Earlier today, Dr. Leonardo Trasande and colleagues from New
York University (NYU) published
yet another in a series of economic studies which they interpret to
indicate that low level general population exposures to some brominated flame
retardants (PBDEs)1 and organophosphate pesticides (OPPs)2
are now causing a larger share of societal economic burden from IQ loss and
intellectual disability than from what they regard as the more traditional
threats of lead and methylmercury.
It is a serious allegation, and for those who conscientiously strive to ensure that currently available products are safe for consumers, it needs to be examined carefully. As this blog will demonstrate, the authors’ results and conclusions are built on a complex web of numerous unverified and questionable assumptions, and thus should be considered to be highly speculative and regarded skeptically.
What the study more convincingly shows is that prenatal and early childhood exposures to these chemicals are continuing to decline in the US as a result of both government regulation and voluntary industry efforts. This is something worth celebrating.
About the Study
To read the accompanying Press
Release issued by NYU, one could easily get the impression that the study
actually evaluated the relationship between these chemicals and cognitive brain
function and that it presents new data, but a close examination of the
published study demonstrates that neither is in fact the case3.
Instead, the authors followed a standard formula that they have repeatedly employed over the past six years and has been subjected to sharp criticism (e.g., see publications by Bolt, 2017; Bond and Dietrich 2017; Gallagher, 2015; Middelbeek and Veuger, 2015; Swaen, 2016 as well as a March 2015 BBC interview with Professor Richard Sharpe). The authors simply used previously published biomonitoring data (a one-time measurement of a few selectively chosen chemicals in blood or urine collected from a sample of Americans) readily available from the Centers for Disease Control and then made some rather bold, unverified, and questionable assumptions about: (1) causal relationships between low-level exposures to those chemicals and adverse impacts on IQ and intellectual disability; (2) the precise dose levels that might actually cause such adverse impacts; (3) the validity and representativeness of the biomonitoring data to the general US population; (4) and the economic consequences of purported levels of IQ loss and intellectual disability.
Let’s scrutinize each of these assumptions in greater detail. You should make up your own mind about their legitimacy.
Correlation Does NOT Equal Causation
The scientific evidence that the authors have relied upon, particularly
for PBDEs and OPPs, to conclude that these chemicals cause IQ loss and
intellectual disability largely derives from a few observational epidemiology
studies that are based on relatively small sample sizes (i.e., a few hundred
subjects), suffered serious methodological issues (e.g., measurement errors),
and produced inconsistent results. For
example, the two most recent epidemiology studies published in 2016 sought a
link between OPPs and IQ loss.
The first study conducted in France and the other study conducted in Ohio, found no link and were not even cited or considered by Trasande and his colleagues. As Stanford professor John Ioannidis has reminded us, the vast majority of published studies, and especially the first few that explore a particular topic, are likely to be wrong, so that independent replication by others is essential to establish a credible body of scientific evidence for a weight of the evidence assessment.
I have written in past blogs (as an example see December
9, 2019) that observational epidemiology studies can establish a statistical
correlation between suspected factors and disease, but only very rarely and under
unique circumstances can they establish a causal relationship. Observational
studies are distinguished by the fact that the investigator does not manipulate
the conditions of exposure for the subjects under study in any way, but instead
merely observes what has often been described as the “natural experiment”.
Chance, bias, and confounding are more likely to occur and must always be considered as possible alternative explanations for an observed association between an exposure and an outcome. Complicating matters in this instance, studies of IQ are particularly prone to measurement errors and must be interpreted very cautiously. Childhood IQ and neurobehavioral development are strongly influenced by the complex interaction of many other well-established risk factors (genetic conditions, alcohol and/or drug use in pregnancy, prematurity and low birth weight, childhood diseases, poverty and cultural deprivation, iodine deficiency, and others), the effects of which the researchers could not sufficiently control in their studies.
The reported differences in test scores between children with presumed high vs. low PBDE or OPP exposures were small enough to be explained by normal variation or by the interactions of other factors.
Even the NYU authors of this most recent study
acknowledge that their assumptions that these chemicals cause IQ loss and
intellectual disability may NOT be true.
However, this important admission is buried in the body of the paper and
you wouldn’t learn of it from reading the abstract, conclusions, or especially
that Press Release, which unfortunately, are the only things the vast majority
of people will ever see.
Paracelsus was right.
Assumptions About Dose-Response
In order to estimate the degree of IQ loss alleged to occur
from exposure to these chemicals in the general population, the authors extracted
limited dose-response relationships from the few observational epidemiology
studies available (often a single study), and then applied them to population
biomonitoring data. This represents an
over-reliance on sparse and therefore unreliable data that is perilous and
implies a level of precision in their data that simply cannot be justified. To their credit, the authors acknowledge that
the assumed dose response relationships they used may not apply to the entire
US population, and again this could profoundly affect their results. However, once again, this admission is buried
in the body of the publication.
Validity and Representativeness of the Biomonitoring Data
for Estimating Exposures
The authors relied on CDC’s National Health and Nutrition Examination Survey (NHANES) for the biomonitoring data (measurement of chemicals in blood and urine) they used to estimate population exposures to the chemicals of interest. NHANES is a nationwide probability sample of the US civilian non-institutionalized population aged one to 74 years. During the past two decades NHANES has oversampled various racial and low-income groups in some of their two-year cycles.
The authors of the subject study evidently combined biomonitoring data across all racial groups without justification and did not discuss how this might have impacted their results. My own cursory review of the CDC biomonitoring data left me with the distinct impression that the measurements likely differ somewhat by race and call into question the authors decision to ignore such differences in their analysis. In fact, in support of my point, ironically several of these same authors published a paper earlier this year wherein they argued that non-Hispanic blacks and Mexican Americans had higher exposures to PBDEs and OPPs and were therefore at disproportionately higher risk for IQ loss and intellectual disability.
By simply combing all of the data across racial groups for the current study it appears to contradict their own earlier findings and is inappropriate. At a minimum, they should have cited their earlier work and discussed the implications for their current study.
NHANES biomonitoring data represents a point in time
measurement of chemicals in the body.
The authors of the subject study have used it to represent a much longer
exposure period (i.e., prenatal and early childhood) which is unlikely to be
valid. Disappointingly, the authors did
NOT even acknowledge or discuss the validity and implications of making such an
assumption. CDC notes that the rate of elimination of PBDEs from the human body depends on the
specific PBDE, with elimination half-lives that can vary from a couple weeks to
several months, thus a single point in time sample may not wholly
representative of prenatal or early childhood exposures.
By contrast OPPs are very short-lived in the body with measurement
of their dialkyl phosphate metabolites
reflecting recent exposure, predominantly in the previous few days, so that
a single point in time measurement is most certainly NOT representative of
longer term exposure. Such assumptions undoubtedly
introduced errors into the study results that could have affected the
The authors restricted their analysis to biomonitoring data from women aged 20-39 for PBDE and methylmercury exposure, because starting in the 2005-2006 cycle the CDC began to pool blood samples from multiple individuals in order to increase their ability to detect the very low levels that exist. They used data for women aged 15-49 to estimate prenatal organophosphate exposure. For lead, they relied on biomonitoring data collected from children up to the age of five. My review of the biomonitoring data for PBDEs and methylmercury revealed a surprisingly small number of pooled samples in any given cycle (less than 40 across all racial groups), which raises questions about the representativeness of the measurement results for the entire US population.
CDC also warns users of their data that the pooling likely leads to overestimating exposures and underestimating the variability of the results. Again, the authors of the subject study did not discuss any of this.
With respect to the class of PBDEs, the authors had biomonitoring data for only one type, PBDE47 (a penta-PBDE which was phased out of production in the US in 2005), and made another unverified assumption that it could serve as a proxy for all types of PBDE flame retardants. The class of PBDE flame retardants represent a diverse set of chemicals that are not all the same toxicologically and prior in-depth reviews conducted by other scientists have noted that the evidence that some of them may be linked to IQ loss is more persuasive than for others. The authors assumed that measuring only one member of the class was sufficient to represent them all and that all PBDEs cause IQ loss following the same dose response curve – both of which seem highly unlikely to be true.
Since organophosphate levels were not available from the CDC
after 2008, the authors used the 2007-2008 levels to estimate IQ loss and
intellectual disability over the next 7-8 years. This could have led to overestimates of the
amount of IQ loss and intellectual disability attributable to those pesticides
use in the US significantly declined over that time period possibly leading
to lower population exposures. It seems
at best disingenuous for the authors to claim that OPP exposures will continue
to climb over time and account for greater IQ loss and intellectual disability
when they did not have any urinary metabolite data for OPPs since 2008 and chose
to use the 2008 data to extrapolate into the future.
Assumptions About Economic Consequences
For the purposes of this review I won’t assess the validity
of the sources of data or methods used to assign societal costs. However, others with specialized expertise
and familiarity with the literature on cost analysis are encouraged to do so
and publish their impressions. I would
point out that the costs of alleged lost IQ points and cases of intellectual
disability are largely due to what was considered lost productivity and owe
their roots in methodology developed by several of the same authors of this
most recent report rather than from any independent scientists. Perhaps for this reason alone they deserve
further scrutiny. These authors do
acknowledge (again in the body of the report) that the cost estimates they
applied in making their economic calculations may not be valid, because they
used data from a single point in time and costs for chronic disease are likely
to have changed over time.
To summarize, the authors of this most recent study have
made some rather remarkable allegations and conclusions about the societal
economic costs of low level exposures to chemicals, but as I’ve shown these
estimates are based on a complex web of numerous unverified and questionable
assumptions. Their findings should be
considered to be highly speculative and regarded skeptically.
Some readers may still have questions: What is the harm
if the cost estimates are grossly exaggerated? Isn’t it better to be safe than
I argue, as have others, that everyone loses when the high standards of the scientific method, which demands impartial objectivity, are not met, as demonstrated by economic papers such as this one. Not only can there be significant and lasting damage to the credibility of individual scientists and organizations, but public acceptance of science, which is already on shaky ground, becomes further eroded. Moreover, public policy choices based on poor science or isolated findings from un-replicated studies, even if well-intentioned, will have significant adverse consequences for individuals and society.
1From EPA’s 2017 Technical Fact Sheet. Polybrominated diphenyl ethers (PBDEs) are brominated hydrocarbons in which 2-10 bromine atoms are attached to the molecular structure. They have been used as flame retardants in a wide variety of products, including plastics, furniture, upholstery, electrical equipment, electronic devices, textiles and other household products. At high temperatures, PBDEs release bromine radicals that reduce both the rate of combustion and dispersion of fire. PBDEs exist as mixtures of distinct chemicals called congeners with unique molecular structures. There are three types of commercial PBDE mixtures, including pentabromodiphenyl ether (pentaBDE), octabromodiphenyl ether (octaBDE) and decabromodiphenyl ether (decaBDE). DecaBDE is the most widely used PBDE globally. The production of octaBDE and pentaBDE in the United States ceased at the end of 2004 after the voluntary phase-out of these chemicals by the only U.S. manufacturer. In 2009, the two U.S. producers and the main U.S. importer of decaBDE announced plans to phase out the compound by the end of 2013.
2According to the CDC, organophosphorus (OP) insecticides, numbering about 40 in a wide variety of formulations registered for use in the USA, are active against a broad spectrum of insects, and have accounted for a large share of all insecticides used in the United States. OP insecticides with the most usage include acephate, chlorpyrifos, malathion, naled, phorate, dicrotophos, phosmet, dimethoate, terbufos, ethoprophos, and tetrachlorvinphos. Although organophosphorus insecticides are still used for insect control on many food crops, most residential uses have been phased out in the United States in the early 2000’s. Certain organophosphorus insecticides (e.g., malathion, naled) are also registered for public health applications (e.g., mosquito control) in the United States. Usage of organophosphates has been steadily declining since 1980, and according to EPA has declined by more than 70% since 2000.
3Unfortunately, large discrepancies between the factual content of academic press releases and the underlying scientific publications which spawn them are all too common in today’s environment where academic researchers are under increasing pressure to compete for a shrinking pool of funding. This was conclusively shown in 2014 by some British researchers who studied 462 biomedical and health related science press releases issued by 20 leading UK universities in 2011 and compared them to the associated peer reviewed research papers and subsequent news stories and found:
• 40% of the press releases contained exaggerated advice
• 33% contained exaggerated causal claims
• 36% contained exaggerated inference to humans from animal studies
When press releases contained such exaggeration, 58% of the subsequent news stories also contained such exaggeration.