Judy S. LaKind, Ph.D., President of LaKind Associates, LLC, Adjunct Associate Professor, Department of Epidemiology and Public Health, University of Maryland School of Medicine and Adjunct Associate Professor, Department of Pediatrics, Pennsylvania State University College of Medicine, Milton S. Hershey Medical Center is a health and environmental scientist with expertise in strategic risk management, assessment of human exposure to environmental chemicals and associated risks, biomonitoring, and environmental regulatory review. You can find her webpage here.
Chemicals and disease in population surveys
In the US, the number of chemicals routinely measured in people’s bodies has grown rapidly – our powerful analytical capabilities now allow us to identify over 1,000 individual chemicals in a single blood sample. Unfortunately, however, our analytical capabilities have outpaced our ability to interpret these data.
The National Health and Nutrition Examination Study (NHANES), a national survey conducted every two years by the U.S. CDC to assess health status, monitors for the presence of over 200 different chemicals in blood and urine samples. Chemical data from NHANES surveys are useful for several important purposes, such as understanding whether exposure to a particular chemical is reduced following a ban on that chemical. Many scientists have also explored the NHANES data looking for links or associations between chemical exposure and various health conditions and diseases. However, it is important to recognize the limitations of such data: studies such as these are cross-sectional, meaning that people are examined only once, providing exposure data for only a snapshot in time rather than over the long-term.
Using NHANES data to determine human exposure to chemicals is especially complicated when considering short-lived chemicals (those that remain in the body for only hours or days), including those that are the subject of widespread media attention such a bisphenol A (BPA), phthalates and benzene. If a short-lived chemical is detected in a person’s blood or urine today, what can we infer about that person’s exposure months or years ago? Unfortunately, not much. Studies like NHANES only capture information on an individual exposure for about a day and are not useful for estimating a person’s lifetime exposure. For this reason, a cross-sectional study cannot establish whether the exposure preceded the disease or whether the disease influenced the exposure.
Although the drawbacks of cross-sectional surveys are well known, some researchers have nevertheless used datasets from studies like NHANES to draw conclusions about whether a short-lived chemical is related to a chronic disease. More alarmingly, the findings from such studies are often used – largely by the popular press, but sometimes even in scientific journals – to draw causal inferences. In our recent paper, we illustrate why NHANES data on BPA and other short-lived chemicals are of limited use for drawing conclusions about links between chemical exposure and chronic diseases such as diabetes, coronary heart disease (CHD) and heart attack – diseases that have multiple causes and can take years to develop.
Interpreting NHANES BPA data: making sense of exposure
Past research on the relation between BPA and chronic disease has yielded inconsistent results. Some studies have found significant associations and others have not, depending on the disease and the year of the survey. We analyzed data from four NHANES surveys from 2003-2010 and, contrary to the findings of some previous studies, found no significant associations between BPA, CHD or diabetes. Can we then conclude that BPA is not causally associated with these diseases? No, because a current measure of a short-lived environmental chemical does not give useful information on past exposures during disease development.
Our statistical analysis of NHANES data on cholesterol levels and heart disease illustrates just how inappropriate these cross-sectional data are for this purpose. We found that cholesterol levels were associated with heart attack and CHD, but the association was opposite of what we would expect, i.e., higher cholesterol levels were associated with lower prevalence of heart attack and CHD. If the role of cholesterol in heart disease was not already understood, one might conclude that cholesterol was protective for heart disease. But, because prospective studies have demonstrated just the opposite, a plausible explanation for our result is that participants were diagnosed with heart attack or CHD before their cholesterol was measured as part of NHANES, and the diagnosis caused them to alter their diet or to use cholesterol-lowering medications.
What is needed?
The results of our statistical analyses demonstrate that data from NHANES surveys are not suitable for drawing conclusions about relations between short-lived environmental chemicals and chronic complex diseases. To understand whether short-lived chemicals or any other short-term exposures play a causal role in chronic diseases, resources are needed for more appropriately designed studies. In addition, better characterization of exposures over longer periods of a person’s life is essential. The obvious take home message is that cross-sectional studies should not be used to draw conclusions about presence or absence of causal associations between short-lived chemicals and chronic disease.