How America became obsessed with BPA
Cherry picking and over-interpretation
There are a few other reporting issues in Chemical Fallout worth noting. On September 17, the Journal Sentinel reported on a new study published in the Journal of the American Medical Association (Lang et al) which claimed that adults with higher levels of BPA metabolites in their urine had “significantly increased risk of heart disease, diabetes and liver abnormalities.”
The piece was written by John Fauber, who was not a regular contributor to the Chemical Fallout series, and he entered a note of caution in interpreting the results to mean that BPA was the cause of these diseases, driven by a similar reservations among the researchers about the limits of a cross-sectional study: “However the methodology used in the research is not definitive and the study did not provide proof that bisphenol A was the cause of the increased disease risk.” Fauber had two scientists weigh in on the study pro and contra – the former an epidemiologist who has long warned about trace exposures to chemicals, the latter warning that people who eat a lot of fast food could be simply exposing themselves to more BPA via the food packaging. In all, it was one of the most balanced, least alarmist pieces in the paper’s coverage of the chemical. But when the journalists who normally wrote about BPA for the paper, Susanne Rust and Meg Kissinger, mentioned the study, its limitations disappeared:
“A second, published today, found that bisphenol A is correlated with an increased risk of heart disease and diabetes in adults.” (September 17, 2008)
“A study published last month in the Journal of the American Medical Association linked bisphenol A to heart disease.” (October 19, 2008)
“Bisphenol A, developed as an estrogen replacement, has been linked to heart disease and diabetes in humans,” (November 1, 2008)
“A study published in the Journal of the American Medical Association in September tied BPA to heart disease in humans.” (November 16, 2008)
“BPA, used to make baby bottles and the lining of metal food cans, including those containing infant formula, is tied to reproductive failures, breast cancer risk, diabetes and heart disease.” (November 18, 2008)
“BPA has been linked to heart disease and diabetes in humans.” (December 4, 2008)
“The chemical has been linked to heart disease and diabetes in humans.” (December 16, 2008)
There turned out to be good grounds for Fauber’s more cautious evaluation of the study. On October 19, the European Union’s Food Safety Authority published a statement on Lang et al, noting that the dataset it used was:
“…an estimation of the exposure to BPA within 24 hrs of sample collection. However, there is no information on exposure during the time needed for development of diseases such as diabetes and cardiovascular conditions or changes in plasma liver-enzyme activities.
Although the study authors attempted to rule out several commonly identified confounders of studies of this type, the observed association between urinary BPA elimination and the conditions mentioned above may have been a chance finding or may be due to non-identified confounders.”
EFSA concluded that the study did “not provide sufficient proof for a causal link between exposure to BPA and the health conditions mentioned above.” The paper did not report this reaction.
In February of this year, the National Institute for Statistical Sciences published a critique of the JAMA paper by Dr. Stanley Young, Assistant Director of Bioinformatics, and an expert on data mining, and Min Yu of the University of British Columbia.
“Young and Yu note that the CDC National Health and Nutrition Examination Survey [2003-2004] that was used in Lang et al’s study measured 275 environmental chemicals and a wide range of health outcomes. Although the Lang et al study focused on one chemical and 16 health outcomes, Young and Yu note that it is important to focus on how many questions were at issue. They point out that with 32 possible health outcomes, including combinations, potentially associated with any of the 275 chemicals, along with multiple confounders and statistical models, there could be as many as approximately 9 million statistical models available to analyze the data. Given the number of questions at issue and possible modeling variations in the CDC design, Young and Yu conclude that the findings reported by the authors could well be the result of chance rather than representing real health concerns.”
Young and Yu’s criticism was published as a letter in the February 18 issue of JAMA.
The Journal Sentinel did not report Young and Yu’s criticism of the study and their warning that mining the CDC survey could throw up all manner of false conclusions. This isn’t especially surprising, as one of the general features of the way the media report scientific research is a rush to publish novel or alarming findings and a corresponding reluctance to follow up when those findings turn out to be more limited than originally advertised.
But the way the JAMA study was absorbed into the Journal Sentinel’s narrative – as with all the other research claiming increased health risks – illustrates the power of journalism’s confirmation bias: anything showing a risk is confirmation not only that there is a risk but that evidence to the contrary is not really evidence at all.