The Ribbon Newsletter

Our knowledge and understanding of the links between environmental factors and breast cancer risk comes primarily from observational epidemiologic studies. These studies are designed to examine the association between a disease and various exposures in the physical, biological, nutritional and social environment. The complex nature of these investigations, however, often makes them difficult to understand and interpret. A broader understanding of the nature of epidemiology, as well as familiarity with the more detailed features of specific epidemiologic study designs, can help the lay person better understand the results of this type of scientific research. The purpose of this brief overview is to describe the key features of the most frequently encountered study designs, using examples of well-known studies that shape current understanding of the etiology, or causes, of breast cancer.

Epidemiology is the study of the distribution and factors involved in the development of a disease in human populations. The primary research aim of an epidemiologic study is to assess the relationship between an environmental factor (exposure) and the disease of interest (outcome). The public health objective of such studies however, is to identify those exposures which are changeable and could lead to prevention, thereby minimizing the prevalence and incidence (see definitions) of the disease in the population. This requires that epidemiologic research must strive to establish that the exposure is a "cause" of the disease. In view of these objectives, it is important to recognize that the basis of all epidemiologic evidence is statistical and probabilistic. Most epidemiologic investigations are not designed to establish cause and effect (causality) between exposure and outcome.

Prevalence: quantifies the proportion of individuals in a population who have the disease at a specific instant and provides an estimate of the probability (risk) that an individual will have the disease at a specific point in time.

Incidence: quantifies the number of new cases of disease that develop in a population of individuals at risk during a specific time.

Epidemiologic studies can be broadly grouped into two types: experimental studies and non-experimental (observational) studies. Experimental studies, including clinical and field intervention trials, are more scientifically rigorous in design in that the exposure (in this context referred to as treatment or intervention) is under the direct control of the researcher. Experimental studies may provide evidence on causality by demonstrating that the exposure is a "cause" of the outcome. Non-experimental, observational studies in contrast, can only demonstrate an "association" between exposure and outcome, and as such do not establish causality. Observational epidemiological studies can be cross-sectional (both exposure and outcome being measured simultaneously at one point in time) or longitudinal (exposure and outcome are measured or estimated at different points in time). The main types of observational epidemiologic studies on breast cancer include ecologic studies, case-control studies and cohort studies, the characteristics of which are described below.

The randomized clinical trial is perhaps the most effective type of experimental epidemiologic study used to test the effects of an intervention, such as a drug, new therapy or other medical intervention. In this type of study the subjects are usually a group of patients identified as "eligible" study subjects (those who meet the study criteria) who are randomly assigned to the experimental and control groups in the study. The clinical trial most closely resembles a controlled laboratory experiment. The investigator assigns study subjects to the treatment and control groups being compared, and thereby decides who is exposed to the treatment or intervention of interest. The main conclusions of these studies are derived on the basis of various tests of statistical significance, such as the well-known t-test. The primary methodological issues in these types of studies are concerned with avoiding systematic bias and increasing the chance of detecting a real difference between groups.

EXAMPLES: Perhaps the most familiar randomized clinical trials on breast cancer are recent studies that focused on the effects of the drug tamoxifen. Although these studies showed a significant inhibitory effect of the drug on breast cancer, the study design could not address the effects of the treatment on other outcomes and conditions. In fact, tamoxifen was later shown to increase the risk of uterine and liver cancer and is no longer considered a "miracle" drug for breast cancer. Other experimental studies however, have yielded potentially useful results for the treatment of breast cancer. For example, conclusions derived from randomized clinical trials comparing radical mastectomy versus lumpectomy have resulted in less invasive surgical interventions for breast cancer patients, i.e. many fewer mastectomies.

Confounding Factors

An important consideration in epidemiologic research is that an observed association (or lack of one) between an exposure and an outcome may be due to the effects of a third factor that is associated with the exposure and independently affects the risk of developing the disease. This is referred to as confounding. The extraneous factor is called a confounding factor or confounder.

In observational epidemiologic studies, confounding factors are most often controlled for in statistical analyses. For example, studies evaluating the effects of a pesticide exposure on breast cancer risk, would control for other breast cancer risk factors (confounding factors). Examples of confounding factors that have been controlled for in studies on chemical exposures and breast cancer risk include: age of menarche, menopausal status, reproductive history, lactation, body mass index, use of hormone therapy, diet, alcohol consumption, family history of breast cancer, and history of benign breast disease.

Studies in which the unit of observation is a group of people or a "population" (most often cities, counties or nations) are known as ecologic studies or population studies. The outcome in these types of studies is most often an incidence or mortality rate for a specific disease, and the exposure is often an overall average or index measure of the environmental factor of interest. These studies are recognized for their contributions in describing differences in populations and for generating novel hypotheses for further investigation. Amongst the main shortcomings of most ecologic studies is the lack of data necessary to control for the effects of confounding factors (see box), as well as the problem of inappropriate inferences, better known as the "ecologic fallacy."

EXAMPLES: The Cornell-China-Oxford Project is a unique ecologic study that combines data on a very large number of outcomes and a comprehensive set of environmental exposures in a singular population at a significant time in its epidemiologic history. Further, unlike other such studies, data are available for two different points in time, tracking changes and trends in the population characteristics. These unique study features address the shortcomings of other ecologic studies. Moreover, this approach allows researchers to examine the strength and consistency of multiple dimensions of a specific relationship in depth within a wider context. This landmark study provides strong evidence indicating that a long-term plant-based diet may offer the best protection for breast cancer risk. Other well-known ecologic investigations include numerous international correlational studies. For example, a widely cited, international correlational study evaluating the relationship between breast cancer and fat intake provides compelling evidence for this dietary hypothesis. This evidence however, is compromised by the fact that breast cancer may vary internationally with many determinants other than fat intake. In other words, the relationship may be confounded by other factors related to both breast cancer and fat intake, and lack of data on these other factors makes the conclusions less reliable.

Case-control studies are designed to compare people with the disease to those without the disease. In contrast to the outcome data that defines the two groups being compared, exposure data in typical case-control studies is collected retrospectively. Thus case-control studies are often referred to as retrospective studies. The majority of studies designed to examine the effects of environmental factors on breast cancer incidence are case-control studies. The advantages of case-control studies are that they are relatively inexpensive and can be completed in a short period of time, allowing researchers to evaluate data on the effects of a range of exposures on the outcome. The main statistic of interest derived from case-control studies is the odds ratio (OR), which indicates the probability of having the disease for the exposed group compared to the unexposed group. The primary methodological issues in these types of studies include issues related to the matching of cases and controls, various types of information bias, particularly recall bias, as well as use of the appropriate statistical adjustment techniques for the effects of known confounding factors.

EXAMPLES: The Long Island Breast Cancer Study Project, designed to evaluate the effects of various environmental contaminants, including organochloride compounds (OCC) and polycyclic aromatic hydrocarbons (PAH), on breast cancer incidence, is a population-based case-control study in Nassau and Suffolk counties. Data are currently being collected and analyzed by researchers and the results are expected to be published over the next few years. Three of the studies on DDE and DDT exposure and breast cancer risk, reviewed in this issue of The Ribbon, are examples of retrospective case-control studies on breast cancer.

Cohort studies are designed to compare people who are exposed to the factor(s) in question, to those who are unexposed. These studies are sometimes referred to as "follow-up" or prospective studies, because the study follows subjects over time to determine outcome differences in the two or more groups being followed. Prospective studies provide stronger evidence of causality than retrospective studies but are often more costly and difficult, especially if the outcome takes a long time to develop or if it is very rare. Unlike experimental studies, controlling the effects of other factors is done by statistically adjusting the relationship between exposure and outcome. The relative risk (RR) of getting the disease for exposed versus unexposed subjects is the primary statistic of interest that is derived directly from the results of prospective studies. Frequently, data collected in cohort studies are used to generate a nested case-control study in which exposure data is available prospectively and a sufficient number of cases have occurred for meaningful statistical analysis. This is a particularly efficient way of utilizing exposure data from biological samples collected when the study was initiated. Expensive laboratory testing is therefore conducted only for the cancer cases and appropriately matched controls, and not for the entire cohort of study subjects.

EXAMPLES: There are several well-known, ongoing prospective studies on women's health that continue to yield valuable information on various outcomes, including breast cancer. The Nurses Health Study initiated by researchers at Harvard University is a prospective study of 120,000 nurses and the results to date have provided strong evidence for the effects of reproductive factors, alcohol intakes, and body size on breast cancer risk. Other prospective studies that promise to provide valuable information for understanding the etiology of breast cancer include the Women's Health Initiative (WHI) and the New York University (NYU) Women's Health Study. The WHI focuses on the effects of low-fat dietary pattern and exogenous hormones, whereas the NYU study emphasizes the role of endogenous hormones and diet on breast cancer incidence and other diseases that affect women's health. The Agricultural Health Study is yet another prospective effort designed to study the effects of pesticides and other environmental contaminants on health outcomes including breast cancer.

The results of such ongoing epidemiologic studies as described above, while not proving direct cause, will undoubtedly broaden our understanding of the relationship between breast cancer and environmental risk factors. BCERF and all those interested in contributing to reducing the risk of breast cancer, look to all of these investigations to help guide these efforts. Written by Banoo Parpia, Senior Research Associate, Division of Nutritional Sciences, Cornell University.

Fundamentals for Interpreting Breast Cancer Epidemiology The Ribbon

Fundamentals for Interpreting Breast Cancer Epidemiology
The Ribbon