Use of Best Available Science Is the Best Way to Protect Health
November 19, 2013
EPA is currently addressing the challenge of updating its hazardous chemical assessment approaches. Here I identify critical components of a chemical hazard assessment that reflect the best available science and the recommendations of the National Academies. These elements should be a part of EPA’s chemical assessment paradigm, and included in the Chemical Safety Improvement Act (CSIA) now circulating in Congress as a reform of the outdated and inadequate Toxic Substances Control Act (TSCA).
The first task in a chemical assessment is to select the relevant studies for inclusion. Initially, selecting studies should be as comprehensive as reasonably possible, by gathering the published, unpublished, and “grey” literature (publicly available government reports, etc.) as part of the literature search. This broad inclusive approach is generally supported by regulators, the chemical industry, and independent researchers.
Next, the scientific quality of individual studies should be determined by evaluating possible bias in the study. This can include selection bias, performance bias, attrition bias, detection bias, and reporting bias – all of which get to the heart of the “believability” of the study results. The chemical industry and the CSIA favor reporting quality as the preferred measure of study quality. This favors industry-generated studies that adhere to Good Laboratory Practices (GLP). GLP is a standard for reporting and record-keeping, animal care and data collection required for industry laboratories in response to fraudulent practices documented in the 1970s. GLP is not necessarily associated with higher quality research, proper study design or correct statistical analysis. Moreover, GLP studies are usually designed to identify major toxic effects like weight loss or cancer or death, rather than early-warnings of potential toxicity like genetic changes or alterations in hormone levels. GLP studies aren’t designed to grapple with the problems of low-dose exposures, endocrine or hormonal effects, behavioral or learning effects, or reduced sperm count that predicts low fertility.
Since statistically underpowered studies do not have the statistical power to detect a real increase in disease risk or adverse effect in populations, they should be excluded from an assessment when they fail to find an effect because they cannot be interpreted. However, if an underpowered study does find an effect it should not be excluded, because the effect is likely real. As an analogy, if you reach into a haystack a few times (an underpowered study) and don’t find a needle (fail to find an effect), you cannot conclude whether or not there may be needles in the haystack, whereas if you do find a needle (an underpowered study that finds an effect), then there is at least one needle, and probably more, in the haystack (the effect is real).
The chemical industry routinely produces statistically underpowered studies that fail to find an adverse effect from exposure to its toxic products. This can be done by design in a number of ways, such as by using too few test subjects in a laboratory study or by mixing exposed and unexposed groups in an epidemiology study. An underpowered study is more likely to fail to find a real risk (a false negative), than to erroneously find an effect that isn’t real (a false positive) – and, that favors industry interests.
The chemical industry and the CSIA favor the requirement that studies reporting an adverse effect be disregarded – or at least treated with great skepticism - if the mechanism for that effect cannot be fully explained, or can be explained away based on mechanistic data.  Health protective restrictions and regulations have been delayed (go to www.nrdc.org to learn more about delay tactics) using this tactic for many toxic and even deadly chemicals, including asbestos, benzene, formaldehyde, phthalates, bisphenol-A, trichloroethylene, and atrazine. A documented example is the industry-promoted hypothesis that kidney tumor findings in rats with exacerbated chronic progressive nephropathy (CPN) are not relevant to humans exposed to these chemicals. In fact, a comprehensive data review across sixty chemicals associated with this type of cancer found that the industry-proposed hypothesis lacked evidence of biological plausibility, and there was an inconsistent relationship between exacerbated CPN and kidney tumor incidence across rat cancer studies.  To be health-protective, chemicals that induce kidney tumors and that also exacerbate CPN in rats should be considered to pose human cancer risks. Evidence of cancer from well-conducted laboratory studies with adequate statistical power should lead to presumptions of human risk and health-protective actions.
There are many uncertainties inherent in conducting a chemical risk assessment. Factors such as age, genetic makeup, diet, socio-economic status, and pre-existing diseases can make some individuals more susceptible to developing a health problem. Current risk assessment practices do not fully account for this variability, leaving many people inadequately protected by regulatory standards. Also, chemicals are assessed one at a time, not in combination, and the assessments do not incorporate nutritional deficiencies, pre-existing conditions, and genetic factors. When information is lacking or unreliable, regulatory agencies should use scientifically-based default assumptions (go to www.nrdc.org to find out more about risk assessment and health protective assumptions) that will protect health, rather than waiting for more data, to accelerate the chemical assessment and decision-making process. This is consistent with recommendation of the National Academies in its 2009 “Science and Decisions” report. In the absence of clear science, a chemical risk assessment should presume harm rather than innocence, to protect health and the environment. In contrast, the chemical industry routinely claims that default health-protective factors are not scientific and should not be used. The CSIA makes no mention of the need for default health-protective factors – an omission that speaks volumes to its favorable chemical industry framework.
 Conrad JW, Jr, Becker RA. 2011. Enhancing credibility of chemical safety studies: emerging consensus on key assessment criteria. Environ Health Perspect. 119:757–764. and Conrad JW, Jr, Becker RA. 2011b. Chemical Safety Studies: Conrad and Becker Respond. Environ Health Perspect. 119: a508–a509.
Becker RA, Janus ER, White RD, Kruszewski FH, Brackett RE. Good Laboratory Practices and safety assessments. Environ Health Perspect. 2009;117:A482–A483. and Becker RA, Janus ER, White RD, Kruszewski FH, Brackett RE. 2010. Good Laboratory Practices: Becker et al. 118A194–A195.A195.
 Myers, J. P., F. S. vom Saal, et al. (2009). "Why public health agencies cannot depend on good laboratory practices as a criterion for selecting data: the case of bisphenol A." Environ Health Perspect 117 (3): 309-15.
 Klitzman R. How IRB leaders view and approach challenges raised by industry-funded research. IRB. 2013 May-Jun;35(3):9-17.
 Michaels, David. Doubt is Their Product: How Industry’s Assault on Science Threatens Your Health. Chapter 6: Tricks of the Trade – How Mercenary Scientists Mislead You. Oxford Press (2008)
 Sass J, Rosenberg D. The Delay Game: How the chemical industry ducks regulation of the most toxic substances. An NRDC report. October, 2011.
 Melnick RL, Burns KM, Ward JM, Huff J. Chemically exacerbated chronic progressive nephropathy not associated with renal tubular tumor induction in rats: an evaluation based on 60 carcinogenicity studies by the national toxicology program. Toxicol Sci. 2012 Aug;128(2):346-56.
Melnick RL, Huff J. 1,3-Butadiene: toxicity and carcinogenicity in laboratory animals and in humans. Rev Environ Contam Toxicol. 1992;124:111-44. Review.
 Janssen S, Sass J, Schettler T, Solomon G. Strengthening toxic chemical risk assessment to protect human health. An NRDC Issue Paper, February, 2012