Unscientific reassurances.

In the study [PDF], the USGS scientists measured the concentrations of 260 commonly used chemicals, "including pesticides, solvents, gasoline hydrocarbons, personal care and house-hold use products, disinfection by-products and manufacturing additives." Levels were measured before and after treatment and purification.

They found low levels of about 130 of the compounds.  "Nearly two-thirds of those chemicals were also detected after treatment," the scientists report.  The data show that while treatment removes some contaminants, others, including the herbicide atrazine, are not affected by treatment.

Each of the stories found in the main stream media about this survey picked up the same metaphor from the USGS for how to think about the levels measured:  "Most of the chemicals found were at levels equivalent to one thimble of water in an Olympic-sized pool." 

This sounds like a small amount. The explicit implication is that these levels are too low to pose risks.

They may pose no risk, but in fact, current scientific understanding does not justify assuming they are safe.  The reporters could have challenged the USGS more thoroughly on their assurances of safety.

There are two big areas of scientific uncertainty that undermine the assumption of safety that would have been useful to explore in writing about these results.

The first is the question of mixtures. A chemical's effects can differ when it is mixed with other chemicals -- typical of real world exposures -- than when it is alone. One plus one may yield much more than 2 (1+1= ?), at least when comes to the impacts of contaminant mixtures on health.

The key point of this body of scientific literature is that mixtures of compounds, each at a level too low to cause an effect by itself, can together cause harm.

Experiments with mixtures show this. A set of classic experiments by Rajapaske et al. (2002) working with estrogenic contaminants showed that mixing 11 different compounds together with estrogen -- each one of the contaminants at a level too low to cause an effect by itself -- doubled the effect of estrogen on their test system.  Similar results have been reported repeatedly since then, with many different chemicals.

The USGS report noted that roughly 50% of the samples they gathered from 9 different river systems around the country had at least 15 different chemicals in them (graph to right; source, USGS report briefing, MS Powerpoint slides), and approximately 5% had at least 30 different contaminants. There is no scientific reason to think these mixtures are immune to the new math (1+1 adds up to much more than 2) that the experiments on mixtures consistently reveal.

 Whether they do remains to be determined, but current science does not justify the blanket reassurance of safety claimed by the USGS and repeated in each of the media articles.

The second point is whether high dose tests used to establish toxicity safety standards are able to predict the consequences of low dose exposures. This might have made sense years ago, but decades of research in endocrinology has shown that what happens at low doses may differ fundamentally from what is found during high dose experiments.

Many decades of research in endocrinology (the study of hormones) has shown that is what hormones cause to happen at high concentrations can't predict what they cause at low concentrations.  The suite of genes that are turned on at one level of hormone concentration can be very different than the suite of genes turned on by another, with very different effects not predictable one from the other.  In fact, high doses can cause effects exactly the opposite of low doses. 

Work during the last 10 years has revealed multiple examples in which endocrine disrupting contaminants can behave in just the same way, showing "biphasic" or "non-monotonic" dose-response curves.

All of the toxicological testing used to establish the safety of the chemicals found by the USGS survey assume that high dose testing is sufficient to identify a safe level of exposure.  None ever directly test what happens at the levels that the high dose tests predict will be safe.  It is simply assumed that those low levels are safe, yet this  assumption has been disproven many times in the scientific literature.

The USGS rationale for reassuring safety for one of the compounds measured, atrazine, is even shakier.  They report levels of atrazine in the White River (Indiana) up to approximately 5 parts per billion (ppb) in the spring in source water and roughly 2 ppb in treated water (graph to right; source).  This may sound like a low dose, but Hayes et al. (2002; 2003) have reported that an amount of atrazine 20 times lower than 2 ppb is sufficient to cause male tadpoles to grow up hermaphroditic, i.e., partly male, partly female. Hayes’s research has been challenged by industry-funded studies, which have claimed it can’t be replicated.

Dave Fahrenthold (Washington Post) challenged the USGS assurances on atrazine, writing:  "But scientists still have questions about the effects of some of the contaminants. One, the weedkiller atrazine, is being examined as a possible cause of the Potomac's "intersex" bass -- male fish that have begun growing eggs."

One additional observation about the coverage: Karen Dillon (Kansas City Star) did the best job capturing the national scope of this story.  Others covered the results from their local river.  In several there was no hint that this was a national story with data from all over the country.  Even a simple mention that the results demonstrated a consistent pattern across 9 different river systems would have helped readers put their own river in a broader context and convey the consistency of the results.