EPA Revisits Atrazine

Atrazine is one of the most commonly used herbicides in the U.S. and has been on the market for 50 years. The Environmental Protection Agency reviewed the safety of the weed-killing chemical in 2006 and declared it to be safe for use on corn and other crops when used as directed. But last fall the agency decided to reevaluate the potential health effects of atrazine because of recent studies that suggest an association between exposure to the pesticide and birth defects, premature births, and low birth weight in humans.

Over the course of this year, EPA’s Office of Pesticide Programs (OPP) will attempt to integrate everything there is to know about the safety of atrazine, including animal toxicity data and, for the first time, human epidemiology data, looking at both cancer and noncancer effects. The agency will then decide whether new restrictions on atrazine are necessary.

To get feedback and provide transparency to the review process, the pesticide office is holding a series of three meetings of its external scientific advisory panel. The first of those meetings was held last month to discuss how to incorporate human epidemiology and incident data into risk assessments.

The second meeting will be held in April to discuss noncancer effects associated with atrazine exposure in animal toxicity studies, and the third meeting will take place in September to examine EPA’s complete weight-of-the-evidence analysis of atrazine. That analysis will “pull it all together” and integrate information from all of the available studies, Steve Bradbury, acting director of OPP, said as he kicked off the February meeting.

In the past, EPA has generally relied on laboratory animal studies for toxicity information about pesticides because of the lack of well-designed human epidemiology studies. But several large-scale epidemiology studies, including the Agricultural Health Study and studies by the Centers for Children’s Environmental Health & Disease Prevention, funded jointly by EPA and the National Institute of Environmental Health Sciences (NIEHS), are starting to yield data that could be useful to the agency.

The studies “are likely to have bearing on our evaluation” of pesticides that come up for review in the coming years, according to Tina E. Levine, director of the health effects division of EPA’s pesticide office.

The Agricultural Health Study is a collaborative effort among NIEHS, EPA, and the National Cancer Institute. It is a prospective study, meaning exposure is assessed prior to the onset of disease. The study is tracking a wide range of health effects related to pesticide and other agricultural exposures in nearly 90,000 people, including farmers and their families in North Carolina and Iowa.

Likewise, the children’s centers studies are starting to provide data on human-health effects of pesticides. These centers have already published several studies that associate pesticide exposure with birth defects and neurodevelopmental problems. “We believe some of these studies can provide important information on the effects of pesticides on children,” Levine said at the February meeting.

Some advisory panel members raised concerns at the meeting about large epidemiology studies like the Agricultural Health Study, particularly with regard to how far data from pesticide applicators can be extrapolated to other populations. “Extrapolation to children and spouses requires a new set of data. I don’t see those data,” noted panel chair Steven G. Heeringa, a research scientist and director for statistical design at the University of Michigan Institute for Social Research.

Ecologic or retrospective epidemiology studies have also emerged since EPA’s last risk assessment of atrazine. Such studies assess exposure at the group level after the onset of disease. EPA intends to integrate some of these studies into the current review of atrazine as well as future risk assessments of other chemicals.

One challenge the agency faces is that epidemiology studies vary in quality. “Not all epidemiological studies are as robust as those of the Agricultural Health Study or of the children’s centers. We often struggle with the most appropriate approach to integrating these less robust studies into risk assessment,” Levine said.

The use of human data eliminates the need to extrapolate from animal studies, but such data “come from wide-ranging sources that vary in purpose, scope, and quality,” added Anna Lowit, a senior scientist in the health effects division of EPA’s pesticide office. “It is our job to evaluate those data and combine them with what we know about the chemical from animal studies, as well as in vitro and high-throughput assays, to make a weight-of-the-evidence finding.”

At the February meeting, EPA presented a draft framework that describes the scientific factors the agency will weigh in evaluating how human epidemiology and incident data can be integrated into pesticide risk assessments. The factors include what kind of approach was used for exposure assessment (indirect questionnaire or direct monitoring), which data sources were used, how data were collected, the statistical significance of the data, and whether exposure to other chemicals was considered.

“The framework is a tool for organizing information in an explicit way to discuss uncertainties and strengths, see places where you can improve the database, and see differences between animals and humans,” Lowit said. “It’s a great place to bring in new information like epidemiology and high-throughput data.”

Although much of the discussion at the meeting focused on atrazine, EPA officials emphasized that the draft framework being developed to organize the human and animal data for the atrazine review will likely be used for other pesticides that come up for reevaluation.

When evaluating the strengths and weaknesses of epidemiology studies, the agency considers whether the study focuses more on generating a hypothesis or demonstrating causality, Aaron Niman, an industrial hygienist in EPA’s pesticide office, told the panel members. Hypothesis-generating studies may be useful in the problem formulation stage of risk assessment, but they do not demonstrate causality, he said.

“Ecologic studies have significant limitations, but they are useful for suggesting hypotheses for future well-designed studies,” agreed panel member Ellen B. Gold, professor of public health sciences at the University of California, Davis, School of Medicine. In addition, such studies “can help provide insight into future analyses, help identify gaps, help drive the research agenda, help in the problem formulation, and help generate better research in the future to overcome some shortcomings of previous research.”

One of the limitations of epidemiology studies is a lack of validation. A lot of toxicological assays have been validated and have well-defined standards. “There’s a real need for looking at the validity of epidemiology studies,” stressed panel member Richard Greenwood, professor of environmental science at the University of Portsmouth, in England.

To demonstrate some of the key issues involved in integrating epidemiology studies into risk assessment, EPA developed a case study consisting of five ecologic and retrospective epidemiology studies that report adverse birth outcomes associated with atrazine exposure in humans. All of the studies were published since EPA’s last human-health risk assessment of atrazine.

Some panel members expressed frustration that EPA was taking the time to do a case study on a small subset of published work instead of looking at all available studies. “These aren’t all the studies that have been done on birth defects associated with atrazine. We don’t have the whole universe here,” noted panel member Frank J. Bove, a senior epidemiologist with the Agency for Toxic Substances & Disease Registry.

Panel member John S. Reif, professor of epidemiology at Colorado State University, agreed. “The weight-of-the-evidence analysis should consider all relevant studies and then triage. The whole strategy is incomplete.” EPA officials reassured the panel that it will have considered all available data by the September meeting.

EPA is also trying to integrate human incident data from poison control centers and other sources into future pesticide risk assessments. Such acute-poisoning incidents provide insight on potential short-term effects from single exposures.

In general, panel members were less than enthusiastic about the idea of using such incident data in risk assessments. “Exposure estimates in these cases are probably very limited quantitatively and of limited reliability. In addition, the incidents are reported on products, not single compounds, so the possible interactions or synergies with the main active ingredient or with other chemicals are unknown,” panel member Janice E. Chambers, director of the Center for Environmental Health Sciences at Mississippi State University, noted at the meeting.

EPA officials developed a case study focusing on the pesticide diazinon to show how human incident data can be used to inform risk assessment. The agency chose this pesticide because it has a wealth of animal toxicology data, a recent history that includes major changes in allowable exposures, and a high frequency of reported incidents.

But many members of the panel dismissed the case study as being irrelevant to the atrazine review. “It’s unlikely that what we discuss in this case of diazinon would be applicable to other pesticides such as triazine herbicides, which do not have acute toxicity,” noted panel member Chensheng (Alex) Lu, professor of environmental exposure biology at the Harvard University School of Public Health.

Although many of the panel members were somewhat critical of EPA’s approach for incorporating human epidemiology and incident data into risk assessments, several of them also commended the agency for undertaking such a difficult task.

EPA’s Lowit acknowledged the concerns about the proposed approach but warned the panel not to lose sight of the alternative: extrapolating data from rats to humans.