If you have an ACS member number, please enter it here so we can link this account to your membership. (optional)

ACS values your privacy. By submitting your information, you are gaining access to C&EN and subscribing to our weekly newsletter. We use the information you provide to make your reading experience better, and we will never sell your data to third party members.



Risk By Association

Scientists often start with an observed link between chemical exposure and a health risk, then design studies to confirm it

by Stephen K. Ritter
October 14, 2013 | A version of this story appeared in Volume 91, Issue 41


Credit: Center For Environmental Research & Children’s Health
A researcher conducts a neurological assessment of a girl as part of a study to gauge the impact of environmental exposure to flame retardants.
Photo of a researcher conducting a neurological assessment of a girl as part of a study to gauge the impact of environmental exposure to flame retardants.
Credit: Center For Environmental Research & Children’s Health
A researcher conducts a neurological assessment of a girl as part of a study to gauge the impact of environmental exposure to flame retardants.

You are what you eat, so the saying goes. But now that expression isn’t so simple. Synthetic chemicals are increasingly entering our bodies through our food and drink and our material surroundings. On top of that, scientific studies are showing that exercise, sleep, stress, and social support affect our biochemistry. Yet the human machine is so complex that any specific beneficial or harmful effect stemming from diet or lifestyle can usually only be guessed at.


Risk By Association

So why do we care? Most people already know that eating and living well have a bearing on one’s state of mind and overall health. But that doesn’t switch off our appetite for wanting to know more precisely whether something we consume or might be exposed to is healthy for us or whether it is going to give us heart disease or cancer, or make our kids hyperactive.

Newspapers, magazines, and Internet reports bombard us with stories on association studies that link a cause, such as exposure to a ubiquitous chemical, with an effect, such as obesity or Alzheimer’s. Advocacy groups often are quoted invoking the precautionary principle and suggesting that for public benefit the food or consumer product containing the suspect chemical should be avoided. And the chemical industry typically points out that the proven usefulness of the chemical outweighs unproven health or environmental risks. With those divergent points of view, it’s difficult to make sense of the findings.

Behind the scenes, scientists are looking more carefully at the associations and using them to guide the design of more definitive studies to solidify the cause-effect link. How researchers start with a simple association and work beyond it is a lesson in the scientific method—forming a hypothesis and then testing predictions to find proof. Meanwhile, regulatory agencies weigh the scientific evidence at hand and new scientific evidence as it becomes available to determine safe levels of exposure and whether a problematic substance should be banned or not when the answer isn’t obvious.

The process isn’t perfect. A guiding principle underlying risk assessment is that causality can’t be proven early—usually it can only be inferred with different degrees of certainty. That constraint doesn’t, however, prevent informed decisions from being made in the absence of absolute certainty.

The associations prompting the headlines typically come from epidemiology studies in which scientists sift through piles of collected data—they are often called data-mining studies. These studies have long been a cornerstone of public health research by suggesting targets for preventive medicine and clinical studies. They usually have limited value in that they gather cross-sectional data, which is to say a snapshot of data at one moment in time. Analyzing the data presents a chicken-egg problem, as one doesn’t know if the outcome being considered came before or after the associated exposure.

One of the most popular data sets for such studies is the National Health & Nutrition Examination Survey (NHANES), a program of the Centers for Disease Control & Prevention. NHANES was established in the 1960s to assess risk factors that may increase the chances of developing a certain disease or reproductive/developmental problem. It is now bulging with information from blood and urine tests, personal interviews, and survey forms. Researchers typically look broadly at the data set or select individuals on the basis of demographic information and look for an association of interest.

For example, one recent study of 766 12- to 19-year-olds found a strong association between the urinary concentration of the phthalate DEHP, which is used as a plasticizer in food packaging and medical equipment, and insulin resistance, a condition that frequently leads to diabetes (Pediatrics 2013, DOI: 10.1542/peds.2012-4022). Globally, diabetes is increasing in young people, and the research suggests environmental exposures to potentially causative substances such as DEHP should be minimized.

Researchers can’t make any claims about whether DEHP actually leads to diabetes later in life, but the data point to a concern nonetheless. To complicate matters, reverse causality is a possibility. That is, higher levels of DEHP in the teenagers might only be a biomarker for a bad diet—the packaged foods they eat might be the sole culprit. Or it could be something else entirely.

The next step in the research progression might be an observational study in which the investigators observe the subjects over time and measure their outcomes. These types of so-called longitudinal studies are more powerful than data-mining because researchers can confirm that the exposure preceded the outcome. In the DEHP case, scientists would monitor diet and health and see how many of the study participants actually develop diabetes.

To confirm the cause-effect relationship, researchers might then conduct a more rigorous interventional study, a type of clinical trial in which they treat the research subjects with a particular intervention, such as a specialized diet to follow. The treated subjects are then compared with members of a control group, who make no changes.

Credit: Proc. Natl. Acad. Sci. USA
This heat map of gene expression, generated from prostate biopsy tissue, shows that some disease-preventing genes are turned on (red, postintervention) and some disease-promoting genes are turned off (green, postintervention) following prescribed diet and lifestyle changes. Understanding gene regulation in response to environmental factors could aid risk-assessment research.
This heat map of gene expression, generated from prostate biopsy tissue, shows that some disease-preventing genes are turned on (red, postintervention) and some disease-promoting gene are turned off (green, postintervention) following prescribed diet and lifestyle changes.
Credit: Proc. Natl. Acad. Sci. USA
This heat map of gene expression, generated from prostate biopsy tissue, shows that some disease-preventing genes are turned on (red, postintervention) and some disease-promoting genes are turned off (green, postintervention) following prescribed diet and lifestyle changes. Understanding gene regulation in response to environmental factors could aid risk-assessment research.

Medical researcher Dean Ornish of the University of California, San Francisco, who directs the nonprofit Preventive Medicine Research Institute, is known for such clinical intervention studies. For more than 30 years Ornish has carried out research showing that comprehensive lifestyle changes involving diet, exercise, stress reduction, and social interaction can delay or even reverse the progression of heart disease, early-stage prostate cancer, and other health conditions (Proc. Natl. Acad. Sci. USA 2008, DOI: 10.1073/pnas.0803080105). Ornish says his body of work is showing that lifestyle changes can moderate gene expression—turning on disease-preventing genes and turning off genes that promote diseases.

Ornish recently led a team that found that lifestyle changes can promote production of telomerase, an enzyme that lengthens and repairs telomeres. Telomeres are DNA-protein complexes that cap the ends of chromosomes to protect them, similar to the plastic tips on shoelaces. As telomeres wear out, they can start to affect cell division and how quickly cells age and die. Shorter telomere length is associated with an increased risk of age-related diseases, including prostate and other cancers, heart disease, obesity, osteoporosis, and diabetes.

In work with UC San Francisco colleague Elizabeth H. Blackburn, who received the Nobel Prize for her telomere research, Ornish’s team compared two groups of men—with and without lifestyle intervention—who had been diagnosed with low-risk prostate cancer and who had not undergone conventional treatment with surgery or radiation (The Lancet 2013, DOI: 10.1016/s1470-2045(13)70366-8). The researchers measured the length of the men’s telomeres at the start of the study and again after five years. In the group that made the lifestyle changes, Ornish says, telomere length increased by an average of 10%, but in the control group, telomere length decreased by an average of 3%.

Although the study shows a cause-effect relationship and provides a plausible mechanism, there are still too many variables and potential biases to determine a precise cause-mechanism-effect chain with a strong degree of certainty. For that reason, the results are still being viewed cautiously by the medical research community.

To make the case for such cause-effect observations and pin down a mechanism, scientists turn to toxicology research studies, using lab animals or human cells to determine whether the suspected causative agent actually leads to the purported beneficial effect or disease when introduced to a healthy organism. Toxicology also has a broader role in testing new drugs and chemicals to determine possible toxicity without any known association. But harmful effects revealed in lab tests often don’t correlate to harmful effects in people, because the biochemistry is not the same or the dosing is not proportional to what people experience.

For example, toxicology studies in the late 1960s suggested that the artificial sweetener cyclamate caused bladder cancer in rats. The Food & Drug Administration banned cyclamate in the U.S., but it remained available elsewhere. The artificial sweetener saccharin, which was around before cyclamate was discovered, had also been under scrutiny for possible toxicity. In the early 1970s, saccharin was also associated with bladder cancer in lab rats. Saccharin was not banned in the U.S., but products containing it were required to carry a warning label. Over time, there’s been no strong evidence to support the idea that cyclamate or saccharin causes cancer in people. Cyclamate is still banned in the U.S., but saccharin was delisted as a possible carcinogen in 2000.

The case of artificial sweeteners points to a difficulty when it comes to evaluating risk: Risk assessment and risk management are two different but linked activities.

“Risk assessment is concerned with the nature and quality of the evidence describing a toxic effect, and it should describe the uncertainties surrounding the evidence,” explains toxicologist Paul Illing, a former U.K. government scientist and now a risk assessment consultant based in England. “Risk management is the decision-taking process associated with evaluating the evidence concerning the risk, the public attitude to the risk, possible control processes, and the costs and benefits of the decisions.”

With risk management, Illing adds, there may be a need to proceed in the presence of limited evidence of causation if the effect is likely to be serious—that is, to apply a form of the precautionary principle as a way to overprotect everyone until more definitive data are available. “But this requires that the right cause is being managed,” he says. “When the effect is mediated by some other cause, the precautionary principle’s application will be, at best, ineffectual.”

Bisphenol A, phthalates, and flame retardants are now in a similar pickle as artificial sweeteners. Among these examples, flame retardants, which are used in clothing, electronics, furniture, and building insulation, provide a good case study.

Epidemiologist Brenda Eskenazi and environmental health scientist Asa Bradman, cofounders of the Center for Environmental Research & Children’s Health at UC Berkeley, have been leading studies designed to determine how exposure to pesticides, flame retardants, and other chemicals can impact human health.

Eskenazi and Bradman’s study participants are part of the Center for the Health Assessment of Mothers & Children of Salinas (CHAMACOS, which in Spanish means young children), a long-term study group of nearly 600 primarily Hispanic women and their children living in California’s agricultural Salinas Valley. When starting out, the research team looked at existing toxicological studies to help determine which research questions they wanted to ask. Some of their early work has revealed links between polybrominated diphenyl ether flame-retardant concentrations in the blood of the women and decreased fertility, changes in thyroid hormone levels when pregnant, and low birth weight.

The scientists are currently tracking the link between exposure to flame retardants and early childhood neurobehavioral development, childhood obesity, and other outcomes. Their most recent published results, on neurobehavioral development, have found that children with greater exposure to flame retardants have lower IQ, shorter attention span, and diminished fine motor skills compared with national averages (Environ. Health Perspect. 2012, DOI: 10.1289/ehp.1205597).

The researchers measured flame-retardant levels in the blood of some pregnant women and later in their children. The children were then evaluated at ages five and seven with a battery of standardized tests to determine their motor and cognitive skills. The physical tests were supplemented with surveys completed by the mothers and teachers on behavior, learning ability, and attention span.

It is at this point, when it comes to pinning down the exact cause of the observed effects, that risk assessment research loses traction. “Epidemiological studies are observational,” Bradman says. “What we are doing is assessing the exposures and pathways to exposure. In our data analysis, we use statistical methods to control for other possible chemical exposures, so we are able to more confidently home in on the independent health effect of flame retardants.”


The team is now looking for a possible mechanism to cement the cause-effect relationship by conducting DNA methylation epigenetic studies on flame-retardant exposures in cells. But the biggest challenge in epidemiological studies, Bradman continues, is that people are exposed to a mix of synthetic chemicals, and individuals are susceptible to the exposures to different degrees. Scientists need more direct proof to verify that a chemical or set of chemicals causes the observed effect. “But we can’t go out and intentionally expose children to a chemical and see what happens,” he says.

To begin to take a hard look at causality and determine an acceptable level of use of a substance, Eskenazi says, the neurobehavioral study must be checked for consistency with multiple other studies, including other study cohorts. It also needs to be consistent with animal toxicology data and evaluated on the basis of the strength of the dose-response relationship. The work has to be considered as part of the overall weight of evidence to determine what any policy steps would be, she says, such as determining a threshold level of exposure to ensure safety.

To that end, the UC Berkeley team discusses its results in meetings with federal research and regulatory agencies, fire-safety officials, insurance industry groups, and chemical manufacturers—mostly outside the public eye. “We talk about the exposure data and the potential risks and fire safety,” Bradman explains. “This is where the information from our studies really gets vetted and used.”

“For me as a consumer, not as a scientist,” Eskenazi adds when summing up her thoughts about risk, “I always consider what the downside is. What would happen if I didn’t have flame retardants in my couch? Would I go up in a puff of smoke? If yes, then we need flame retardants. And if I wouldn’t, then why am I being exposed to something I am suspicious of, even if the science on the risk is not definitive?”


This article has been sent to the following recipient:

Chemistry matters. Join us to get the news you need.