Analysis of Cancer Risks in Populations Near Nuclear Facilities: Phase 2: Pilot Planning (2014)

Appendix B

NAS Staff Progress with
Responding to the Statement of Task

The activities described below were carried out by NAS and were overseen by the expert committee.¹

APPOINT THE STUDY COMMITTEE

Members of the committee were chosen on the basis of their knowledge and expertise in the scientific disciplines needed to carry out the study. The NAS provisionally appointed the expert committee for the study on October 23, 2013. An additional expert, Dr. Christie Eheman, was provisionally appointed to the committee on January 27, 2014, to provide expertise in data availability and release criteria from cancer registries. The slate of provisional committee appointments was open to public comment for 20 calendar days and all appointments were finalized February 17, 2014.

IDENTIFY THE PROCESSES FOR SELECTING QUALIFIED INDIVIDUALS AND/OR ORGANIZATIONS TO PERFORM THE EPIDEMIOLOGY AND DOSIMETRY TASKS

The collection and analysis of dosimetry and epidemiology data will be carried out by individuals and/or organizations whose work will be overseen by NAS and a NAS advisory committee. Qualified individuals and/or organizations to perform data collection and analysis will likely be identified by issuing a request for proposal (RFP). An RFP is a type of bidding solicitation in which NAS will announce that funding is available for investigators to provide research support for the pilot study and will outline the bidding process and contract terms. The RFP will be open to a wide range of bidders and will create open competition among qualified individuals and/or organizations. The decision on who will be awarded the contract will be made by NAS.

INITIATE EFFLUENT RELEASE AND METEOROLOGICAL DATA COLLECTION

NAS staff considered multiple sources of information to obtain effluent release and meteorological data. Data collection efforts primarily focused on obtaining effluent release reports for the seven pilot facilities because it was

________________

¹ With the exception of the first activity, Appoint the Study Committee, which was carried out by NAS staff using the NAS process for committee appointments. See http://www.nationalacademies.org/studyprocess/.

expected that they would contain the data required to estimate doses to the study populations. In the case of the Dresden Nuclear Power Plant, a large portion of the population living within 50 kilometers (30 miles) from the plant is also exposed to releases from two neighboring plants: Braidwood and LaSalle. These neighboring plants will also be considered in estimating doses to that portion of the population; therefore, effluent release reports from these nuclear power plants need to be collected as well.

EFFLUENT RELEASE DATA

Required first by the Atomic Energy Commission and now by the U.S. Nuclear Regulatory Commission (USNRC) (according to 10 CFR § 50.36(a)(2)), licensees submit effluent release reports during operation of the facility and during decommissioning. The licensee’s technical specifications (required by 10 CFR § 50.36(a)(2)) also contain reporting requirements for radioactive effluents. The licensee’s final safety analysis report identifies commitments regarding the content of nuclear power facilities’ effluent release reports, which include the quantity of principal radionuclides released to unrestricted areas in gaseous and liquid form including additional information needed to estimate maximum potential doses to the public. This information includes the locations of the release points, information on batch and/or episodic releases, and meteorological data such as wind speed, direction, and stability. The effluent release reports from recent years routinely list releases for 20 to 35 radionuclides. Carbon-14, a radionuclide of particular interest today, was first consistently reported in licensees’ effluent release reports beginning in 2010. For licensees of facilities processing special nuclear material (e.g., Nuclear Fuel Services), the USNRC technical specifications (10 CFR § 70.59) define similar effluent release reporting requirements.

The effluent release reports were collected from the following three sources:

1. The USNRC’s official record-keeping system known as the Agencywide Documents Access and Management System (ADAMS).²
2. The USNRC’s offsite archival storage facility.
3. The nuclear facilities.

At the time this report was written,³ 68 percent of effluent release reports produced by the pilot facilities (including Braidwood and LaSalle) from the start of operations to 2013 were retrieved and determined to be human-readable. As shown in Table B.1, a large percentage of the reports from 1995 to 2013 (92 percent) and the majority of reports from 1975 to 1994 (74 percent) have been retrieved and determined to be human-readable. However, very few reports prior to 1975 have been collected (fewer than 25 percent). The USNRC is searching its offsite storage facility and working with its licensees to retrieve these early effluent release reports.

The effluent report retrieval efforts that took place during the pilot planning study have resulted in the public access to over 100 additional effluent release and other related reports.⁴ Approximately 70 of these are Nuclear Fuel Services reports previously restricted from public access. The remaining reports were provided by three licensees in response to a request from NAS staff (Oyster Creek) or USNRC staff (San Onofre and Dresden) for effluent release reports from early (pre-1975) years of operation.

Efforts to collect additional effluent release reports or identify better-quality copies of the reports are ongoing. However, it is possible that some gaps will remain. Effluent release data for the years when reports are missing will need to be interpolated from available reports, or data from annual summary reports⁵ may be used. In 1977, the

________________

² Information on ADAMS and its various libraries can be found at http://www.nrc.gov/reading-rm/adams.html. The web-based ADAMS Publicly Available Records System (PARS) Library was used to identify and retrieve the most recent effluent release reports (reports dating from approximately 1995 and later). Reports retrieved directly from ADAMS PARS Library are considered official reports. The web-based ADAMS Public Legacy Library was used to identify earlier effluent release reports; these are stored on microfiche. Reports retrieved from microfiche are not considered official effluent release reports.

³ That is, as of 11/12/2014.

⁴ These reports can be downloaded from the ADAMS PARS Library.

⁵ Several sources of annual effluent release summary reports exist, including the USNRC’s NUREG 2907 series. This series, which was originally produced by Brookhaven National Laboratory for the USNRC, contains yearly summaries of nuclear power facilities’ effluent releases. NAS staff has collected summaries from 1980 through 2009.

TABLE B.1 Summary Description of Effluent Release Reports Retrieved

^aAccounts for the operation of the first unit at a given site.

Environmental Protection Agency (EPA) produced a report on annual summed measurements of effluent releases from nuclear power facilities. The report includes a listing of yearly summed releases of 27 isotopes. For some radionuclides, including carbon-14, it may be possible to fill gaps in effluent release data by scaling to power production.

METEOROLOGICAL DATA

Although nuclear facilities collected meteorological data⁶ as frequently as every hour, only quarterly or semiannual joint frequency distributions are reported in the effluent release reports, and the data are not always readable primarily because of the small font size. Also, data often are not available over the entire period of interest for this study.

Other sources of information for meteorology need further exploring. One such source is the National Center for Atmospheric Research Reanalysis Project. This project has used measured data from different sources (e.g., weather stations, ships, aircraft, satellites) to forecast, at a minimum daily average, meteorological information at different locations. Information from this source is available from 1948 onward.

INVESTIGATE AVAILABILITY OF EXISTING MODELS OR NEED TO CREATE A NEW MODEL FOR DOSE ESTIMATION

A number of models exist that can be adapted and used to obtain estimates of atmospheric and aquatic dispersion of effluent releases, internal and external radiation exposure, and age-dependent absorbed doses to individual organs. However, considering that carbon-14 emissions may have been a major contributor to dose in recent years, an improved model for estimating dose from carbon-14 may need to be developed.

Off-the-shelf modified Gaussian plume dispersion models such as those recommended by the EPA or the USNRC⁷ will be sufficient for use in estimating average air concentrations of radionuclides released. However,

________________

⁶ The meteorological data included in the facility’s effluent release reports are wind speed, direction, and stability. They do not include precipitation.

⁷ For example, REXOQ2, a PC version of the USNRC program XOQDOQ (Sagendorf J.F., Goll J.T., and Sandusky W.F., 1982, XOQDOQ: Computer Program for the Meteorological Evaluation of Routine Effluent Releases at Nuclear Power Stations, NUREG/CR-4380, U.S. Nuclear Regulatory Commission), GASPAR (Eckerman K.F., Congel F.J., Roecldein A.K., and Pasciak, W.J., 1980, User’s Guide to GASPAR Code, NUREG/-0597, U.S. Nuclear Regulatory Commission); Strenge D. L.T., Bander J., and Soldat, J.K., 1987, GASPAR II—Technical Reference and User Guide, NUREG/CR-4653, PNL 5907, Pacific Northwest National Laboratory, Richland, WA), AERMOD (Cimorelli A.J., Perry S.G., Venkatram A., Weil J.C., Paine R.J., Wilson R.B., Lee R.F., Peters W.D., Brode R.W., Paumier J.O., 2004, AERMOD: Description of Model Formulation, EPA-454/R-03-004, U.S. Environmental Protection Agency, http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf).

some nuclear facilities may require more complicated models because of their geographies. Models such as the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model⁸ or the Regional Atmospheric Transport Code for Hanford Emission Tracking (RATCHET)⁹ model could be used for estimating air concentrations at these facilities.

Reviews of models of aquatic transport in rivers, lakes, estuaries, oceans, and other water bodies are available.¹⁰ The suitability of a model to a given site will depend on the availability of model input data such as direction and speed of flow of currents, both natural and plant-induced; intensity of turbulent mixing; size, geometry, and bottom topography of the water body; and characteristics of suspended and bottom sediments.

IDENTIFY STATE REQUIREMENTS FOR DATA SHARING AND TRANSFER OF HEALTH INFORMATION

Health information needed to conduct the recommended ecologic and case-control studies is collected and maintained by the state cancer registries and vital statistics offices. These two entities are typically found in the state’s department of public health and are separate in terms of administrative processes, statutes, and regulations. Cancer registries collect information related to incident cancer cases. Vital statistics offices collect information related to deaths (including cancer deaths) and births.

The seven pilot sites are located in six states as shown in Table B.2. These six states will need to approve release of health information that is relevant to the study. Additional states whose populations reside within 50 kilometers (30 miles) of the pilot nuclear facilities will also need to approve release of information on their portion of the population that lives near these facilities.

NAS staff visited several pilot states¹¹ and engaged in discussions with representatives of the state’s cancer registries and vital statistics offices to better understand data availability and release criteria. The information collected by NAS staff from its visits with the state offices is summarized below in terms of data availability and completeness, linkages, and policies and mechanisms.

DATA AVAILABILITY AND COMPLETENESS

The terms data availability and data completeness are used here as an estimate of totality of features of the datasets to be requested from the cancer registries and vital statistics offices to conduct the pilot study. Because doses to the populations and/or individuals will be based on their place of diagnosis (ecologic study of cancer incidence), death (ecologic study of cancer mortality), or birth (case-control study), availability of address at time of diagnosis, death, or birth is essential for conducting the pilot.

Address at Time of Diagnosis

Table B.3 summarizes the availability and completeness of information on address at time of cancer diagnosis from the pilot states. Note that the state representatives were not asked to query their databases to provide precise information on address availability; instead they were asked to provide rough estimates of address information completeness. In addition, the percentages given by the states on address information completeness (noted as per-

________________

⁸ Draxler R.R., Hess, G.D. 1998: An overview of the HYSPLIT_4 modeling system of trajectories, dispersion, and deposition. Aust Meteor Mag, 47, 295-308.

⁹ Ramsdell J.V., Jr., Simonen C.A., and Burk K.W., 1994. Regional Atmospheric Transport Code for Hanford Emission Tracking (RATCHET). PNWD-2224 HEDR, Battelle, Pacific Northwest National Laboratories, Richland, WA.

¹⁰ USNRC Guide 1.113, Estimating Aquatic Dispersion of Effluents from Accidental and Routine Reactor Releases for the Purpose of Implementing Appendix I from Light-Water-Cooled Reactors, Revision 1. April 1977; Till J.E., Grogan H.A. (eds), 2008. Radiological Risk Assessment and Environmental Analysis, Oxford University Press.

¹¹ Information related to the North Carolina cancer registry and vital statistics office and the New Jersey vital statistics office was collected by a phone interview. At the time of this writing (November 2014), NAS staff has not collected the relevant information from Rhode Island or New York.

TABLE B.2 States That Will Contribute Data to the Pilot Study

TABLE B.3 Availability of Address at Cancer Diagnosis Information from Pilot States

^aPart of the state or a selected population within the state is part of the Surveillance, Epidemiology, and End Results (SEER) program.

^bEntire state is part of the SEER program.

^cEntire state is part of the National Program of Cancer Registries.

cent records missing address on Table B.3) may not represent completeness of the information in the study areas (i.e., 50 kilometers around the pilot nuclear facilities). In fact, a number of state representatives commented that in rural areas such as those near nuclear facilities, address information completeness may be lower than state averages.

As shown in Table B.3, the time periods for which address at time of cancer diagnosis is available electronically is about 1995, although there is some variability across states. Addresses are missing or are incomplete for some records (i.e., described as P.O. Box or rural route number). All states geocode¹² address information to census tract at least for the more recent years. Some state cancer registries geocode address information in-house whereas others contract with geocoding specialists. Cancer registries may have used different geocoding tools to geocode their data, leading to different match rates¹³ and level of positional accuracy of the geocoded data.

Veterans Affairs Cases

The U.S. Department of Veterans Affairs (VA) changed its policy regarding the sharing of VA cancer data in 2007. This policy change results in incomplete reporting of VA hospital patients to some state cancer registries

________________

¹² Geocoding is the process of assigning geographic coordinates to an address so that it can be placed as a point on a map.

¹³ Defined as the ratio between the total number of automatically geocoded addresses and total number of addresses to be geocoded. See Zhan F. B., Brender J. D., De Lima I., Suarez L., Langlois P. H., 2006, Match rate and positional accuracy of two geocoding methods for epidemiologic research, Ann Epidemiol, 16(11):842-849.

and the inability of the states to share data with third parties requesting cancer incidence data from state cancer registries.

NAS staff requested information on the proportion of VA hospital cancer cases within the pilot states to roughly estimate the impact of the missing VA hospital patients in the pilot study. The annual percentage of estimated VA hospital cases ranged between 0.5 and 2 percent of the total number of cases reported annually within the state.

Address at Time of Death

Of the seven state vital statistics offices (California, Connecticut, Illinois, Michigan, North Carolina, New Jersey, and Tennessee) that provided information on availability of address at time of death, all currently have that information in electronic form. However, the first year for which the information is available electronically varies greatly: 2005, 1949, 2008, 2000, 2000, 2006, and 1990, respectively. Although the information is not available electronically prior to the years shown above, it is available in paper copies of the death certificate. Transfer of the information from the death certificate to an electronic database is estimated to cost about $1.00 per death certificate.¹⁴ Given that there are about 15,000-55,000 cancer deaths annually in each of the pilot states, transfer of the information is an expensive and time-consuming task. Most states reported that address at time of death is not geocoded.

Address at Time of Birth

Address at time of birth is needed for the case-control study and selection of appropriate cases and controls. This address is actually the address of residence of the mother at the time of delivery of the child and for the purposes of the study is also assumed to be the mother’s residence during pregnancy and the child’s residence until he/she is 15 years old. Typically, this address is available electronically in vital statistics office databases from 1995 onward with some exceptions. For example, in Illinois the information only exists electronically since 2010. Most states reported that address at time of birth is not geocoded.

Other Data

NAS staff inquired about availability of additional information that is relevant to the ecologic and case-control study designs. A sample of the variables of interest is listed in Table B.4. States reported that they collect this information.

There may be variability across and within states in missing or incorrect information on the death certificates. Examples include lack of information on cancer site or, if a cancer metastasizes, listing of the underlying cause of death as the metastatic site instead of the primary cancer site.

POLICIES AND MECHANISMS

The health information required for the pilot study generally is not publicly releasable because of privacy and patient protection federal and/or state regulations. Cancer registries and vital statistics offices will need to review the research proposal and protocol for the NAS pilot study before approving the release of information. Some states have multiple levels of protocol approvals; in general it would take 3-6 months for a protocol to be approved by the state IRB.

All of the data discussed previously (e.g., address at time of diagnosis or death, address at time of birth, and those listed in Table B.4) could be released to the investigators upon approval of the research protocol by the state’s IRB. States reported that there are no different restrictions for release of the information for children and adults. An exception was health information and other identifiable information related to the mother and reported

________________

¹⁴ Communication with state of Michigan cancer registry and vital statistics office director. This estimate was supported by representatives of other state vital statistics offices.

TABLE B.4 Sample Variables of Interest Collected from the States

N/A (not applicable) means that the specific variable is not needed to carry out one of the two recommended study designs.

in the Illinois birth certificates.¹⁵ NAS staff did not discuss with the Illinois vital statistics office representatives whether this barrier could be overcome.

DATA LINKAGES

To conduct the case-control study, birth records will be linked with cancer registration records to identify suitable cases and controls. None of the pilot states reported to routinely perform the type of linkages required for this study. However, all states had some experience with data linkages. Since no unique identifier such as the child’s or the mother’s social security number is typically present in both administrative databases, the linkage would be (at least partially) probabilistic using variables such as name, date of birth, gender, race of the child, and possibly address at time of birth.

The pilot states generally indicated that the linkages would happen in-house. However, some states indicated that they could release the data to the investigators to perform the linkages in their facilities, and other states were open to the idea of the investigators performing the linkages in the state’s facility. Although the specifics were not discussed, cost and time frame for release of the linked information would vary based on the approach.

OBTAIN IRB APPROVALS FOR THE STUDY, AS APPROPRIATE

Studies that involve health information of individuals require IRB approvals to ensure ethical conduct and protection of information. The IRB has the authority to approve, require modifications to secure approval, or disapprove the activities of the pilot study. During the pilot planning step, no health data were requested from the

________________

¹⁵http://www.ilga.gov/commission/jcar/admincode/077/077005000000200R.html.

cancer registries and vital statistics offices; therefore, there was no need to obtain IRB approvals. However, IRB approvals will be required for the pilot execution step when health data will be requested from the state cancer registries and vital statistics offices. All entities involved in carrying out the pilot study, that is, the state cancer registries and vital statistics offices, NAS, and NAS contractors, will be required to obtain IRB approvals from the states and from their institutions. Some states may require additional approvals, for example, from affiliated regulatory groups. State representatives estimate that it will take about 3-6 months to obtain the needed approvals to carry out the pilot study. IRB (and other approvals) will need to be renewed annually and whenever modifications are made to the previously approved protocol.

IDENTIFY KEY STAKEHOLDERS AND PROCESSES FOR COMMUNICATING WITH THEM

The committee and staff used several processes to communicate with and invite the participation of interested members of the public during the pilot planning. These included:

1. The NAS current project website (http://www8.nationalacademies.org/cp/) which lists committee biographies and meeting dates and agendas.
2. A dedicated project website (http://nas-sites.org/cancerriskstudy/) supplementing the NAS website to provide additional information of interest to the public about the study and further enable interested parties to submit information for the committee’s consideration.
3. A listserv to notify interested parties about project milestones such as appointment of the study committee and meeting dates, locations, and agendas.
4. Web conferencing for remote participation of interested members of the public unable to be present at the committee’s information-gathering sessions.
5. Public-comment sessions scheduled at the end of the committee’s information-gathering meetings in Washington, D.C., and Irvine, California.
6. A public meeting near the Oyster Creek Generating Station located in New Jersey.
7. Creation of a Frequently Asked Questions document¹⁶ that discusses several issues related to the pilot, the methods and nuclear facilities selected, processes of the NAS, and ways for interested members of the public to be kept informed about the study and provide comments.

OTHER ACTIVITIES

NAS staff performed additional activities not specified in the statement of task but deemed necessary for planning the pilot. Sources of information for these additional activities were:

• The February 18, 2014, conference call with Dr. Benjamin Zhan, professor and director, Texas Center for Geographic Information Science, and Dr. Francis Boscoe, research scientist, New York State Cancer Registry.
• Representatives of the state cancer registries and vital statistics offices.
• NAS staff research.

INVESTIGATING THE PROCESS FOR GEOCODING ADDRESS INFORMATION

The pilot study will likely require a substantial effort to geocode addresses because:

1. Most cancer registries only geocode data for the most recent years.
2. Vital statistics offices typically do not geocode data.
3. States and the different offices within a state are not required to use the same protocol for geocoding data.

________________

¹⁶ See: http://nas-sites.org/cancerriskstudy/.

Representatives of a number of pilot state cancer registries suggested that it is important for a multistate study such as this pilot to use the same protocol for geocoding address at time of diagnosis across states. (The same applies for address at time of cancer death and address where the mother lived at the time of delivery of the child.) This is because different geocoding methods may result in different match rates and level of positional accuracy of the geocoded data. Also, it is easier to compare and integrate data created using the same protocol.¹⁷ A number of commercial and open-source geocoding tools are available today. Eight tools were evaluated for their performance and compared for cost and licensing.¹⁸

The quality of the original address data is an important factor in determining the performance of a geocoding tool. Typically, about 80-90 percent of addresses from health-related datasets can be automatically geocoded using current geocoding tools.¹⁹ It would take about 2 seconds on average to automatically geocode an address using a typical computer. Addresses that cannot be geocoded automatically require manual intervention/interactive geocoding. The manual intervention/interactive geocoding process can range from checking an address for obvious errors that can easily be corrected to investigating means to find more information on the address from the record, websites, or other administrative databases.²⁰ Representatives of the Illinois cancer registry said that it may take up to an hour to manually geocode a single address.

Dr. Benjamin Zhan, Texas Center for Geographic Information Science, who briefed the committee, said that the cost of a geocoding project correlates to the number of records that need to be processed manually.

INVESTIGATING AVAILABILITY OF CENSUS DATA

Linking a geocoded cancer record (or other) address with census areas is used to determine characteristics of the populations living in an area in which an address is located, such as demographic, socioeconomic, household, and health insurance data. This information is available at the census tract level in the 2000 and 2010 census data. Subscription sources such as SimplyMap have additional information on lifestyle factors of the populations residing in a census tract, such as percentage of nonsmoking population, percentage of people who eat healthily, percentage of people who exercise regularly, and percentage of nonalcoholic population for the same years. For earlier time periods (1960, 1970, 1980, and 1990), some of these variables are available at the census tract level through commercial vendors, for example, Geolytics.²¹ However, since the U.S. Census Bureau did not fully tract the United States until 1990, pre-1990 data are available only for urban areas.

In addition to the decennial data available from the sources mentioned above, the U.S. Census Bureau performs the American Community Survey (ACS), an ongoing statistical survey that samples a small percentage of the population every year. The estimates at the census tract level are based on rolling 5-year data periods starting from 2005. Example variables in the ACS data include age, gender, race, family and relationships, income and benefits, health insurance, education, veteran status, disabilities, location of work and how people get to their office, and how much people pay for some essentials.

________________

¹⁷ Goldberg, D.W., 2008, A Geocoding Best Practices Guide. North American Association of Cancer Registries, https://www.naaccr.org/LinkClick.aspx?fileticket=ZKekM8k_IQ0%3D&tabid=239&mid=699 (accessed November 2014).

¹⁸ Swift J.N., Goldberg D.W., Wilson J.P., 2008, Geocoding Best Practices: Review of Eight Commonly Used Geocoding Systems. University of Southern California GIS Research Laboratory Technical Report No 10, http://spatial.usc.edu/wp-content/uploads/gislabtr10.pdf (accessed November 2014).

¹⁹ Wang Y., O’Leary L.A., Rickard R.S., Mason, C.A., 2009, Geocoding capacity of birth defects surveillance programs: Results from the National Birth Defects Prevention Network Geocoding Survey. J Registry Manage, 37(1):22-26; Zhan F.B., Brender J.D., De Lima I., Suarez L., Langlois P.H., 2006, Match rate and positional accuracy of two geocoding methods for epidemiologic research, Ann Epidemiol, 16(11):842-849.

²⁰ Goldberg D.W., Wilson J.P., Knoblock C.A., Ritz B., Cockburn M.G., 2008, An effective and efficient approach for manually improving geocoded data. Int J Health Geogr, 7(1):60.

²¹ See http://www.geolytics.com/USCensus,Census-1970-1980-1990,Categories.asp.

This page intentionally left blank.