Media fact checking time - again. Two studies came out recently that looked at obesity rates, and reports on them are concluding that we have some discordant data on childhood obesity rates. I'm seeing articles going around saying that childhood obesity rates are dropping, while other reports pit studies against each other as if officials can't decide on what's happening to childhood obesity rates - see here, here. So what's up with Childhood Obesity?
In reality, these studies aren't saying say too much that's different, in reference to childhood obesity. However, cherry-picking a few pieces of data can paint a pretty, share-able picture of conflict over where we're at with reducing Childhood Obesity rates. This is an important issue to clear up, as having good on data on this is pretty important, to understand the impact of public health efforts, and the relative funding that should be given to reducing/maintaining the current levels. Let's take a look at them:
The first study (1), published in JAMA, came out of the CDC and used NHANES data from 9120 participants to analyze obesity's prevalence between 2003 and 2012. Their results reported no significant chances in high weight for recumbent length among infants and toddlers, obesity in 2-19year olds or obesity in adults. The overall obesity rates for 2-19 year olds was 16.9%, but they report a decrease in obesity among a subset of this group, 2 to 5 year olds, dropping from 13.9% to 8.4% (P=.03). This latter point has gotten quite a lot of attention, and I'll address it further below.
The second study (2), published in JAMA Pediatrics, also used NHANES data; however, they looked at 14years worth of data from 26,690 children ages 2-19. Their results show that obesity rates, from 2011-2012, in this age group were 17.3%. All classes of obesity rose in this age group over the past 14 years, although 2011-2012 was not significantly different than 2009-2010. They did not break up their age group data into further sub-groupings.
So what's going on with childhood obesity then- are we seeing decreases in 2-5yo and are scientists really debating as the NYT (3) makes one think? Not really. Both studies are reporting generally the same overall obesity rates in the 2-19yo age group. The major issue with the interpretations here are that reporters are looking at the decreasing obesity rates in the first JAMA study in children ages 2-5yo and concluding that these values are significant. In this case, a p value of .03 is not statistically significant. Whenever you're breaking up a group into smaller subsets, as they did with 2-19 yo (the authors partitioned them into age groups of 2-5, 6-11, 12-19), you have to correct for multiple comparisons. Using a p-value significance threshold of .05 lends itself to a lot of statistical error - think about it, I could break the 2-19 yr olds off into 20 different groups (2-4, 2-6, 2-8, 2-9, 2-10 etc etc) and one of them would've been significant by chance at the p=.05 value. When you do a Bonferonni correction to account for the fact that they split the 2-19 year old group into 2-5yos, 6-11yos, and 12-19yos, you see that a significant p value would be more like .05/3 (you could argue that since they really have 6 age groups that the correction would need to even higher, and therefore a lower p value). Either way, when you correct for multiple comparisons, this decrease in childhood obesity ages 2-5 is no longer statistically significant, and we really can't say that this decrease is representative of the population at large. Indeed, the authors of the JAMA Pediatrics article have noted in interviews that there was an odd increase in obesity rates in 2003, and the subsequent decrease afterwards created a false appearance of decline (3).
The picture the media is painting with these reports is not helpful. Spreading the sentiment that we've been wildly successful, while carrying a tone of comforting optimism, paints the wrong, overly-lax picture. Certainly, there have been some reports of improvements, both in vulnerable populations (4) and by state (5), and overall, obesity rates at the population level seem to be leveling off, but they have certainly been increasing since 1999. The big picture (pun ✓) is still pretty clear - obesity, at all ages, is a pervasive issue and needs just as much, if not more attention, from government and public health officials, as well as medical practitioners and lay individuals.
1. https://jama.jamanetwork.com/article.aspx?articleid=1832542
2. http://archpedi.jamanetwork.com/article.aspx?articleid=1856480
3. http://www.nytimes.com/2014/04/15/health/obesity-studies-tell-two-stories-both-right.html?rref=health&module=Ribbon&version=origin®ion=Header&action=click&contentCollection=Health&pgtype=article&_r=0
4. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6231a4.htm?s_cid=mm6231a4_w
5. http://pediatrics.aappublications.org/content/129/5/823.long
The first study (1), published in JAMA, came out of the CDC and used NHANES data from 9120 participants to analyze obesity's prevalence between 2003 and 2012. Their results reported no significant chances in high weight for recumbent length among infants and toddlers, obesity in 2-19year olds or obesity in adults. The overall obesity rates for 2-19 year olds was 16.9%, but they report a decrease in obesity among a subset of this group, 2 to 5 year olds, dropping from 13.9% to 8.4% (P=.03). This latter point has gotten quite a lot of attention, and I'll address it further below.
 |
| Bonferroni Correction |
So what's going on with childhood obesity then- are we seeing decreases in 2-5yo and are scientists really debating as the NYT (3) makes one think? Not really. Both studies are reporting generally the same overall obesity rates in the 2-19yo age group. The major issue with the interpretations here are that reporters are looking at the decreasing obesity rates in the first JAMA study in children ages 2-5yo and concluding that these values are significant. In this case, a p value of .03 is not statistically significant. Whenever you're breaking up a group into smaller subsets, as they did with 2-19 yo (the authors partitioned them into age groups of 2-5, 6-11, 12-19), you have to correct for multiple comparisons. Using a p-value significance threshold of .05 lends itself to a lot of statistical error - think about it, I could break the 2-19 yr olds off into 20 different groups (2-4, 2-6, 2-8, 2-9, 2-10 etc etc) and one of them would've been significant by chance at the p=.05 value. When you do a Bonferonni correction to account for the fact that they split the 2-19 year old group into 2-5yos, 6-11yos, and 12-19yos, you see that a significant p value would be more like .05/3 (you could argue that since they really have 6 age groups that the correction would need to even higher, and therefore a lower p value). Either way, when you correct for multiple comparisons, this decrease in childhood obesity ages 2-5 is no longer statistically significant, and we really can't say that this decrease is representative of the population at large. Indeed, the authors of the JAMA Pediatrics article have noted in interviews that there was an odd increase in obesity rates in 2003, and the subsequent decrease afterwards created a false appearance of decline (3).
The picture the media is painting with these reports is not helpful. Spreading the sentiment that we've been wildly successful, while carrying a tone of comforting optimism, paints the wrong, overly-lax picture. Certainly, there have been some reports of improvements, both in vulnerable populations (4) and by state (5), and overall, obesity rates at the population level seem to be leveling off, but they have certainly been increasing since 1999. The big picture (pun ✓) is still pretty clear - obesity, at all ages, is a pervasive issue and needs just as much, if not more attention, from government and public health officials, as well as medical practitioners and lay individuals.
1. https://jama.jamanetwork.com/article.aspx?articleid=1832542
2. http://archpedi.jamanetwork.com/article.aspx?articleid=1856480
3. http://www.nytimes.com/2014/04/15/health/obesity-studies-tell-two-stories-both-right.html?rref=health&module=Ribbon&version=origin®ion=Header&action=click&contentCollection=Health&pgtype=article&_r=0
4. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6231a4.htm?s_cid=mm6231a4_w
5. http://pediatrics.aappublications.org/content/129/5/823.long
Comments
Post a Comment