Analysis of Cancer Risks in Populations Near Nuclear Facilities: Phase 1 (2012)

Summary

The U.S. Nuclear Regulatory Commission (USNRC) requested that the National Academy of Sciences (NAS) provide an assessment of cancer risks in populations near USNRC-licensed nuclear facilities that utilize or process uranium for the production of electricity (see Sidebar 1.1 in Chapter 1 for the complete statement of task). These facilities presently include 104 operating nuclear reactors at 65 sites in 31 states and 13 fuel-cycle facilities in operation in 10 states. The operating fuel-cycle facilities include four in situ uranium recovery facilities, one conventional uranium mill, one conversion facility, two uranium enrichment facilities, and five fuel fabrication facilities (see Sidebar 1.2 in Chapter 1 for a description of these facilities). There are additional state-licensed conventional uranium milling facilities and in situ leaching facilities.

This USNRC-requested assessment is being carried out in two consecutive phases. The focus of the Phase 1 scoping study, which is the subject of this report, is to identify scientifically sound approaches for carrying out an assessment of cancer risks associated with living near a nuclear facility. The results of this Phase 1 study will be used to inform the design of the cancer risk assessment, which will be carried out in Phase 2. This report provides the committee’s judgments about the strengths and weaknesses of various study approaches; these approaches differ in their broadness of approach, anticipated statistical power, ability to assess potential confounding factors, possible biases, and required effort.

Three findings and three recommendations emerged from this study. These are presented and discussed below. Additional supporting information can be found in the report.

FINDING 1: There are several challenges for carrying out epidemiologic studies of cancer risks in populations near U.S. Nuclear Regulatory Commission-licensed nuclear facilities in the United States, including the following:

• Uneven availability and quality of data on cancer mortality and incidence at geographic levels smaller than a county. Cancer mortality and incidence are tracked by individual states, and the availability and quality of data varies from state to state. In general, cancer mortality data are available electronically from about 1970, but subject address at time of death is not captured until much later in some states. (In the absence of subject address at time of death, mortality data cannot be geocoded at levels of geographic interest for an epidemiologic study, such as census tracts.) Cancer incidence data of known quality are generally available from about 1995, although such data are available for earlier times in some states. These data include address at time of diagnosis and have been widely geocoded, although there are residual problems associated with post office boxes and rural delivery addresses.
• Uneven availability and quality of data on nuclear facility effluent releases. Effluent release data may not be available and data quality may be poor for some nuclear facilities. Effluent releases from many nuclear facilities were much higher in the past and their radionuclide compositions have changed over time. Uncertainties in dose estimates may be much higher in years when effluent releases were highest.
• Inability to reliably capture information on population mobility, risk factors, and potential confounding factors. There is no centralized source of information on residential histories or lifestyle characteristics of individuals who live in the United States. The U.S. Census provides decadal snapshots of some population characteristics, including population size and distribution with respect to age, race/ethnicity, gender, educational level, and income. However, data on population lifestyle risk factors, including exposure to cigarette smoking and access to healthcare, are limited to state-level health surveys and are not consistently available from state to state at the same level of resolution. Moreover, populations near nuclear facilities receive radiation doses from multiple sources that are unrelated to facility effluent releases, for example, doses from natural background radiation and medical radiation. There may be other risk factors and potential confounding factors, for example, exposures to toxic chemicals and unidentified lifestyle factors, that can influence cancer risks.

• Low expected statistical power. Doses resulting from monitored and reported radioactive effluent releases from nuclear facilities are expected to be low. As a consequence, epidemiologic studies of cancer risk in populations near nuclear facilities may not have adequate statistical power to detect the presumed small increases in cancer risks arising from these monitored and reported releases.

The committee paid close attention to these challenges as it assessed the scientific merit of various epidemiologic study designs.

FINDING 2: An assessment of cancer risks in populations near nuclear facilities could be carried out using several study designs. Each design has strengths and limitations for estimating cancer risks.

• Risk-projection models estimate cancer risks by combining population radiation dose and/or dose surrogate (e.g., distance and direction from a nuclear facility) estimates with risk coefficients derived from epidemiologic studies of other exposed populations, for example, Japanese atomic bombing survivors. Risk-projection models can be used to estimate population-based cancer risks for any facility type, population size, and time period. However, because risk estimates are based on extrapolations from other epidemiologic studies and not on actual cancer incidence and/or mortality rates in populations near nuclear facilities, risk-projection models cannot be used to assess whether any predicted excess cancer risks correspond to observed patterns of cancer incidence or mortality.
• Ecologic studies estimate cancer risks by comparing observed cancer incidence and/or mortality rates in populations, considered as a group rather than as individuals, as a function of average radiation doses and/or dose surrogates for those populations. This design allows for the study of multiple cancer types during past and recent times, which helps to improve statistical power and provides a comprehensive picture of cancer risks. However, ecologic studies involve a large number of comparisons among population age groups, nuclear facilities, years of operation, and cancer types. This can lead to false associations resulting from chance alone. Moreover, ecologic studies can account only for population characteristics and potential confounding factors using group averages that are available from the decennial census and from survey information that can be linked to the census data (such as the American Community Survey). Individual characteristics can diverge sharply from group averages.
• Cohort studies estimate cancer risks by following individuals for a

• specified period of time to determine the rate or risk of cancer as a function of doses and/or dose surrogates. In a prospective cohort study, subjects are followed from the present to a future time; in a retrospective cohort study, subjects are followed from a past time to a more recent time, usually via records. Prospective cohort studies can in principle provide the least-biased estimates of associations of multiple cancer types and radiation doses and/or dose surrogates compared to studies that rely on retrospective collection of information, such as case-control studies (described below) or retrospective cohorts. However, prospective cohort studies need to follow subjects for long time periods and could therefore require decades to complete. Retrospective cohort studies are more efficient than prospective studies because the follow-up period has already occurred. However, such studies rely on linkages such as those between birth certificates and state cancer registries; logistical and administrative barriers to such linkages could limit the feasibility of this study design in some states. Moreover, in- and out-migration issues need to be considered.
• Case-control studies estimate cancer risks by comparing radiation dose and or dose surrogates between individuals selected because they have (cases) or do not have (controls) cancer. The individuals under study and cancer outcomes of interest must be predefined and for practical reasons would be limited to one or a few cancer types (for example, pediatric cancers). A challenge in case-control studies is to select suitable controls in a way that does not bias the study results.

In the absence of information on residential history, most studies by necessity make assumptions about relevant exposures based on information about location of residence at one time point in the lifetime of the study cases, such as place of residence at time of birth or place of residence at time of diagnosis or death, with the equivalent time for controls. This single time point of place of residence may not be the most relevant regarding exposure from the nuclear facilities.

Studies that are based on individuals, such as cohort and case-control studies, can potentially provide stronger evidence for or against an association between radiation exposure and cancer compared to an ecologic study that is based on groups of individuals (i.e., populations). However, such studies are likely to involve fewer cancer cases than an ecologic study due to the effort involved in subject selection and individual data collection. The effort involved in conducting a cohort or a case-control study could be reduced by partnering with existing multistate cancer studies that have already linked cancer and birth registration data.

Case-control studies can involve contacting subjects to collect residential history and lifestyle information through interviews and questionnaires. Such studies would need to be limited to recently diagnosed cancer cases (i.e., diagnoses made during the past 5 years) and would likely be subject to additional selection and information biases. There are added difficulties in obtaining appropriate approvals from the cancer registries before subjects could be contacted. However, such studies can also be carried out without subject contacts by using information from birth and other administrative records.

FINDING 3: Effluent release, direct exposure, and meteorology data, if available, can be used to obtain rough estimates of annual variations in dose as a function of distance and direction from nuclear facilities.

Effluent release and direct exposure data collected by facility licensees are likely to be sufficiently accurate to develop a population-level dose reconstruction that provides rough estimates in annual variations in dose as a function of distance and direction from nuclear facilities. However, such data would not be sufficient to support detailed reconstructions of doses to specific individuals living near nuclear facilities. However, it will be necessary to develop a methodology for estimating releases of carbon-14 prior to 2010 to support dose estimation (carbon-14 may be a significant contributor to dose from nuclear plant releases, especially in recent years). Moreover, facility-specific evaluations will be required to determine the quality and availability of effluent release and meteorology data as well as meteorology data for batch releases. Obtaining and digitizing effluent release and meteorology data for use in an epidemiologic study will be a large and costly effort.

Environmental monitoring data have limited usefulness for estimating absorbed doses from effluent releases around nuclear plants and fuel-cycle facilities. Almost all environmental measurements reported by facilities are either below the minimum detection limits or are not sensitive enough to allow for the development of useful dose estimates.

Computer models have been developed to estimate absorbed doses resulting from airborne and waterborne radioactive effluent releases. These models combine information on effluent release timing and magnitude, transport of the released effluents through the environment, and the exposure of individuals to radiation from these releases to estimate absorbed doses. Such models could be used to obtain rough estimates of doses to support an epidemiologic study. An existing model could be adapted for this purpose or a new model could be developed. Regardless of the approach used, it is essential that the model reflect modern practices for dose reconstruction, including approaches for estimating uncertainties.

Absorbed doses near nuclear facilities are anticipated to be low, in most cases well below variations in levels of natural background radiation in the vicinity of individual facilities. Absorbed doses are also anticipated to be below levels of radiation received by some members of the public from medical procedures and air travel. Consequently, dose estimates used in an epidemiologic study would ideally account for these other sources of radiation exposures and possibly for other risk factors such as exposure to hazardous (and potentially carcinogenic) materials released from nearby industrial facilities.

RECOMMENDATION 1: Should the U.S. Nuclear Regulatory Commission decide to proceed with an epidemiologic study of cancer risks in populations near nuclear facilities, the committee recommends that this investigation be carried out by conducting the following two studies, subject to the feasibility assessment described in Recommendation 2: (1) an ecologic study of multiple cancer types of populations living near nuclear facilities and (2) a record-linkage-based case-control study of cancers in children born near nuclear facilities.

Brief descriptions of these recommended studies are provided below. A list of strengths and weaknesses of the recommended studies and additional details on the study designs can be found in Chapter 4.

The ecologic study should assess cancer incidence and mortality in populations within approximately 50 kilometers (30 miles) of nuclear facilities for the operational histories of those facilities to the extent allowed by available data. A study zone of this size would incorporate both the most potentially exposed as well as essentially unexposed regions to be used for comparison purposes. The study should examine all relatively common cancer types by age interval and gender, including cancers that are not considered to have a radiogenic origin (presumed nonradiogenic cancers such as prostate cancer can serve as useful negative controls) and also take into account temporal changes in estimated radiation dose. A subanalysis should specifically be carried out for highly radiogenic cancers such as leukemia in children. The study should examine associations between (i) cancer and distance and direction from the nuclear facility and (ii) cancer and estimated radiation dose, both at the census-tract level. The committee recommends that absorbed doses to individual organs be estimated using the methodology outlined in Chapter 3.

The record-linkage-based case-control study should assess the association of childhood cancers (diagnosed at younger than 15 years of age) in relation to maternal residential proximity at the time of birth of the child, among those whose address at time of delivery was within a 50-kilometer radius of a nuclear facility. The study period for individual facilities should

be based on the quality and availability of cancer registration in each state. Controls born within the same 50-kilometer radius as the cases should be selected from birth records to match cases on birth year at a minimum. Absorbed doses and/or dose surrogates should be based on address of the mother’s place of residence at time of delivery, as determined from birth records.

These recommended studies are complementary: The ecologic study would provide a broad investigation of both cancer incidence and mortality over the operational histories of nuclear facilities to the extent allowed by available data. The analysis will be based on place of residence at time of cancer diagnosis or at time of death from cancer. The committee’s recommended approach for carrying out this study would improve on the 1990 National Cancer Institute survey¹ (these improvements are described in Chapter 4). The record-linkage-based case-control study of childhood cancers would attempt to provide a more focused assessment of the association of these cancers in relation to early life exposure to radiation during more recent operating periods of nuclear facilities. An analysis based on maternal residence at time of delivery of the child may be considered more appropriate for capturing relevant exposures.

The committee has recommended these two studies based primarily on scientific merit, feasibility, and utility for addressing public concerns about cancer risks. However, the decision about whether to carry out one or both of these studies is the responsibility of the USNRC. In making this decision, the Commission will consider a number of factors, some of which are outside the charge for this Phase 1 study such as cost and priority of addressing public concerns about cancer risks near Commission-licensed nuclear facilities versus other agency priorities. As noted in this summary and discussed in detail in Chapter 4, the statistical power of epidemiologic studies of cancer risks in populations near nuclear facilities is likely to be low based on currently reported effluent releases from those facilities. Moreover, the magnitude of the variation of other risk factors that may not be measurable such as smoking or exposure to medical radiation may surpass the expected effect from the releases of the nuclear facilities and therefore overwhelm the actual effect attributed to the releases. Nevertheless, there may be sound policy reasons for proceeding with these studies: They can help to address public concerns about cancer risks and also demonstrate the USNRC’s commitment to working constructively with its stakeholders.

RECOMMENDATION 2: A pilot study should be carried out to assess the feasibility of the committee-recommended dose assessment and epidemiologic studies and to estimate the required time and resources.

Additional work beyond the scope of this Phase 1 study will be required to assess the feasibility of these recommended studies and to estimate the time and resources needed to carry them out. The recommended pilot study is designed to develop this information. The pilot study should focus on the four activities described below. Additional details can be found in Chapters 3 and 4.

• Obtain effluent release and meteorology data for six nuclear plants and one fuel-cycle facility (the committee suggests Dresden, Millstone, Oyster Creek, Haddam Neck, Big Rock Point, San Onofre, and Nuclear Fuel Services; see Chapter 2) and digitize these data into a form that is usable for dose estimation. The pilot should also develop a methodology for estimating releases of carbon-14 from the six nuclear plants for all years of operations for which effluent release data are available.
• Develop a computer model (i.e., by modifying an existing model or developing a new model) to obtain estimates of absorbed doses to individual organs resulting from airborne and waterborne effluent releases, and use this model to obtain dose estimates as a function of distance (0 to 50 kilometers from the plant) and direction for each of these seven facilities. Methodologies should also be developed to account for natural background radiation and, to the extent feasible, other sources of radiation in the dose estimates, especially medical radiation. An analysis should be carried out to estimate dose uncertainties.
• Retrieve cancer incidence and mortality data at the censustract level within 50 kilometers of these seven facilities to assess feasibility of the recommended ecologic study.
• Confer with investigators who are conducting linkages of cancer and birth registration data to identify eligible cases of pediatric cancers and matched controls to assess feasibility of the recommended record-linkage-based case-control study. Where such linkages are not already in place, link birth registration and cancer incidence data to identify eligible cases of pediatric cancers and matched controls.

RECOMMENDATION 3: The epidemiologic studies should include processes for involving and communicating with stakeholders. A plan for

stakeholder engagement should be developed prior to the initiation of data gathering and analysis for these studies.

Stakeholder engagement is an essential element of any risk assessment process that addresses important public interests (see Chapter 5). Several approaches were used in this Phase 1 study to engage with stakeholders. The Phase 2 study can build on these Phase 1 efforts to achieve effective collaboration with local people and officials and increase social trust and confidence. To this end, the Phase 2 study should develop and execute an engagement plan that includes processes to:

• Identify key stakeholders and stakeholder groups with whom engagement is essential.
• Assess stakeholder concerns, perceptions, and knowledge.
• Communicate the questions that the Phase 2 study can address and its strengths and limitations, and communicate the results from the Phase 2 study in forms that are useful to different stakeholder groups.
• Make the information used in the Phase 2 study publicly accessible to the extent possible.

It is important that the plan be developed prior to the initiation of data gathering and analysis to ensure early engagement with stakeholders in the Phase 2 study.

This page is blank