You have to make a decision on whether you will use a certain product or not. Would you rather base your decision on hearsay or how much hard information there is on the product?

Most of us would answer "evidence." But what kind of evidence? And how much is enough? For some materials, like cigarettes or lead, we have direct human evidence from the medical histories of consumers or workers of the dangers of exposure over long periods of time. And we certainly know what substances can sicken or even kill quickly, which we refer to as poisons. But what about the other substances, including both natural and man-made materials, that are around us every day and which find their way into our bodies in trace amounts? How can we determine whether we should be worried or not?

Society and science have no perfect answer. Testing on humans (e.g., clinical testing of pharmaceuticals) is done in only the most controlled of conditions and after some idea of toxicity is obtained beforehand. However, there are two indirect ways science can get at possible effects of human exposure to substances in the environment. One is to observe human populations in their usual environment and investigate after the fact what factors might have caused a certain effect. The other is to conduct controlled tests on animals, primarily rodents, if there is reason to believe something in the environment may be harmful to living things. Neither way is perfect. Both ways can be inconclusive, misleading or misinterpreted. But they are currently the best tools we have. A brief look at both methods:

• A way of studying possible risks in humans by observation is called epidemiology. Epidemiologists are medical detectives, usually known to us for their efforts to track down the source and spread of a disease. Epidemiologists can take a number of approaches. One observational tool used by epidemiologists is called a "cohort study," which tracks a group of people and monitors the influences on them - diet, exposure to pollutants, personal habits, and the like. If over time they develop a disease or a medical condition, (say, lung cancer) the incidence of the disease may link up with a certain environmental influence (say, smoking). Statisticians call that a correlation, or an association. Another observational tool, called a case-control study, begins with a number of people suffering from the condition whose cause you are trying to identify. They are compared to similar people without the condition, and the epidemiologist looks for differences in lifestyle and environmental exposures (i.e., associations) that might explain the cause of the disease or condition. Both methods look for an association or correlation that relates exposure and outcome.

  Now for the hard part. Does a statistical correlation prove anything? Statisticians point out that "correlation is not causation." In epidemiology, only when the association is strong and repeatedly found in many studies does the probability of the link to a disease become convincing. There are very few sure bets in epidemiology. "Probable" is not proof. There are many cases where an association that was "statistically significant" -- that is, having only a 5 percent chance of being a coincidence - proved to be nothing but a coincidence. Among the most famous cases was the report that coffee consumption increased the risk of pancreatic cancer. More thorough follow-up studies found no link, and the health scare went away.

  Nevertheless, a high probability of a link between possible cause and effect is enough to give the medical detective a "suspect." Science can then follow up with other types of studies to help push the suspicion toward "guilt," or "innocence," and to help solve another problem in human health. For example, about 30 epidemiology studies of smoking and a number of animal experiments were conducted before the Surgeon General in 1964 said smoking causes lung cancer.

• Another way to investigate possible cause-effect linkages is with animal studies. Typically, researchers expose groups of laboratory animals, usually rodents, to a range of doses of the material in question (e.g., a chemical, or electromagnetic radiation). The higher doses are usually far higher than what humans might experience in daily life. A high enough dose of just about anything will increase the number of rodents that are affected, when compared to a control group that gets no dose at all. Researchers then look for the highest dose that has no effect on the animals over periods that can range from 60 days to two years - two years being the average life span of a rodent. The "no observed adverse effect level," or NOAEL, then becomes the basis for regulators to set the maximum permissible level of exposure for humans. It is divided by a number ranging from 100 to 1,000 to account for differences between humans and animals and to build in an added margin of safety. There may still be no proof that a certain substance is harmful at some level, but setting a safety level is the way regulators employ a careful approach to guarding public health by providing a wide margin of safety.

Epidemiology and animal testing rarely offer proof. They offer evidence, and it is the evidence that leads to decisions on how to safeguard public health. A responsible and rational regulatory framework in government is based on science and evidence, not on public or political opinion. There is a need to replicate findings in several different situations, rather than to alarm the public by jumping to conclusions on the basis of the results from just one study. The scare stories, fad diets, and miracle cures we hear about every day do not serve public health and safety; in fact, they often do the opposite. The tools available to public health officials to separate actual risk from needless fear are not perfect, but they produce evidence, and it is the weight of that evidence which helps science to separate the real from the imaginary.

Last Updated: October 16, 2003