Thursday, February 21, 2013

Bisphenol A and Childhood Obesity - Evidence for Reverse Causality?

The paper: Association between Urinary Bisphenol A Concentration and Obesity Prevalence in Children and Adolescents
View the abstract
The author:  Leonardo Trasande
The Journal: Journal of the American Medical Association 2012





Background:
Bisphenol A







This is bisphenol A (BPA), a chemical that's been used the in manufacturing sector for about 50 years in United States to make everything from CDs to car parts (1). Recently, BPA has received a lot of exposure (pun intended) in the media because it is found in the linings of aluminum cans, and other food products.

In the US, 92.7% of people aged 6 and above have detectable levels of BPA in their urine (1); similar numbers can be seen here in Canada from the Canadian Health Measures Survey (2). The human health effects of BPA at low, environmentally relevant doses (meaning, what we would typically be exposed to in our environment) are currently unknown. Of course, giving mice an incredibly high dose acutely can cause the mice to develop health consequences...but this can be said about almost anything at a high enough dose (See Toxicology 101: the dose makes the poison).

Finally, the primary route of exposure to BPA is oral (by mouth), and the primary source of exposure is through food containing BPA (1). The half-life is estimated to be 4-43 hours, with 24 hours being the average for most people.

Toxicokinetics:
Absorption: Readily absorbed orally (primary route of exposure)
DistributionThere is little free BPA in blood, and there may be some storage in adipose tissue
Metabolism: It is rapidly metabolized by the liver (glucuronidation)
Excretion: Rapidly excreted through the urine

The paper:

This paper uses data from a sub-sample of  the 2003-2008 National Health and Nutrition Examination Survey (NHANES), including 2838 participants aged 6-19. They first looked at correlations between BPA and weight status (overweight/obese). Then they used multivariate logistic regression to predict the odds of being obese if you are in the 1st, 2nd, 3rd, or 4th quartile of BPA concentration (so, if you're in the highest category of BPA concentration, are you at an increased risk of being obese?). They also measured three other chemically similar, but non-food related phenols as an analysis of specificity - perhaps it's just phenols in general, or perhaps there is something specific about BPA. Despite being cross-sectional, the paper is written implying forward causality (referring to the mechanisms by which BPA can modify adipose tissues, examining BPA by quartile, not BMI...etc). It's not until the end of the discussion that there is a small paragraph explaining that reverse causality may play a role too.

The study measured BPA using a one-spot urine sample. The within-day and between-day variation of urinary BPA concentration is quite large, and it would have been ideal to collect more than one sample. Seasonality might even come into play here as well - do people eat higher amounts of BPA-containing foods at different times of the year (ie/ Christmas, summertime, etc). One randomized cross-over study showed that serving people canned soup for 5 days led to a 1221% increase in urinary BPA, as compared to soup prepared without canned ingredients (3). Needless to say, multiple measurements are absolutely needed.

In lieu of multiple measurements, the authors controlled for urinary creatinine, which sort allows them to control for a dilution factor - there are two parts to a concentration measurement, the solute and and solvent, or in this case BPA and water. If you drank more water that day your BPA concentration would be lower, but this is not related to the dose, but instead to the dilution. Good on em'.

Other covariates include:
-24 hour dietary recall - they were not able to control for physical activity levels of children, so they assumed everyone to have a high physical activity level for the evaluation of caloric intake as being 'normal' or 'excessive' - a more conservative approach on their part.
-daily hours of television watching (a known correlate of obesity in children)
-serum cotinine (a metabolite of cigarette smoke) - not sure how many 6 years olds are smoking, but fair enough.
-race/ethnicity, and
-socio-economic status (education level and income)

However, they did not control for frequency of consumption of BPA-containing foods. This is a very important consideration that I'll get into later.

Essentially, they found that you are more likely to be obese if you are in the 2nd, 3rd, or 4th quartile of urinary BPA concentration, as compared to the 1st quartile, in both unadjusted and fully adjusted models (see paper for logistic regression analysis). Also the prevalence of obesity increased with the increasing quartiles. It's not exactly linear (the 2nd and 3rd groups are switched), but the authors themselves suggest that this is because of the large variation in using a one-spot urine sample.

Finally, the other three chemically similar and non-food related phenols were not associated with obesity, and did not perturb the models when they were added as covariates, both individually and together.

Up to now, it's a pretty good study. Some things they probably should have controlled for, maybe taken a few more measurements, but overall it's not bad. The interpretation is where it gets weird...

From these results the authors suggest that there is something specific about BPA that is causing it to be related to obesity. Something that other chemically similar non-food related phenols do not have.

Essentially, this is the model that the authors have come up with. BPA is related to obesity, with increased urinary BPA probably causing obesity more than obesity is causing the increase in urinary BPA. Again, it isn't until the end of the discussion that the authors finally suggest the little arrow in this relationship, saying there is a chance that this could be reverse causal as well.






So, I took some of the premises that the authors used and came up with an alternate interpretation.

Let's assume that:
1) The primary source of BPA exposure is through food, as the authors claim in the introduction. This is supported by the literature.
2) Other chemically similar, but non-food related phenols are not related to obesity.

So what we're looking for are commonly consumed foods that contain high amounts of BPA. Low and behold, an analysis using 2005-2006 NHANES data (very similar to the data that the authors here are using) suggested that “Consumption of soda is significantly associated with higher urinary BPA” (4). There is much evidence to suggest that one of the primary sources of BPA exposure in children is sugar-sweetened beverages.

Making the assumption that increases urinary BPA are not causing children to consume more sugar-sweetened beverages, if we modify the model, we get this:














The question is, are sugar-sweetened beverages associated with childhood obesity? Here, the evidence has mounted over the past several decades. Sugar-sweetened beverage consumption patterns have paralleled the rise in obesity in the United States (5), and the results  from a large meta-analysis published in the British Medical Journal in 2013 suggest that sugar-sweetened beverage consumption (along with other free sugars) is a determinant of body weight (6). It's understood in the literature that this relationship is bi-directional. If you consumer more sugar-sweetened beverages, you're more likely to be obese, and obese children tend to consume more sugar-sweetened beverages. I would argue that it's most likely the former, and that the latter is the result of behavioral patterns that the formed during the process of the former...but I'll leave this for another post.

Thus, we have the following (note the reversed arrows in the box):













Where the unidirectional association between urinary BPA and sugar-sweetened beverages, along with the bidirectional association between sugar-sweetened beverages and obesity in children, combined with the finding that other chemically similar non-food related phenols are not associated with obesity, provides evidence to suggest that the relationship between BPA and obesity may be more about reverse causality than about forward causality.

BPA and Obesity:

Forward causality?

Reverse causality?

Or both?

What do you think?

Saturday, February 16, 2013

Comments

Thank you everyone for all of the wonderful comments. I'm glad you're enjoying the content.

To answer some questions, no I did not make this myself- I use Blogger, which is an open source tool for publishing blogs.

I didn't design it myself either...I'm using one of the many templates provided by Blogger.

Hope this answers your questions.

Mike