Unit 6: Risk, Exposure, and Health // Section 4: Using Epidemiology in Risk Assessment
When scientists perform risk analyses, the best source of information on specific contaminants' health effects is data from epidemiologic studies. Epidemiologists analyze how health-related events are distributed in specific human populations—who gets sick with what illnesses, when, and where. By comparing groups with different illness rates and looking at demographic, genetic, environmental, and other differences among these groups, epidemiologists seek to determine how and why certain groups get sick. These studies are designed to inform public health policies and help prevent further harm.
Epidemiologists may consider many possible determinants to explain patterns of illness, including physical, biological, social, cultural, and behavioral factors. In each case they seek to explain associations between certain exposures, risk factors or events, and illnesses or outcomes. Over the past half-century epidemiological studies have documented linkages between smoking and lung cancer, intravenous drug use and HIV/AIDS infection, and poor indoor air quality and health problems, to cite just a few examples.
To explore these associations, analysts have two basic study design options. Cohort studies follow a group of individuals who share some common characteristic such as age, place of residence, or exposure to a hazard, and study the frequency of illness in this group to see how strongly certain risk factors are associated with becoming sick. Researchers may also follow a control group that does not share the common factor with the cohort that is the study's subject. Whether they involve one group or two, cohort studies start with exposures and follow subject through time to find the outcomes.
For example, scientists have studied survivors of the Hiroshima and Nagasaki bombings to see how atomic bomb radiation exposure affects cancer rates in survivors and the incidence of genetic effects in survivors' children. Researchers in the Framingham Heart Study, launched in 1948, have assessed over 10,000 participants from Framingham, Massachusetts, spanning several generations to identify major risk factors for cardiovascular disease (Fig. 9). Many epidemiologic studies focus on workplace exposures, which are generally higher and more frequent than other human exposures to environmental contaminants and therefore are more likely to show associations between exposure and illness.
Figure 9. Four generations from one family participating in the Framingham Heart Study and associate studies
See larger image
Source: © Tobey Sanford.
In contrast, case-control studies enroll a group of people who already have the disease of interest (the case group) and a group of people who do not have the disease but match the case group members as closely as possible in other ways (the control group). Researchers then work backwards to identify risk factors that may have caused the case group to get sick, and compare the groups to test how strongly these risk factors are associated with illness. Case-control studies start with the outcome and look backward to explain its causes.
In an early example of a case-control study, anesthesiologist John Snow investigated an 1854 cholera epidemic in London by mapping where victims lived, then marking the sites of public water pumps on the map (Fig. 10). Unlike area health authorities, Snow believed that contaminated water was a source of infection. Pump A, the Broad Street Pump, lay at the center of a cluster of cholera cases. Snow determined through interviews that other nearby pumps, which he labeled B and C, were used much less frequently than the Broad Street pump, and that all of the local cholera patients had consumed water from Pump A. Accordingly, Snow concluded that Pump A was the source of the infection. When he convinced local officials to remove the pump handle, cholera cases (which were already declining) stopped (footnote 9).
Figure 10. Snow's original map (shows cases of cholera around water pumps)
See larger image
Source: Courtesy Wikimedia Commons. Public Domain.
Each of these approaches has strengths and weaknesses. Cohort studies let researchers see how outcomes develop over long periods of time, but they require large groups to make the findings statistically significant and are expensive to administer. Case-control studies are a more effective way to study rare diseases, since researchers can select members of the exposed group instead of waiting to see which members of a cohort contract the disease, and are quicker and less expensive than cohort studies. However, since they usually look backward in time to reconstruct exposures, results may be skewed by incomplete data or participants' biased recollections.
Even if an exposure and a disease are associated, researchers cannot automatically assume that the exposure causes the disease. In 1965, pioneering British epidemiologist and statistician A.B. Hill proposed nine criteria for citing causal relationships between environmental threats and illness.
- Strength: Groups exposed to the threat have much higher rates of illness than unexposed groups.
- Consistency: The association is detectable consistently in different places, times, and circumstances by different observers.
- Specificity: The association is limited to well-defined groups, particular situations, and specific illnesses.
- Temporality: It is clear over time that the threat occurs first and leads to the outcome.
- Biological gradient: A consistent relationship exists between the size of dose and the scale of response.
- Plausibility: The proposed causal relationship makes biological sense.
- Coherence: The relationship does not conflict seriously with existing historical and scientific knowledge of the disease.
- Experiment: An experimental step (such as shutting down the Broad Street Pump) produces results that support the existence of a causal relationship.
- Analogy: The association is similar to documented causal relationships between threats and diseases (footnote 10).
What if the risk comes from a chemical that has not been studied yet, or has only been studied in a few small groups? In such cases analysts use information from animal toxicology studies, which can measure associations between contaminants and health effects in thousands of animal subjects quickly and inexpensively (relatively speaking—major animal studies can take several years and cost millions of dollars).
But animal data also has its drawbacks. Toxicology studies typically use large doses to produce a measurable response quickly, while environmental exposures usually occur at low levels over long periods of time, so analysts have to extrapolate from high study doses to low real-world doses. They also have to extrapolate from observed results in animals to expected results in humans, which assumes that a contaminant will affect humans in the same way. However, epidemiology and animal studies can inform each other. For example, if epidemiologic studies show that workers in a specific industry are developing cancer at higher than normal rates, researchers may carry out animal studies to see whether a specific material that those workers use causes illness.