skip navigation
Environmental Factor, August 2015

Whole Issue PDF
This issue's PDF is still being created and should be available 3-5 business days after the first of the month. Please check back in a few days.

Researchers rise to the challenge of studying mixtures

By Virginia Guidry

Joseph Braun

“One of the greatest challenges is that there is really not a consensus yet on how we should approach these questions,” Braun said. “This workshop will help us develop a framework for selecting and applying available [statistical] tools.” (Photo courtesy of Steve McCaw)

Danielle Carlin and Cynthia Rider

Danielle Carlin, Ph.D., left, and Cynthia Rider, Ph.D., both of NIEHS, led the planning for the workshop, which built on outcomes from a 2011 NIEHS mixtures workshop. (Photo courtesy of Steve McCaw)

Real-world chemical exposures occur in mixtures, yet researchers typically study exposures to single chemicals, due to analytical limitations. NIEHS invited scientists from across the country to a workshop July 13-14, to evaluate new statistical methods for studying exposures to mixtures of chemicals in the environment.

Statistical Approaches for Assessing Health Effects of Environmental Chemical Mixtures in Epidemiology Studies brought together epidemiologists, biostatisticians, toxicologists, and exposure scientists eager to tackle the topic.

The need to study mixtures

Studies of mixtures are tricky because of the need for complicated measurements, the difficulty of determining which components are of concern, and the lack of well-established statistical methods. “It is imperative that the study of environmental mixtures be driven by our research questions and not by the statistical tools at our disposal,” said Joseph Braun, Ph.D., of Brown University during his opening talk.

“Times have changed dramatically with our computing power and our understanding of the environment and the complexities of exposure,” said Gwen Collman, Ph.D., director of Extramural Research and Training at NIEHS. She added that scientists need to use emerging tools to better understand how the web of environmental exposures may contribute to a host of health endpoints.

Meeting the challenge

NIEHS organized the workshop around a challenge to researchers. Last year, a planning committee, led by Danielle Carlin, Ph.D., of NIEHS, invited scientists to analyze three epidemiological data sets and submit abstracts describing their approaches and results. More than 30 abstracts were submitted and displayed in the poster session. The committee selected 20 of them for workshop presentations.

Two of the data sets the committee provided contained simulated data, generated by meeting planners so that the analyses would be more straightforward and the results easier to compare. The third data set came from the Health Outcomes and Measures of the Environment (HOME) study, based at the Cincinnati Environmental Health Center and funded in large part by NIEHS.

Attendees shared solutions using a variety of statistical methods (see sidebar), and lively discussion characterized the event. Veronica Vieira, D.Sc., from the University of California, Irvine, echoed the enthusiasm many attendees expressed about the opportunity to compare approaches. “[It’s great] to see so many different methods applied to the same data sets, because replication is a big deal,” she said.

The organizers commended participants for the time they spent analyzing data for the challenge. “When you consider that the environmental health community was willing to put hundreds of person-hours into this, that response says something about the field,” said Braun, as he reflected with fellow planning team member Russ Hauser, M.D., Sc.D., of Harvard University. Organizers also celebrated the involvement of both junior and senior scientists in the workshop.

Continuing development of statistical methods

The meeting served as an important springboard for determining how to analyze chemical mixtures in epidemiological studies. “I strongly support the use of more challenges like this to move the field forward,” said Gaurav Pandey, Ph.D., from the Icahn School of Medicine at Mount Sinai. He has seen similar meetings fuel advances in genomics and computational biology.

According to Carlin, the event highlighted the need for more data sets to be made available, so researchers can conduct comparisons and refine their statistical methods.

Several participants recommended keeping focused on public health research questions as new statistical methods emerge. Others emphasized the need to match statistical methods with the research questions they are best suited to answer.

Hauser reminded scientists to generate results that are useful to policymakers, a goal that may be more complicated with analysis of mixtures. “We have to remember when we publish our results, whether on mixtures or individual chemicals, ultimately, they will be used by risk assessors to protect public health,” he said.

Carlin said the organizers will prepare a report and commentary on the workshop.

(Virginia Guidry, Ph.D., is a technical writer and public information specialist in the NIEHS Office of Communications and Public Liaison.)


  • Russ Hauser
    1/9

    Hauser said the goal of the workshop was to bring researchers together to systematically evaluate different statistical approaches and methods for studying exposures to chemical mixtures. (Photo courtesy of Steve McCaw)

  • Pam Factor-Litvak and attendees at event
    2/9

    Pam Factor-Litvak, Ph.D., center, of Columbia University, was among the 160 attendees in the Rodbell Auditorium, while 75 more joined via the webcast. (Photo courtesy of Steve McCaw)

  • Birgit Claus Henn
    3/9

    Birgit Claus Henn, Sc.D., from Boston University, described her team’s use of BKMR to study one of the simulated data sets. (Photo courtesy of Steve McCaw)

  • Webster and Coull in attendence
    4/9

    Webster, seated between Gennings, left, and Coull, contributed to several of the discussions that made the workshop so productive. (Photo courtesy of Steve McCaw)

  • Changchun Xie, Ph.D. explained his technique of using LASSO.
    5/9

    Novel statistical methods were featured in oral presentations, panel discussions, and poster sessions, where Changchun Xie, Ph.D., of the University of Cincinnati, explained his technique of using LASSO. (Photo courtesy of Steve McCaw)

  • Speakers at conference
    6/9

    Coull, foreground, discusses the first simulated data set with panelists, from left, Katrina Waters, Ph.D., Pacific Northwest National Laboratory; Sung Kyun Park, Sc.D., University of Michigan-Ann Arbor; Gennings, not shown; Vieira; Shuo Chen, Ph.D., University of California, Irvine; and Ghassan Hamra, Ph.D., Drexel University. (Photo courtesy of Steve McCaw)

  • Sarah Kreidler
    7/9

    Sarah Kreidler, Ph.D., of Neptune Inc., spoke about the use of Bayesian networks for mixtures analysis. (Photo courtesy of Steve McCaw)

  • Park described shrinkage methods
    8/9

    Park described shrinkage methods used to analyze the real-world data set, including LASSO, Elastic Net, and LARS. (Photo courtesy of Steve McCaw)

  • James Nguyen, Abhra Sarkar, and Susan Teitelbaum
    9/9

    James Nguyen, right, of the U.S. Environmental Protection Agency explained his approach to Abhra Sarkar, Ph.D., from Duke University, left, and Susan Teitelbaum, Ph.D., of the Icahn School of Medicine at Mount Sinai. (Photo courtesy of Steve McCaw)



"Gibbons calls for reimagining ..." - previous story Previous story Next story next story - "Environmental influences on breast ..."
August 2015 Cover Page

Back to top Back to top