samples of a population, and studies use data from other samples of a population. And often the samples are not even from the same population. Some of these samples, like for the CDC curves, are fairly close to random samples of a population. For most studies, including WHO, the selection of samples is not random at all. So there are lots of reasons why a study would not show 5% of people above the 95% percentile. In fact, it’s pretty rare that things fit so cleanly.
Tuesday, March 4, 2014
Thank you Carrie for taking the time to answer my question. This is actually the first time I read any explanation regarding those “percentiles” (though I see them used in reasearch and media all the time, without anyone pointing out the inconsistencies).
So, if I get it right, the definition for childhood obesity is outdated, it does not apply to the populations studied, the numbers were fudged to remove to inconvenient data points, there are (at least) two possible references, it uses the debatable BMI benchmark, but (silver lining?) the definition is the same year after year. So, the absolute numbers are questionable, but the trends should reflect a true phenomenon (assuming that the researchers use the same reference [CDC vs. WHO] in all their papers).