Chapter 4
Epidemiology: The Science of People

Epidemiology is the most direct method of assessing risk to humans, but like any scientific method, it has its own limitations and problems of interpretation. Knowing the principles and pitfalls of epidemiology will help you interpret epidemiological studies.

What is Epidemiology?

Epidemiology is the study of patterns of disease in human populations*. Because epidemiological studies look directly at humans rather than extrapolate from animals, they provide the most compelling evidence for measuring environmental risks to humans.

Most studies in recent decades that have linked environmental factors to human diseases were designed using principles of epidemiology.

Epidemiological studies have provided the critical evidence to link:

• toxic shock syndrome to tampon use.
• leukemia to on-the-job exposures to benzene.
• heart attacks to cholesterol.
• lung cancer, heart attacks, and low birth weight to cigarette smoking.
• Legionnaires' disease to contaminated cooling units.

Epidemiological studies provide evidence, not proof. Uncertainty is inherent in the tools that epidemiologists use. While the uncertainty can be very small, it can never be zero, because epidemiologists cannot be absolutely sure that the effect they see corresponds to the suspected cause.

Epidemiological Research

Epidemiologists compare two or more groups of people to determine what characteristics distinguish groups who get disease from groups who do not.

These distinguishing characteristics are then examined to determine how and why they are associated with disease.

Some of the characteristics epidemiologists look at are:

• consumption of certain foods.
• contact with bacteria, chemicals, or viruses.
• gender, race, or socioeconomic status.
• daily activities and behaviors.
• genetic background.
• metabolic characteristics, such as cholesterol level and blood pressure.

Risk Factors and Exposures

Epidemiologists prefer not to use the word "cause" when looking for clues to disease, because many characteristics associated with disease are not true causes.

For example, cigarette smoking is asso-ciated with heart attacks because chemicals in the smoke trigger the attacks. Race, gender, and socio-economic status also are strongly associated with heart attacks, not because they directly cause the attacks, but because they are proxies for many hard-to-define behaviors, envi-ronmental factors, and genetic factors that increase the risk of heart disease.

So epidemiologists use the term "risk factor" to describe anything that in-creases the risk of disease.Cigarettes, race, and socio-economic status all are risk factors for heart disease.

Risk factors also are called exposures. A person with a risk factor is said to be exposed; a person without that particular risk factor is unexposed.

(However, it is not usual to describe risk factors that are inherent characteristics of an individual, such as sex and race, as exposures.)

Types of Epidemiological Studies

Epidemiologists favor two types of studies for searching out risk factors for disease, case-control studies and.; cohort studies.

Case-Control Studies

Epidemiologists survey a group of people with disease (cases) and a group without disease (controls) about their histories. The survey may involve direct questioning or examination of medical or other records.

The basic question: What differs in the histories of these two groups that could explain why one is diseased and the other is not?

Example of a case-control study: In the spring of 1980, U.S. doctors diagnosed hundreds of cases of toxic shock syndrome (TSS), a potentially fatal, previously rare disease. Most cases occurred in young women during their menstrual periods. Investigators at the Centers for Disease Control questioned 50 women with toxic shock syndrome (cases) about their use of sanitary products in the month before they got sick. Then they asked each woman for the names of three friends who did not have TSS (controls), and asked them the same questions. Women with TSS were more likely than their friends to have used tampons; in particular they were almost 8 times as likely to have used one brand: Rely. This brand was withdrawn from the market in September 1980, and the incidence of TSS decreased dramatically.

Cohort Studies (Follow-up Studies)

A cohort study begins with a group of people who do not have the disease being studied. Group members differ on one or more characteristics suspected of causing the disease (for example, some may smoke while others do not). The group is followed over time to see if members with the suspect characteristic are more likely to develop the disease.

The basic question: Are the people with the suspect characteristic at greater risk of getting disease?

Example of a cohort study: To evaluate the effect of environmental lead exposure on children's IQs, researchers followed 516 children in the lead-smelting town of Port Pirie, Australia, from birth to age seven, periodically taking blood samples to measure lead levels. At age seven, children with highest blood lead levels over the years had the lowest IQs.

Which Kind of Study is Better?

Case-control studies are more common than cohort studies because they are faster and cheaper. Also, for relatively uncommon diseases like childhood leukemia, they often are the only practical way to look for causes of disease.

Cohort studies are more convincing for two reasons:

• they provide a much better opportunity to establish a cause-effect relationship because they begin with the exposure (cause) and move forward in time to the disease (effect). In contrast, case-control studies begin with the disease (effect) and look back to the exposure (cause). It is not always clear that the identified cause actually did come first.
• case-control studies are more prone to certain study design problems, such as bias or chance (see Chapter 5 ).

But cohort studies have their own drawbacks:

• they are very expensive.
• they take a long time (because they start with well people and wait for them to get sick).
• they are difficult to conduct properly because study subjects tend to drop out of the study over time.

Two other types of epidemiological studies-cross-sectional studies and clinical trials-are often in the news. While these studies serve valuable purposes, epidemiologists generally do not use them to investigate risk factors for disease.

Cross-Sectional Studies

The cross-sectional study identifies a population of interest (people in a particular neighborhood, people coming to a clinic) and asks its members about current diseases and current exposures.

Cross-sectional studies offer epidemiologists a quick way to determine whether a problem exists that warrants further study-whether, for example, workers in a particular industry have an unusually high rate of disease.

But this kind of study is not useful for establishing cause and effect because it is difficult to determine whether the exposures actually caused the disease.

Example of misinterpretation from a cross-sectional study: It is well known that cigarette smoking increases the risk of a heart attack. But if researchers did not know this and surveyed a city's residents to determine who had heart disease and who smoked, they might find that healthy people smoke more than people with heart disease. The real reason for this result is that people tend to quit smoking after they are diagnosed with heart disease. (In effect, the outcome is influencing the cause.) However, to the researchers it might appear that cigarette smoking protects against heart disease. Many cross-sectional studies suffer from this chicken-egg problem.

Clinical Trials

A clinical trial is a study done to test the effectiveness of a drug or other treatment.

Patients with a particular disease are randomly assigned to receive either the treatment under study or an inactive placebo (or the standard treatment, if one exists). Patients are then followed for a specified period to determine whether patients receiving the new treatment do better than those getting the standard treatment or the placebo.

Clinical trials are the best of epidemi-ological studies in terms of the quality of the information they provide. However, as a rule, they can't be used to explore causes of disease because it is unethical to assign people to be exposed to suspected toxins. However, such trials may be very useful for studying preventive measures, such as vaccines.

Example of a clinical trial

Estimating Risk

At the end of a study, researchers calculate the risk ratio or relative risk, by comparing the occurrence of disease in two groups-one group with a suspect characteristic, and one group without. This is the source of statements like "people who smoke are 10 times as likely to get lung cancer as people who do not."

Risk ratio close to 1 suggests the characteristic has no effect on disease.

Risk ratio greater than 1 suggests the characteristic increases risk of disease.

Risk ratio less than 1 suggests the characteristic protects against disease.

Example of risk ratios: In a landmark study, scientists followed 34,445 British male physicians from 1951 to 1961 to see if those who smoked had a higher rate of lung cancer. At the end of 10 years, the statistics looked like this:

Lung cancer death rate:

    among nonsmokers:	                                  7 per 100,000    among those smoking up to a half pack daily: 	 54 per 100,000    among those smoking up to a pack daily: 	        139 per 100,000    among those smoking more than a pack daily:    	227 per 100,000 

Dividing rates among smokers by the rate among nonsmokers yields ratios which show that, compared to nonsmokers:

Smokers of up to a half pack daily were almost 
8 times
 as likely todie of lung cancer-(54/7=7.7; the 
risk ratio
 was 7.7).Smokers of up to a pack a day were almost 
20 times
 as likely todie of lung cancer-(139/7=19.9; the 
risk ratio
 was 19.9).Smokers of more than a pack a day were more than 
32 times
 as likely todie of lung cancer-(227/7=32.4; the 
risk ratio
 was 32.4). 

Causation Criteria

If an association has been observed between an exposure and a disease, and bias, confounders, and other possible errors have been reasonably accounted for (see Chapter 5 ), then researchers can address the question of whether the association is likely to reflect a true cause-effect relationship. Some commonly used criteria are:

Strength of association . The exposure is associated with a large increase (or decrease) in the risk of disease. (The stronger the association, the less likely it is to be due to bias or an unknown confounder.)

Dose-response relationship . Higher doses of the exposure are associated with higher rates of disease.

Biologic credibility . A plausible biologic mechanism is available to explain how the exposure causes disease.

Consistency . Other studies done in different ways and in different populations have found the same association.

Time sequence . The exposure can be shown to occur before the disease.

Specificity . The exposure is associated with a specific disease. The above criteria are guidelines, not rules. Some toxicants that clearly cause disease do not meet all the above criteria. For example, cigarette smoking does not meet the specificity criterion, for it is associated with many diseases.

Cancer Clusters

When contamination is discovered in a community, citizens often look for health effects. They may notice a lot of people with cancer and conclude that this represents an unusually high incidence of disease. Public health agencies are often called upon to investigate the reported cluster-a group of individuals living in a limited area and manifesting a particular disease.

In many instances, scientists find that people have underestimated the background incidence of cancer and that the number of cancers is really just what would be expected.

In other situations, it is clear from the variety of cancers occurring that there is not a cluster that can be associated with a particular source. In the vast majority of cases, public health epidemiologists find that the suspected clusters do not represent unusual events.

Some cases are more complex than described on the previous page and it may not be possible to determine whether a cluster is present. Reasons why a firm conclusion is not possible include:

• the population is so small that any variations from population averages may be due to chance.
• it is usually impossible to reconstruct past exposures to the agent of concern to determine which cancer victims have been exposed and how much they have been exposed.
• the history of individual exposures to other possible cancer-causing agents, such as workplace chemicals or radon, cannot accurately be determined. In summary, investigations of cancer clusters are very unlikely to establish a relationship between a local contaminant and the disease. However, public health epidemiologists often undertake cancer cluster analyses in response to strong public reaction to contamination incidents and deep public fear of cancer.

Conflicting Studies

What if researchers do not agree? Reporters frequently are faced with conflicting studies. (A recent example is the question of whether alcohol consumption increases the risk of breast cancer; some studies say yes, some say no.)

One possible explanation is that one study was larger and therefore had more power to find an effect. Other possibilities are bias or confounding in one or both studies. (See Chapter 5 for discussion of how to evaluate the validity of competing studies.)

Often, there is no obvious resolution to the conflicts; they reflect the frustrating fact that most diseases have complex, intertwined causes that are difficult to tease apart. The answers come slowly, through the accumulation of research results that eventually tip the balance in favor of a particular explanation.

|| Previous Page | Next Page ||

|| Table of Contents ||

The Foundation for American Communications (FACS)
3800 Barham Boulevard, Suite 409, Los Angeles, CA , 90068
Questions? Suggestions? email us: facs@facsnet.org