Phased Approach to Respond to Community Inquiries: Phase 3

Phase 3: Considerations for Epidemiologic Studies

Feasibility Assessment

The feasibility of designing and conducting an epidemiologic study is dependent upon a variety of factors. For that reason, a feasibility assessment should be conducted to determine whether it is practical to conduct a study that would provide meaningful answers. The feasibility assessment should address multiple issues, including, but not limited to, data availability, adequate sample size to detect meaningful differences or associations, staff capacity, and other resources. In addition, to support studies and investigations it may be important for STLT health departments to consider all sources of funding and resources that may be available to them, including funding at the federal and state or local level, as well as nontraditional sources of funding.

When it has been determined that an epidemiologic study is feasible, do the following:

• Schedule meetings to discuss the initial findings and proposed approach for further assessment with the initial inquirer and other community members.
• Provide a detailed description of the next steps for evaluation. These next steps should be based on summary findings from the analysis of the 10 criteria [PDF – 169 KB] in Phase 2.
• Develop a summary that includes a rationale for continuing the investigation. For example, the assessment identified a potential spatial cluster or environmental factors that warrant further exploration.

Establish a Community Advisory Committee

A community advisory committee (CAC) can contribute valuable sources of information and liaise with other community members. A CAC can provide specific details on the community, including help with cultural sensitivities or knowledge of current and/or historical environmental concerns in the area. A CAC can serve as a communication link back to the community at large.

Open communication will be needed to set goals, establish timelines, and discuss other issues (e.g., possible study limitations) so that expectations are clear. As previously mentioned, it may be important to also discuss issues related to resource requirements and other barriers that may impact conducting an epidemiologic study. Every community is different; however, agencies may be able to work with some existing organizations (e.g., neighborhood or religious organizations) to establish a CAC.

CAC and other community members can provide important insight and assist investigations:

• Help promote participation in studies and conduct outreach
• Identify specific tasks for community members that would contribute to the success of the investigation
• Identify additional data sources within the community (e.g., local lists that may be maintained which would identify previous residents)
• Identify population mobility patterns and housing development considerations

Consider the Following for Study Design and Protocol Development

The following sections highlight the most important considerations for study design and protocol development. Consider developing a peer review committee with external partners to assess study design issues. Committee members should have a breadth of related skill sets in epidemiology, cancer, biostatistics, toxicology, and environmental health. The final draft of the protocol should be shared with the CAC in advance of initiating the study.

Case Definition

Generate a formal case definition ahead of further analyses. Include the type of cancer, the population, the geographic area of concern, and the time period of interest. Review the study area and revise (if necessary) before starting additional analyses. For example, early assessments may have indicated that a particular contaminant was of concern and that more populations were found to be at potential risk for exposure to this contaminant. Thus, the study area would be expanded to include the other populations who were at potential risk for the exposure of concern.

Hypothesis

A hypothesis regarding cancer cases or potential environmental contaminants was likely already generated during the assessment phase. However, hypotheses can evolve during an investigation,  particularly if a potential exposure source has not been identified (30). Re-review and update hypotheses regarding the cancer cases and contaminant(s) of interest (i.e., based on observations within the assessment phase to formulate analyses a priori). The hypothesis helps to guide the collection of data, the analysis plan, and interpretation of results.

Study Population

Define the population of interest for the study. Ask the following questions to help define the study population as well as the time period of interest:

• How are the boundaries defined for the geographic area of interest?
• What are the characteristics of the population? Are there specific characteristics that are noted for the people with cancer (i.e., age group, race/ethnicity)?
• What is the latency period for the particular cancers of interest? What years of residence would need to be included to reflect the latency/development period assumed for the cancers of concern?
• How long have the residents lived in the area? What is the mobility and migration of former residents?
• Is there a potential for in utero exposures for the cancer(s) of concern based on where the mother lived during pregnancy?

Latency

Latency and change of residence add to the complexity of these investigations. Because residential history data are generally not available, collecting such data is critical as part of any epidemiologic study. Given the long latency period associated with cancers in adults, behaviors and exposures that might have contributed to the development of cancer in a person typically occur years to decades before the diagnosis. For example, malignant mesothelioma, a tumor of the lining of the lung, is associated with asbestos exposure, and the latency period between first exposure to asbestos and death from mesothelioma is often 30 years or longer (36).

Latency in an epidemiologic investigation influences the exposure period relevant to the investigation. For example, if a person with cancer did not live in the suspected area of cancer concern during the relevant exposure period, then that person’s cancer would not likely be related to an exposure in the area of concern. Conversely, the latency period might limit the ability to detect unusual patterns of cancer or identify cancers related to an environmental exposure that occurred in the past. In a mobile population, patterns resulting from an environmental contamination occurring years or even decades earlier might go undetected because exposed residents may have moved away from the community before the cancer developed. Thus, as persons move in and out of different communities, their cumulative exposure profile will change.

Because childhood cancers generally have shorter latency periods than cancers in adults, changes of residence might be less of an issue in the investigation of unusual patterns of childhood cancers. However, childhood cancer investigations may have the same limitations as adult counterparts. For example, in one California study of 380 children with a diagnosis of leukemia, approximately 65% of the study participants changed residence between birth and diagnosis (37), indicating that even among cancers with short latency periods, migration might be an important factor. Account for latency when designing any additional analyses or studies.

Additional Environmental Data

If additional environmental data are needed, partner with state or local environmental regulatory agencies. ATSDR, including the regional offices, can also provide technical assistance, such as review of environmental sampling plans. Outline the types of sampling needed and resources available or necessary to conduct such sampling. Also, consider the following:

• Potential sources and routes of exposure (e.g., through air inhalation, water ingestion, in utero)
• Contaminant transport (e.g., if exposure occurs via ambient air, meteorological factors such as wind direction and speed are important)
• Other contributing sources of exposure
• Boundaries of potential spread/movement (important to understand the potential at-risk population)
• Any prior or ongoing remediation or reduction activities related to the environmental concern of interest
• Role of participating agencies in data collection and identification of resources to support the collection of environmental data.

Existing Health Data for Case Finding

Primarily, cases of cancer among the study population are identified from a state’s cancer registry, using the case definition. Consider multiple existing data sources for use to help identify cases. Forming partnerships with health information exchange or network entities may help with identifying other existing data sources. For additional support and guidance, CDC/ATSDR can provide technical assistance.

Contributing Risk Factors

An environmental factor may have been identified as a potential risk factor for the cancer of interest; however, cancers may be caused by several different risk factors or through multiple causal mechanisms (8). Review additional risk factors for the type(s) of cancer under investigation, such as social determinants of health, behavioral risks, occupational exposures, and in some cases genetic factors.

Types of Epidemiologic Studies to Consider

Methods used in environmental epidemiology are observational, not experimental. This makes the process of identifying a causative agent more difficult because researchers may not have access to information about other exposures or other variables that can create bias in the analysis (29,30,38). However, additional studies can potentially help make inferences about cancer cases and potential associations. Consider the following types of descriptive or analytical studies:

• Case-series study: A case-series study is a descriptive analysis of persons with a similar cancer diagnosis. This study can be particularly helpful when there are small numbers of cases and it is not feasible to pursue other studies (e.g., statistical challenges associated with a small number of individuals diagnosed with the disease). Case-series studies are often designed to collect more information about each person to identify any commonalities. For these studies, identify as many relevant cases as possible to avoid selection bias. Depending on the findings, a further analytical study may be possible or warranted (see case-control or cohort below).
• Ecological study: An ecological study compares aggregated environmental data to aggregated cancer data to examine general associations. Exposures and outcomes are generally grouped to a geographic area, such as a census tract, for comparison. Individual exposures are not included in an ecological analysis. Ecological studies are most often done initially to explore potential associations but alone cannot determine causality. Epidemiologists must use caution when interpreting this type of analysis because the association with a particular environmental contaminant might not be true for individual cases, especially if there is heterogeneous distribution of the exposure over the geographic area. The related bias is known as ecological inference fallacy and could result in exposure misclassification.
• Cross-sectional study: A cross-sectional study is another descriptive study that measures health outcomes and exposure factors during a specific point in time. This type of study is helpful in describing the frequency of different characteristics in a study population during the time frame of interest. However, it may not be the ideal study design for investigating unusual patterns of cancer because this type of study design only looks at data distributions in a specific time frame, and temporal patterns cannot be established. In instances when biomarker data are available, a cross-sectional study may provide insights about key exposures among the population.
• Case-control study: Consider a case-control study when the etiology of the cancer is unknown. This type of analytical study can also be used to assess the association between cancer and a presumed exposure while being able to control for confounding factors. A case-control study collects information from cancer cases meeting the case-definition and controls within the same study area. Exposures and risk factors are compared between the two groups, for example using regression analysis methods. Case-control studies are suggested for rare cancer outcomes. Unlike the ecological study, case-control studies allow for the collection of individual-level data and risk factors to assess within the analyses. The primary disadvantages of a case-control study are the inability for individuals to recall historic events (e.g., exposures) and the difficulty of providing a direct estimate of risk. Risk is often estimated as an odds ratio, showing the odds for a particular cancer to have occurred given a particular exposure, compared to odds of the same cancer without the exposure. For this type of study, the sample size needed should be calculated ahead of time considering the power needed to detect statistical differences between the populations (Appendix A).
• Cohort (retrospective or prospective): Consider a cohort study when the exposure source is known or being investigated. Select people who have varying levels of exposure for the study. Prospective cohort studies, while one of the strongest types of studies to examine exposure-disease relationships, are very time consuming and expensive. Depending on the latency of the cancer(s) of concern, a prospective cohort study may need to be 10–20 years or longer to collect enough data for analyses. The length of time needed often must account for study attrition and thus require a very large sample size. A retrospective cohort study may be feasible, but also is reliant on time-related information. For example, a cohort study may want to compare a population exposed to chemicals from Factory X, which was operating in the early 1990s. To establish a study cohort, a population within that community would have to have been in the study area ~30 years previously. Because of the length of time needed for these studies, some challenges may arise that need to be accounted for within the study design. For example, a community may have experienced a substantial percentage of movement or migration. Also, as previously mentioned, the latency of the cancer will also need to be accounted for within the study design. For example, the average latency for mesothelioma is generally thought to be decades (39). More recently, data from the World Trade Center Health Program suggests that the minimum latency for mesothelioma is about 11 years and the minimum latency for thyroid cancer is 2.5 years (8). Studies assessing these types of cancers would have to account for the number of years that specific types of cancers develop following exposures. Additionally, limitations exist with quantifying historic exposures for retrospective cohort studies.

Ongoing technical and scientific advancements in areas such as data science, analytic and geospatial methods, and cancer genomics may be available in the future (e.g., genomic research, gene-environment interaction).

Other Considerations for Your Study

When evaluating the feasibility of different epidemiologic studies, consider the following:

• Funding: Funding may be needed to help with the investigation. Seek state and federal opportunities to assist with funding the investigation or with providing technical assistance. In addition, academic partnerships may help to support the investigation.
• Outreach: For more robust analyses, reach out to other states to discuss case-finding and/or environmental sampling activities, specifically in neighboring states if suspected environmental contaminants cross state borders.
• Communication: Ultimately, the goal of an epidemiologic investigation of unusual patterns or excesses of cancer is to understand the potential relationship between environmental and other risk factors identified and observed cases of cancer. Some epidemiologic investigations have identified potential associations between certain cancers and risk factors including exposures to certain chemicals (40–42). However, methodological limitations and data limitations (such as unknown levels of exposure) often limit the ability to demonstrate that such relationships exist. Regardless of the outcome of an epidemiologic investigation, continuing to communicate with individuals most concerned about exposures and health outcomes is important.