Review of Health Studies Relevant to Lawrence Livermore National Laboratory and the Surrounding Community
LAWRENCE LIVERMORE NATIONAL LABORATORY, MAIN SITE (USDOE)
LIVERMORE, ALAMEDA COUNTY, CALIFORNIA
Lawrence Livermore National Laboratory (LLNL), a research facility conducting nuclear weapons research and development, is located in Livermore, California, in the San Francisco Bay Area. In 1987, the LLNL Site was placed on the federal National Priorities List ("Superfund"). Releases of volatile organic chemicals, metals, plutonium and tritium have occurred at this site.
The California Department of Health Services (CDHS) has produced this health consultation with support from the Agency for Toxic Substances and Disease Registry (ATSDR). The review of available health outcome data for plausible outcomes and outcomes of community concern is a standard component of the ATSDR health assessment process, and this review was requested by members of the site team, which consisted of community residents, stakeholders, and agencies working with CDHS at this site.
This health consultation summarizes health studies and reviews regarding LLNL and is not intended to provide a comprehensive analysis of those studies. Although in most cases the study is simply summarized, in some cases we comment directly on the analysis to provide the reader with a fuller interpretation, and these comments are noted as our own. The studies fall into several areas: (1) studies of cancer incidence in workers (many on melanoma, a type of skin cancer); (2) case-control studies of workers to identify possible causes of elevated melanoma; (3) reviews of studies; (4) studies of cancer incidence among community members; and (5) a review of birth defects in the community.
In general, the incidence of cancer among LLNL employees was not higher than expected. However, studies from the 1970s through the mid-1980s found melanoma rates among employees of the LLNL to be approximately three times higher than expected. Some of this excess might be partly due to earlier diagnosis of LLNL employees, as publicity about melanoma prompted heightened awareness. The search for workplace risk factors for melanoma found several factors to be associated, including work around ionizing radiation.
Ionizing radiation is radiation that has enough energy to strip away electrons from atoms or break some chemical bonds. Although there may not be adverse health effects, cells may become malignant and the chance of cancer increases. The main types of ionizing radiation are alpha, beta, and gamma (e.g., nuclear explosions are one source of gamma rays, plutonium emits alpha rays, and tritium emits beta rays). Tritium is a radioactive isotope of hydrogen that binds with oxygen to form tritiated water, which can be readily absorbed through the skin, inhaled, or ingested. In the Livermore community, historic tritium releases from LLNL are one possible exposure source for the community to ionizing radiation.
Overall rates of cancer in community residents were not elevated over three decades. However, statistical reviews of health data among community residents found elevations of melanoma among children and young adults. More recently, melanoma rates among residents of Livermore were not significantly elevated, nor were more birth defects found than expected.
In 1984, LLNL instituted on-site medical screening for melanoma in order to detect and treat pre-cancerous skin lesions. LLNL has stated that melanoma rates have fallen to Bay Area averages since 1985.(1) Data on melanoma rates should be periodically reviewed to see if this trend is continuing. Also, because of the findings that working around radiation was associated with higher risk of melanoma within the LLNL workforce, it would be useful for future epidemiological studies of radiation in other settings to consider investigating melanoma specifically.
This document was originally issued as a Public Comment Draft in February 2003 and has been modified in response to comments received. Original comments and responses are included in the Appendices.
Lawrence Livermore National Laboratory Site
Lawrence Livermore National Laboratory (LLNL), a high-energy research facility conducting nuclear weapons research and development, is located in Livermore, California, in the San Francisco Bay Area (Figure 1, Appendix A). The site was originally used by the military as a flight training base and aircraft assembly and repair facility, before the Atomic Energy Commission converted the former base to LLNL in 1951 (1).
LLNL was placed on the National Priorities List (NPL; "Superfund") of hazardous waste sites in 1987 because of groundwater contamination from volatile organic chemicals (1). Spills and other releases of metals, fuel hydrocarbons, plutonium and tritium also had occurred at the site (1,2). Developing and testing nuclear weapons and conducting other energy research at LLNL has involved numerous and varied chemicals, processes, and radiation sources (3).
Above-background levels of plutonium in a community park near LLNL were discovered, and an investigation was conducted to determine how the contamination might have occurred (4). It has also been determined that plutonium-contaminated sludge (due to releases from LLNL) from the municipal water treatment plant was distributed to the public and other entities for use as a soil amendment in the community. The practice of distributing sludge occurred from the late 1950s until the early 1970s (5). The levels of plutonium were estimated and evaluated by ATSDR, and deemed well below a level that would cause health problems (5), but as the data are incomplete, the possibility that the evaluation may not be definitive in all cases has been raised (6).
Most of the tritium released from LLNL has been to the air, with the Tritium Facility as the major source of these smaller, routine, day-to-day airborne releases, which occur in the course of conducting experiments with radioactive gases (7). Other sources include sewage discharge (7). Tritium has been found in rainwater at LLNL, with levels seven times the drinking water standard (8). Elevations of tritium were also observed in off-site rainwater samples, and the levels were found to be lower than the on-site samples (8). Besides the routine releases, there were several large releases of tritium to the Livermore community, one in 1965 and one in 1970 (9). An expert panel determined that 80% of the releases from LLNL occurred during those two accidents (9). These air releases created completed exposure pathways with the nearby population (a completed exposure pathway means that the contaminant reached a population that had contact with it) (9).
Ionizing radiation is radiation that has enough energy to strip away electrons from atoms or break some chemical bonds, creating ions which can be hazardous to living tissue (10,11). In some cases there may not be adverse effects (11). The body has ways to repair some cell damage, but if cells become malignant, the chance of cancer increases (11). The main types of ionizing radiation are alpha, beta, and gamma (e.g., nuclear explosions are one source of gamma rays, plutonium emits alpha rays, and tritium emits beta rays) (10). Tritium is a radioactive isotope of hydrogen that binds with oxygen to form tritiated water, which can be readily absorbed through the skin, inhaled, or ingested. (Additional information on ionizing radiation is included in Appendix D.)
Public Health Assessment at LLNL Site
CDHS, in a coop process with ATSDR, began a health assessment process in September 1996 to evaluate whether or not site-related contaminants from LLNL were affecting the local community. At the start of the process, CDHS convened a site team of community members, stakeholders, and agency representatives to help identify and prioritize health topics to be addressed, review and comment on documents prepared about the site, and act as a conduit between the site team and the agency or community that each member represents. The review of available health outcome data for plausible outcomes and outcomes of community concerns is a standard component of the ATSDR health assessment process, and a health consultation reviewing and summarizing the different health studies relevant to the Livermore community was requested by members of the site team. The CDHS has prepared this health consultation with support from the ATSDR. ATSDR is preparing a document which will summarize the evaluation of site-related contaminants.
Children's Health Considerations
CDHS and ATSDR recognize that in communities with contaminated water, soil, air, or food, infants and children can be more sensitive than adults to chemical exposures. This sensitivity results from several factors: (1) children might have higher exposures to environmental toxins because, pound for pound of body weight, children drink more water, eat more food, and breathe more air than adults; (2) children play outdoors close to the ground, increasing their exposure to toxins in dust, soil, surface water, and ambient air; (3) children have a tendency to put their hands in their mounts, thus they might ingest potentially contaminated soil particles at higher rates than adults; (4) children are shorter than adults, which means they can breathe dust, soil, and vapors close to the ground; (5) children's bodies are rapidly growing and developing, thus they can sustain permanent damage if toxic exposures occur during critical growth states; and (6) children and teenagers more readily than adults may disregard no trespassing signs and wander onto restricted property. Regarding ionizing radiation exposure specifically, because children are growing more rapidly, they have more cells dividing, posing a greater opportunity for radiation exposure to disturb this process (11).
Scope and Purpose of this Health Consultation
This health consultation (HC) summarizes publicly available information from a number of health studies and health data reviews regarding LLNL and the surrounding community. The document is not intended to provide a comprehensive analysis of those studies. The HC includes both worker and community studies of cancer because findings among workers may be relevant as an indication of potential effects in the community. Although in most cases, the studies (or critiques) are simply summarized, in a few cases it seemed relevant to comment directly on the analysis to provide the reader with a fuller interpretation of the findings, and these comments are indicated as our own.
The studies reviewed in this document fall into several categories:
- LLNL worker cancer incidence rates;
- LLNL worker case-control studies (to identify possible causes of elevated melanoma rates in the workforce);
- Reviews of worker studies;
- Community health studies of cancer incidence among residents in the area; and
- Birth defects (review of incidence in the community).
This review first summarizes LLNL worker studies. Included are both studies of cancer incidence rates and case-control investigations to determine potential workplace risk factors for melanoma. Many of the studies concern melanoma, which became the subject of a series of investigations that found an excess of this cancer in the LLNL workforce. Further investigations were undertaken in an attempt to verify the previous findings, assess whether studies had been correctly performed, understand the importance of personal risk factors (such as lifestyle or physical characteristics), evaluate alternative explanations for the elevation (such as aspects of data reporting), and identify specific workplace exposures that could account for the elevation (such as radiation or chemicals). Reviews of the key studies were conducted to evaluate them and confirm the results, which are reported here as well. All existing community studies are also reviewed; these include reports on cancer incidence and birth defects. A summary table of studies reviewed is also provided; these are listed approximately in chronological order (Table 1, Appendix B).
This document was originally issued as a Public Comment Draft in February 2003, and has been modified in response to comments received. Most of the comments fell in several main areas, so we have written a general response to each area that addresses a number of specific points. These responses are in Appendix C. Additional responses to minor comments can be found in Table 4. The original comments received are also included (Appendix F).
Community Health Concerns
Some community members were concerned about what they felt was an excess of cancers among people in the community. In addition to mentioning melanoma, other cancers of concern cited by residents included leukemia, brain cancer, and ovarian cancer. Some community members were concerned about and/or knew of children with birth defects. Other conditions mentioned included fibromyalgia, irritable bowel syndrome, psoriasis, rashes, diarrhea, stomach ache, headache, and fatigue. Community health concerns are discussed more comprehensively in a health consultation on that topic (12). Because this health consultation is a review of existing information, only cancer and birth defects are addressed.
Background Information on Cancer (please see Appendix E for additional information)
Although incidence of and mortality from many cancers have been decreasing in California, cancer overall is still the second leading cause of death (13). According to current rates, one of every two men and two of every five women will develop some form of invasive cancer in their lifetimes (13). Breast cancer is the most common cancer among women, prostate cancer is the most common cancer among men, and lung cancer is the second most common cancer for both men and women (13).
History of Melanoma Concerns at LLNL and in Livermore
Community concern about LLNL-related melanoma first came to the attention of CDHS in 1978, when a prominent LLNL scientist died of melanoma. In a 5-year period, there had been 13 LLNL employees who had been diagnosed with melanoma (14). These findings prompted the chief laboratory physician of LLNL to request that the Resource for Cancer Epidemiology (RCE) of CDHS (now the Cancer Surveillance Section of the California Cancer Registry) investigate the incidence of melanoma among LLNL employees (3,14). These events also raised questions about children in the area who had developed the disease around the same time (14).
In 1984, LLNL established a workplace melanoma screening program to ensure that employees were regularly screened for skin diseases (15). All employees were systematically asked to examine their moles (a risk factor for melanoma) and report their findings to LLNL Health Services personnel. A physician still runs a weekly clinic at LLNL to examine employees and surgically remove any suspicious lesions before they develop further. Many melanomas can be prevented if earlier skin changes are identified and treated, and treatment will be more successful for melanomas identified in early stages (15).
Malignant Melanoma of the Skin
Melanoma is one of three common types of skin cancer; the other two are basal cell carcinoma, and squamous cell carcinoma. These skin cancers usually develop in the outer layer of skin, the epidermis. Basal and squamous cell carcinoma are often grouped together as non-melanoma skin cancer. The top layer of the epidermis is made of dead cells that contain keratin. The cells that produce keratin can develop into squamous cell cancer. The deeper layers of the epidermis has cells called melanocytes, which produce melanin, the pigment made when skin is exposed to sun. These are the cells that become cancerous in malignant melanoma. At the bottom of the epidermis are the basal cells, which are the cells that normal skin develops from, but can develop into basal cell cancer, which is the most common type of skin cancer. The thickness of the epidermis and dermis (the skin layer below the epidermis) together can vary from 2-4 millimeters (mm).
Epidemiology of Malignant Melanoma of the Skin
Although malignant melanoma of the skin (cutaneous melanoma) was once considered rare, its incidence has risen dramatically (16). In 2002, 1.3 million Americans were diagnosed with skin cancer; although only around 4% of these skin cancers were melanoma (26). In the United States, the risk of developing this cancer is estimated to be about one of every nine persons (17). Recent data show 11.5 new cases per year for every 100,000 persons in California (13). Over time, incidence rates have increased among white populations around the world (18), a development that does not appear to be merely the result of greater public awareness (17).
This increase in rates has also been followed by an increase in the number of well-designed studies covering occupational and other exposures, and constitutional risk factors (i.e., physical characteristics such as skin type) (26). Information from these studies has provided evidence for the role of solar ultraviolet light exposure in development of melanoma (19), as well as evidence implicating exposure to artificial sources of ultraviolet light (20). There is also some evidence that persons with suppressed immune systems may be at greater risk for melanoma, such as persons who had received immunosuppressive treatment for a primary cancer (21), or organ transplant recipients (22).
Melanoma incidence and mortality rates vary by social class, with higher rates in the upper social classes (23,24,25). This is hypothesized to be associated with lifestyle factors, perhaps because wealthy individuals might be more likely to vacation in sunny locations (19,24). Constitutional risk factors must also be considered: higher risk is associated in persons with light or red hair, and persons who have a tendency to sunburn easily (23).
Having a large number of nevi (moles) is now recognized as one of the strongest predictors of melanoma risk (26,50). However, atypical dysplastic nevi (irregularly shaped moles) appear not only a marker for increased risk of melanoma, but a potential precursor of melanoma (27). This means that they share a common causal pathway, such as sun exposure, but nevi would develop before melanoma, as an intermediate step in the evolution of melanoma (26).
Ionizing radiation is considered among other possible occupational exposures as to its ability to cause melanoma, but the evidence has been mixed (28). Because melanoma has not been historically associated with ionizing radiation (only with ultraviolet radiation) (28,37), opportunities could have been lost in rigorously assessing this relationship. For example, a 1990 report of atomic bomb survivors reports that skin cancer except melanoma was evaluated (29). It was similarly not evaluated in a large study of British atomic weapons workers (30), nor another study of radiation workers in the United States (31).
Consideration of Biological Plausibility for Melanoma being caused by Ionizing Radiation
Ionizing radiation is known to cause many human cancers (28). Because ionizing radiation removes electrons from atoms, it has the ability to produce permanent damage after a relatively small amount of energy is absorbed (28). Radiation is thought to induce carcinogenic change by either 1) inducing mutations and altering the structure of single genes of chromosomes; 2) changing gene expression without mutations; or 3) weakening the immune system, allowing certain viruses in turn to cause cancer (32). [Please refer also to the Appendix D for information on ionizing radiation.] Although almost all cancers can be caused by radiation, the amount of radiation required to induce a particular type of cancer varies (33). Some cancers are very common in the population generally, but are not easily caused by radiation (relatively more radiation is required to induce them) (e.g., colon); others are infrequent but can be caused by lower amounts of radiation (e.g., thyroid) (33). In the opinion of CDHS, one possibility why melanoma has been overlooked in the past is that even though it could be caused by ionizing radiation, it may not have been recognized because it may not be as easily caused by radioactivity as other cancers. Another reason is that melanoma is much less frequent in comparison to the other types of skin cancers, and detecting an increase in a frequent event will be easier than in a rarer outcome, at least from a statistical perspective.
As discussed above, releases of a variety of chemicals and radionuclides have occurred at the LLNL site, including plutonium and tritium. In the event of human contact with tritium, because it is essentially a type of water molecule, it is easily absorbed by the skin (10), or can damage living tissues and organs it comes in contact with after inhalation or ingestion (34). In the opinion of CDHS, the fact that tritium can be absorbed into the skin supports the plausibility that this type of exposure could contribute to melanoma, although it is not known at this time if this exposure is related to melanoma occurrence.
Plutonium-239 is another site contaminant. Although plutonium is considered to be usually absorbed by the dead, outer layer of human skin (11), recent research on skin decontamination methods found that when plutonium was applied to skin samples, most of its radioactivity was found to be in the epidermis, with 2-4% of its radioactivity penetrating through the epidermis to reach the deeper, dermal skin layer (35). Depth of penetration would in part presumably depend on individual skin thickness. Because the melanocytes (the cells which develop into melanoma) are located within the epidermis, the fact that melanoma develops from the melanocytes and that plutonium-239 may penetrate into and beyond the skin layer containing melanocytes, would support the plausibility that plutonium exposure may have the potential to contribute to the development of melanoma, in the view of CDHS.
Incidence is a measure that is used to quantify the number of new cases of a disease that develop in a population at risk during a specific time interval (36). Studies can measure disease incidence among different groups of individuals who are defined by the presence or absence of exposure to a suspected risk factor. The incidence of disease can then be compared between these groups as a method of evaluating the relationship between disease and exposure. This comparison between two incidence rates is called an "incidence rate ratio."
Table 2 shows the elevation in melanoma in the LLNL workforce compared to other populations, from different studies. Figure 2 shows melanoma cases among LLNL workers over time. This figure, and the studies which contributed these data, are discussed later in this report.
Original Investigation of a Perceived Excess of Melanoma Among LLNL Employees (A Study of Cancer Incidence in Lawrence Livermore Laboratory Employees [Report #1]; Malignant Melanoma Among Employees of Lawrence Livermore National Laboratory (37,38))
The initial study (Report #1) of melanoma incidence among LLNL employees was conducted by the RCE and LLNL. The findings were issued as a report by CDHS in 1980 (37), and subsequently as a journal article in 1981 (38). The study examined incidence of melanoma among the 5,100 white employees of LLNL during 1972-1977 who lived in Alameda or Contra Costa counties. For each study year, members were grouped by 5-year age group, sex, and census tract of residence. Each year of residence in the study area by a study member (employee of LLNL during that year) was counted as 1 year of observed time for the group to which the member belonged. The number of observed cancer cases was compared to the number that would be expected. The expected number was based on the rates among all white persons of the same age group and sex who lived in the same census tracts as the employees during the given year (of employee diagnosis).
Nineteen cases of malignant melanoma were identified at LLNL during the study period. The number of melanomas observed was three times higher than expected. The statistical likelihood of an elevation of this magnitude occurring by chance alone is about two in 100,000 (for males).
The statistical likelihood of certain results occurring by chance can also be expressed using a confidence interval (CI), which represents the range within which the true magnitude of effect lies with a certain degree of assurance (often 95% is chosen) (36). For this analysis, the best estimate of the incidence ratio of melanoma among males at LLNL was 3 (i.e., males at LLNL were 3 times more likely to have melanoma compared to Alameda and Contra Costa county residents during 1972-1977). However, a confidence interval can add information about how much uncertainty exists in this estimatethe wider the confidence interval, the less precise the estimate. The width of the confidence interval reflects the sample sizethe greater the number of people studied, the more precise the estimate becomes. The degree of precision for the estimate of 3 is shown by its confidence interval, which indicates that with 95% assurance (from a statistical perspective), the true incidence ratio lies between 1.5 and 5.9 (i.e., the true incidence ratio is at least 50% higher than expected and might be nearly six times higher than expected.)
Another component of the analysis assessed the incidence rate for malignant melanoma among white males aged 20-64 in the LLNL study group. The rate was compared to the corresponding rate for all white males in Alameda County (excluding LLNL employees). The LLNL worker rate was 48.8 cases per 100,000 persons versus 11.7 cases per 100,000 persons for Alameda County. Thus, melanoma incidence among male LLNL employees in the study group was over 4 times higher than that of other white males in Alameda County. It also was elevated when compared to the community rate in the cities of Livermore and Pleasanton, excluding LLNL workers. This community rate was 13.5 cases per 100,000 persons, indicating male LLNL workers experienced about 3.6 times the melanoma rate of the surrounding community.
Cancer Incidence in LLNL Workers [Report #2] (Cancer Incidence Among Employees of the Lawrence Livermore National Laboratory, 1969-1980 (39))
A second cancer incidence review spanning a greater number of years (1969-1980) was undertaken by RCE to assess whether there was an excess of melanoma or any other cancer in this occupational cohort. The study examined the incidence of cancer among active employees of LLNL between 20-69 years of age. Person-years of observation were categorized by sex and 5-year age groups. The age categories contributing the greatest number of person years were in the middle of this age range, with the 40-44 year old category highest among these. The San Francisco-Oakland Standard Metropolitan Statistical Area (SMSA) rates were used for comparison. For all types of cancer combined, the ratio of the observed number of cancer cases to the number expected was 0.95, indicating that fewer cases of cancer occurred than would typically be expected for this population. Among the approximately 100 comparisons made, the only statistically significant excesses found were salivary gland tumors (among females), nervous system tumors other than brain tumors (among males), and malignant melanoma of the skin (males and females). The observed number of melanomas was over 3 times that expected. Although the excess of salivary gland tumors was statistically significant among women, this was based on only a few cases. For other nervous system tumors (other than brain cancer), three cases occurred, but less than one was expected (0.23). When a large number of analyses are made simultaneously, it is possible that some will be elevated by chance alone. Thus, one possibility is that some of these findings could have been due to chance.
Cancers were also analyzed according to their "radiosensitivity;" that is, how likely they are to be caused by radiation exposure. Cancers were further grouped according to whether they were considered "highly radiosensitive" or "moderately radiosensitive." No excess cancer was found among the highly radiosensitive category (the ratio of the observed to expected was 0.86). The "moderately radiosensitive" cancers include lung and female breast, and the ratio of observed to expected cancers was significantly less than expected (0.62).
Men at LLNL had less lung cancer than the general population (the difference was statistically significant). More educated men tend to smoke less than the general population, and, as the LLNL workforce may be generally well educated, this could be a reason for this difference. Thus, in general, the study did not find cancer rates, other than melanoma, to be elevated in active LLNL employees.
Investigation of Melanoma Elevation and Alternative Explanations Using Kaiser Permanente Medical Care Program (KPMCP) Population (The Possible Effect of Increased Surveillance on the Incidence of Malignant Melanoma (40))
The question of whether melanoma rates were elevated among LLNL employees was also examined among members of Kaiser Permanente Medical Care Program (KPMCP), a large, prepaid health plan with eight facilities in the San Francisco Bay Area. A substantial portion of the LLNL worker population subscribed to KPMCP. Melanoma incidence among LLNL-employee KPMCP members was compared with rates among other KPMCP members. The years studied were 1976-1980.
Melanoma incidence rates for LLNL employees were uniformly higher than rates among members at any of the other KPMCP facilities, including others in sunny areas such as Walnut Creek, Vallejo, Santa Clara and Sacramento. The rates by facility included LLNL employee cases, so the true difference would be expected to be higher. Compared to the KPMCP members at one of the facilities (Walnut Creek), the relative risk of melanoma for LLNL employees was 3.2 (95% confidence interval (CI) = 1.7-6.0). This means that for an LLNL employee the risk of having melanoma was 3.2 times higher than the risk for melanoma among all KPMCP members attending the Walnut Creek facility.
In addition to evaluating LLNL employees, the study calculated melanoma rates among LLNL dependents who were KPMCP members. The relative risk of melanoma for LLNL dependents was non-significantly elevated at 2.1 (95% CI = 0.9-4.7). The number of cases was small, and from a statistical perspective, this result could have occurred by chance.
The number of physician visits for skin appointments was also analyzed to assess if there was a difference in care between LLNL employees and other KPMCP members. This analysis for the years 1974-1980 found LLNL employees had a greater number of physician visits for skin appointments to KPMCP than non-LLNL employees, and that the number of biopsies for skin lesions among persons without melanoma was greater among LLNL employees than KPMCP members. In 1976 and earlier, (before publicity about the melanoma excess), the number of biopsies was somewhat greater among LLNL employees, but not significantly so. After 1976, the number of biopsies was significantly greater among LLNL employees. The authors suggested that awareness of LLNL's excess melanoma incidence could have increased the propensity of physicians to take biopsies among LLNL employees, which would contribute to the excess incidence of melanoma. Alternatively, although this was not suggested by the authors, it would be possible that higher rates of skin conditions among LLNL employees led to a need for more frequent biopsies, e.g., the greater presence of pre-cancerous skin cell changes. Finally, the authors also stated that there could be an environmental agent (or agents) in the laboratory environment that could be responsible for an underlying increase in melanoma incidence.
Investigation of Alternative Explanations for Melanoma Elevation (Melanoma Surveillance and Earlier Diagnosis (41))
In a letter to Lancet, LLNL, with Drs. Jeffrey Schneider and Richard Sagebiel of the University of California at San Francisco's (UCSF) dermatology and pathology department, published data on the distribution of melanoma thickness among employees at LLNL over three time periods: 1969-1975 ("pre-awareness"), 1977-1984 ("early awareness"), and 1984-1986 ("aggressive surveillance"). They showed that the rate of melanoma increased by time period, and that lesion thickness decreased over these time periods as well. They suggested that the increasing proportion of thin melanomas is a result of greater awareness within the population as well as the results of medical efforts to identify skin disease earlier.
Investigation of Alternative Explanations for Melanoma Elevation Using a Community Pathology Laboratory for Comparison (Early Diagnosis of Cutaneous Malignant Melanoma at Lawrence Livermore National Laboratory (42))
In an effort to understand possible reasons for the increase in melanoma incidence among employees, LLNL conducted a study in collaboration with Dr. Schneider of UCSF. This study compared the thickness of melanoma lesions in LLNL employees with other melanoma cases in the community. The study used cases from 1976-1984. The study years were primarily after the original period for RCE Report #1 (1972-1977) and after publicity had raised public awareness. Comparisons with lesions seen at a community laboratory were made for 31 lesions from LLNL employees originally diagnosed with melanoma, which were confirmed to have cutaneous malignant melanoma and had information available on lesion thickness.
The average thickness of lesions diagnosed at LLNL was found to be thinner than those diagnosed in the community (there were 476 community member lesions for which thickness information was available). Initially during the 3-year period around 1976 there was little difference in tumor thickness between LLNL and community lesions. However, subsequently lesion thickness declined in both the community and LLNL lesions, although the LLNL decline was greater. The authors suggest that it is possible that earlier diagnosis among LLNL employees would result in thinner lesions being reported and that this might have contributed to the excess of melanoma seen among LLNL employees, at least in these years following increased publicity and awareness.
Investigation of Alternative Explanations for Melanoma Elevation (Cutaneous Melanoma at Lawrence Livermore National Laboratory: Comparison with Rates in Two San Francisco Bay Area Counties (43))
In collaboration with the Northern California Cancer Center and researchers at Stanford University, LLNL undertook another study to investigate whether the increased melanoma rate was due to either (1) under-reporting of community melanoma, or (2) heightened awareness of melanoma leading to early detection of LLNL melanomas. LLNL employees from 1974-1985 were studied. Early detection could lead to a greater number of cases that would otherwise not have been discovered until a later stage. (The total number of cases would not change, but more of them would be discovered sooner and fewer discovered later.)
The study also calculated incidence ratios for melanoma, finding rates to be 2.9 times higher than expected for men at LLNL and 2.4 times higher for women. When the tumors were separated into different thickness categories, the excess was found among the thinner categories.
To address the first hypothesis of under-reporting, investigators identified several dermatopathology laboratories that the state cancer registry had not included. Although this resulted in finding some additional community melanoma cases that had not been reported in the registry, the number of cases was small and did not explain the apparent excess incidence.
To address the second hypothesis of possible early detection, the analysis separated tumors by thickness to look at rates within thickness categories for LLNL employees. The analysis was conducted only among cases for which thickness information was available. Among LLNL employees, excess melanomas were found in the categories of very thin tumors (<0.75 mm) and medium thin tumors (0.76-1.50 mm), but not in moderate to thick tumors (>1.50 mm). The authors conclude that "...no excess of melanoma is observed if we limit our attention to thicker, more invasive melanomas." The authors suggested as a possible interpretation that melanomas among LLNL employees were detected earlier (when they are still thin), than those among community members, due to the educational campaign conducted by LLNL.
To evaluate whether this hypothesis is supported by the data, it is important to look at what happens to rates in the next time period as well. If in reality no more LLNL tumors are occurring than usual, and the apparent excess is only due to identifying tumors earlier at thinner stages, then we would expect that they would be missing from the later time period during which they would otherwise have been counted. This would mean that rates in the next time period would drop, because those early-detected tumors would be absent, as they would have already been counted in the previous time period (when they were thinner). However, although rates in this later time period decreased, they remained elevated, which would not support the hypothesis that early detection was the reason for the increased melanomas.
However, there is one other possible alternative explanation that could reconcile the early detection theory with the finding that rates did not drop in the subsequent time period. This would only be possible if some thin but invasive tumors actually disappear. This hypothesis is based on the idea that some melanomas do not develop into more invasive tumors but instead regress. The authors suggest this as a possible explanation. Although this phenomenon exists, it is very rare (44), which does not suggest it would account for the elevation.
In support of the idea that melanomas are detected earlier among LLNL employees than in the general population, the authors pointed to two differences they view as noteworthy. First, they described the proportion of very thin melanomas as increasing with calendar period, both at LLNL and in the community, but more strongly at LLNL (early detection affecting LLNL and community equally could not account for any difference in rates). Second, they concluded that lesions were thinner at LLNL compared to the community.
CDHS conducted statistical tests of their reported numbers to assess whether there were differences in lesion thickness between the two time periods, or between LLNL and the community, finding that any differences could not be distinguished from chance in this small group of cases.(2) In our view, the findings support the existence of an excess of melanoma in the LLNL workforce, but are also consistent with a possible effect of early detection. Given the publicity concerning the cases of melanoma and the LLNL screening program (in later years), it would not be unusual for lesions to have been thin, and it would perhaps have been surprising if many LLNL cases continued undetected long enough to progress to thicker, more advanced stages.
Investigation of Alternative Explanations for Melanoma Elevation Using KPMCP Population (Surveillance Bias and the Excess Risk of Malignant Melanoma Among Employees of the Lawrence Livermore National Laboratory (45))
A study was conducted among members of KPMCP to determine if early detection of melanomas among LLNL employees (surveillance bias) was the reason for elevated rates. KPMCP's Division of Research, in collaboration with others, conducted the study. They hypothesized that if surveillance were increased among LLNL members, then biopsies would be performed at an earlier stage of tumor development and the thickness of the tumors would be less than that of other KPMCP members. The study included 20 LLNL melanoma cases and 36 matched melanoma cases of KPMCP members and covered the years 1970-1984. The analysis compared tumor thickness between the early period before publicity (1970-76) to a later time period (1977-1984). The analysis found that tumors of LLNL employees were thinner than the comparison group until 1977, after which there was no difference, as the tumor thickness among the non-LLNL employees had decreased in the second time period of the study.
The authors cited a number of possible reasons for the difference in findings between this study and those of the previous analysis by LLNL with Dr. Schneider (Dr. Schneider's study found LLNL melanomas to be thinner than his comparison group during his study period [1976-1984]). One possibility is that the two studies included somewhat different groups of participants. Also, some very thin lesions were excluded in the KPMCP's study, as they were not considered melanomas by the blinded, three-person panel of expert dermatopathologists who reviewed the slides. Another possible explanation is potential differences in the nature of the comparison population. The LLNL-Schneider study used cases referred to a laboratory, and it is not known if the laboratory population was representative of cases overall.
The authors concluded their findings were consistent with an effect of surveillance bias, contributing to the excess melanoma rates up to around 1976, but did not support a role for surveillance bias in explaining the excess incidence after that time.
LLNL Announces Completion of New Cancer Review (Health Review Reported in Press: "Lab Workers During 1974-1997 Have High Rate of Cancer" (46,47))
Most recently, a review of cancer rates among LLNL employees compared to persons in the five-county Bay Area was completed on behalf of LLNL by an outside consultant. The review covered the years 1974-1997, and was limited to diagnoses that occurred during the time of LLNL employment. This summary is based on information contained in news articles published in November 2001 that contained some report findings, but stated that the report was not yet available. CDHS requested this report, however, as of this time of preparation of this report, per LLNL, the study is pending publication in a journal article, and so is not yet publicly available. The study reportedly found that there were overall fewer cases of cancer than expected among LLNL employees, for both men and women. The press release also stated that the analysis found that men who worked at LLNL during the years covered had a 38% higher rate of melanoma, although the frequency of melanoma has declined in the past 15 years, reportedly falling to Bay Area averages since 1985. Another finding presented was that testicular cancer was 107% higher than expected among LLNL employees.
The study also reportedly found a statistically lower rate of "invasive genital organ cancer in women." "Genital organ cancer" would include cervical cancer. As cervical cancer is primarily caused by a sexually transmitted virus, the main risk factors are social/behavioral (48). A low rate of cervical cancer would be expected for a well-educated population with good access to health care, such as LLNL.
A summary of melanoma incidence rate ratios among LLNL workers from different studies is shown in Table 2. A value of 1.0 means the rate was not elevated in comparison to what would be expected based on a standard or comparison population. Values above 1.0 indicate elevated melanoma rates (more melanoma occurred than expected). Elevations in melanoma rates exist for both men and women, within different time periods, and using different comparison groups. The consistency of this findings lends support to its validity.
Similarly, this elevation is portrayed in Figure 2, which shows the number of melanomas among LLNL employees over time, separating those that were in situ (non-invasive) and those that were invasive. The figure is a reproduction of one published in the LLNL journal article of their case-control study (3). The figure displays lines representing the number of expected in situ and invasive melanomas. The actual number that occurred exceeds the expected for both invasive and in situ melanoma. This shows an excess of melanoma that was not confined to thin tumors, suggesting that the excess was not simply due to greater detection among LLNL employees.
A case-control study is a type of analytic epidemiological study in which participants are selected on the basis of whether or not they have the disease that is being studied (36). The groups are then compared to assess the proportion in each group with a history of a certain exposure or a particular risk characteristic.
Initial Investigation of Potential Risk from Occupational Variables (Report #1) (Malignant Melanoma Among Employees of Lawrence Livermore National Laboratory (37,38))
The initial investigation by RCE (Report #1) of melanoma among LLNL employees also included a small case-control study of the 19 cases of melanoma that were identified among employees. The study sought to identify workplace risk factors associated with melanoma. Random selection from all other LLNL employees of the same sex, race, 5-year age group, and area of residence was used to identify four controls for each case. Cumulative gamma radiation exposure above background was compared between cases and controls; other forms and specific sources of radiation (e.g., neutron, tritium, alpha, beta) were not tested. No relationship was found between melanoma and monitored gamma radiation exposure during LLNL employment, job classification as a scientist, or length of employment at LLNL. However, an excess risk for melanoma was observed among chemists (relative risk=6.97; p-value=0.011).
Follow-Up Investigation of Melanoma Risk Factors (Report #3) (Investigation of an Excess of Melanoma Among Employees of the Lawrence Livermore National Laboratory (49,50))
A broader and more rigorous case-control study using interviews was subsequently conducted by RCE to gather more detailed information on familial cancer health history, socioeconomic status, sun exposure, sun sensitivity, and occupational exposure history. The findings were issued first in a report in 1984 ("Report #3") and later published as a journal article in 1997. This study involved all known cases of malignant melanoma identified among active LLNL employees between 1969 and 1980. Random selection was used to match the 31 cases to 110 controls from the appropriate age group (+ 5 years), race (white), and sex as indicated in the annual employee file for the year of diagnosis of the respective case. Data sources included the LLNL personnel and medical clinic records, excluding dosimetry records; a mailed questionnaire; and an extensive in-person interview with the respondent (or next of kin if the subject was deceased). Occupational data came from two sources: the personnel file classification of the position type and project assignment, and the respondent's description of his or her duties and activities in the work history portion of the interviews. Participants completed an occupational history, and a detailed description of the jobs held during the 10-year period before the date of the melanoma diagnosis was obtained. These jobs and activities were then coded according to the National Institute for Occupational Safety and Health 1980 Census classification. The coder was blinded to the case versus control status of each participant's data. Participants were also asked about their participation in nuclear events at specific locations, and work assignments at Site 300, a weapons testing area about 15 miles from LLNL.
Several occupational factors were found to be strongly associated with risk status, resulting in an elevated odds ratios (OR). The odds ratio is interpreted as: the likelihood of having a certain risk (e.g., having chemist duties) among persons with melanoma compared to the likelihood of that risk among persons without melanoma. Thus, the odds ratio for chemist duties was 8.0 (95% confidence interval: 1.4-46.0). This is interpreted as: persons with melanoma were 8 times more likely to have had chemist duties than those without melanoma. Other ORs were as follows: exposure to radioactive materials (OR=3.7; 95% confidence interval: 1.6-8.6), and exposure to high explosives (OR=3.0; 95% confidence interval 1.0-9.5). In a multivariable model adjusting for occupational risk factors, the odds ratio of working around radioactive materials (ever vs. never exposed) was 2.8 (95% confidence interval 1.1-7.1). After adjustment for constitutional risk factors (i.e., physical traits such as skin fairness, tendency to sunburn) and occupational risk factors of interest, the odds ratio associated with reported work around ionizing radiation remained elevated, although this analysis did not quite reach statistical significance at the 0.05 level of significance (OR=2.3; 95% confidence interval: 1.0-7.6). This means that the odds of a person with melanoma having worked around ionizing radiation was 2.3 times higher than that of a person without melanoma after accounting for other risk factors. The authors concluded that the results were sufficient to warrant additional studies of occupational factors and risk for malignant melanoma of the skin.
Workplace Investigation of Increased Diagnosis of Malignant Melanoma Among Employees of Lawrence Livermore National Laboratory (3,51)
In addition to the case-control studies described above, LLNL conducted its own case-control study of melanoma. The findings were first issued in a LLNL document in 1994, and subsequently published as a journal article in 1997. Although we present the conclusions that the LLNL authors made, this report also discusses some key limitations in the study methodology we have identified.
The study included 69 cases among LLNL employees in 1969-1989, and collected extensive information on a variety of known and hypothesized constitutional and occupational risk factors for melanoma. Participation was limited to those living, so deceased participants in the earlier RCE study were not included. In all, the cases in this study represent only a subset of all cases among LLNL employees. There were other differences between the studies, some of which are discussed in the article in an assessment of what could have accounted for the difference in findings.
The LLNL study did not find occupational factors (exposure to ionizing radiation and chemicals) related to melanoma risk among employees. LLNL found that constitutional factors explained most of the excess: cases tended to sunburn more than controls, cases had more moles than controls, and controls sunbathed more frequently than persons with melanoma. Although it would seem contradictory for controls (persons who did not get melanoma) to sunbathe more frequently than the persons who got melanoma, the authors suggested that cases avoided sunbathing probably due to their hypersensitivity to ultraviolet radiation. This measures approximately the same characteristic as the first risk factor: cases tended to sunburn more than controls. However, the second finding, that cases had more moles than controls, does not necessarily mean that the excess melanoma was due to the moles; it could reflect an exposure that was causing both the moles and the melanoma (52). LLNL's findings regarding constitutional factors were consistent with those identified in the RCE Report #3 issued in 1984.
A LLNL nurse conducted a standardized interview regarding constitutional factors, and a dermatologist conducted a physical examination. Participants underwent an "Occupational Factors Interview," which consisted of an interview conducted by a former leader of the Hazards Control Department who had been associated with LLNL and Lawrence Berkeley Laboratory for 40 years. The study used an open-ended interview format to gather occupational information (rather than standardized questions), with a shorthand reporter transcribing the conversation. The text was then reviewed and edited by the interviewee and the investigators, and occupational factors were assigned numeric scores. The analysis also included a computerized word-usage analysis to count the number of times in an interview a participant used various exposure or melanoma-related words. Dosimetry records from the Hazards Control Department were also obtained for gamma, neutron, tritium, skin, and hand radiation, as well as for doses before employment at LLNL. The report presented extensive information on many variables; a detailed description of all the material presented is beyond the scope of this review. The analysis did not present estimates of risk ratios for any of the risk factors, which would have provided a quantitative assessment of the degree of risk associated with a specific factor, such as having fair skin, or exposure to radiation. Several other key factors limit the interpretation of the results.
First, in a case-control study, the selection of controls is fundamental in determining the overall results of the study. LLNL used an algorithm to select controls on the basis of the best total score of five criteria: (1) sex, (2) age, (3) start date at LLNL, (4) years of education, and (5) tenure at LLNL. Race was not a matching factor, despite large disparities in risk by race (13). The factors were weighted differentially.
This algorithm was created after modification based on the experience using the first 11 cases as a pilot study. The 11 cases were re-analyzed along with the complete group using the modified criteria for the actual study (51). Different algorithms would have resulted in different controls. There were more than 16,000 possible controls from which to choose, and only one control was selected per case. Which controls are selected is important as the overall results depend on the characteristics of controls. The final algorithm emphasized some factors more heavily than others(3). For example, sex was mismatched in 11 pairs, but even the pair with the poorest overall match score had a start date and tenure within a month of each other. The results may have been quite different using a less restrictive control selection process.
Secondly, it seems that within the LLNL workforce, matching on start date would have obscured critical exposure information. This is because persons hired earlier had higher exposures to ionizing radiation and chemicals than employees hired later, as the LLNL report states (51):
"Over the years, the LLNL workforce has seen a decrease in exposure to both chemicals and ionizing radiation. Thus, employees with earlier start dates were likely to have higher exposures than those with later start dates. We can demonstrate that this was true for all of the 138 members in our case control study."
Exposure assessment in occupational studies is typically a combination of job type, or job type and concentration of substance to which a person is exposed, and duration of exposure. Matching on both start date and employment tenure will increase the likelihood that major components of exposure assessment will be the same (selecting controls with similar exposures to cases). Also, there would be no a priori reason why cases would be more likely to have earlier start dates than controls, and understanding those differences should be an important part of the data analysis.
Thus, in effect, the LLNL study appears to have matched cases to controls using a surrogate for exposure (start date). Matching should not be used for any variable that one might wish to explore (36). If a matching algorithm results in controls with similar exposures to cases, then one cannot find exposure differences between case and control groups.
When publishing their report as a journal article, the investigators attempted to address the concern about overmatching (e.g., matching on start date) by statistically assessing the relationship between start year, years of education, and occupational exposure variables. The variables were based on those in the RCE study: work with ionizing radiation, work at Site 300, work with photographic chemicals, presence at Pacific Test Site, and duties as a chemist. The aim of the analysis was to investigate how much of the variation in these occupational variables of interest was explained by the factors start year and education.
Highly statistically significant relationships were found between start year and the following four occupational factors: ionizing radiation, work at Site 300, presence at Pacific Test Site, and photographic chemicals. The relationship with chemist duties approached the standard level for statistical significance. However, the authors concluded overmatching was not a problem because less than 30% of the total variance of any of the occupational factor scores was explained by the factors start year and education. Specifically, start year and education accounted for 27% of the variance of the factor "Presence at the Pacific Test Site," and for 23% of the variance of the factor "Exposure to ionizing radiation."
The probability of a relationship this strong occurring by chance is less than 1 in 10,000 (i.e., for the relationship between start year and the factors ionizing radiation, Site 300, and Pacific Test Site, p <0.0001). In our interpretation, the great unlikelihood of this strong an association occurring by chance indicates that start year is in fact very strongly associated with the exposure variables (e.g., ionizing radiation). (This interpretation is also supported by the above quotation regarding the decrease in ionizing radiation and chemical exposures over time at LLNL.) Although it is possible that in an idealized dataset, there would be adequate variability within a given start year for persons to have varying levels of exposure, in reality at this site, start year appears to be so strongly correlated with radiation exposure and that start year very closely represents this exposure.
In our view, matching on start date may have resulted in a group of controls with very similar exposures as cases, which would have obscured a potential relationship between melanoma and ionizing radiation (and any other occupational factors that correlated strongly with start year).
However, the study authors describe the use of matching on start date as a technique to prevent the effects of confounding (a problem in epidemiologic studies that can obscure the relationship between the risk factor and the outcome). Matching is often performed for potential confounding variables with known associations with the outcome, like age, sex, and race. Because it is impossible to investigate how matched factors might contribute to causing the disease, statistical adjustment during analysis (rather than matching) is the preferred method of controlling confounding factors (36,53).
Confounding occurs when there is a factor other than the exposure being investigated that:
- is associated with the exposure or factor being investigated; and
- independently affects the risk of developing the disease (36).
A hypothetical example can be used to explain confounding. Suppose you are investigating if coffee drinking causes lung cancer. You might find that coffee drinkers have more lung cancer. However, this could be because coffee drinkers tend to smoke a lot more than other people, and it really is their smoking that is causing lung cancer. In this example:
- smoking (the confounder) is associated with the factor being investigated (coffee drinking); and
- smoking independently affects the outcome being investigated (lung cancer). That is, smoking itself causes lung cancer.
Thus, in this situation, smoking is a confounder, and to understand the effects of coffee drinking on lung cancer it is necessary to take into account smoking behavior. Even though smoking is a confounder in this example, it would still be preferable to control for it in the analysis rather than match on it. That way, it is possible to learn how smoking contributes to cancer, how coffee drinking may affect cancer, and exactly how these factors may be related; a matched design precludes the possibility of gaining this information.
Applied to the LLNL situation, the two conditions that must be met if start date is a true confounder are:
- start date (the possible confounder) is associated with the exposure being investigated (occupational factors such as ionizing radiation); and
- start date independently affects the development of the disease (melanoma).
The first condition is met: start date is associated with occupational factors (e.g., ionizing radiation) (occupational factor scores decrease with start date). What is not clear is whether the second condition is met: whether start date itself independently affects the development of melanoma. The article states: "...we matched for start date in order to minimize the effects of a general change in the background environment, which may contribute to the well-documented trend of increasing melanoma rates in recent decades." It seems questionable whether start date is associated with whatever changes exist in the general background environment that are responsible for the increased rate of melanoma over time. It is of concern that start date seems likely to control for a general change in the LLNL environment, as this would result in controlling for, not investigating, the exposures of interest. Age would be the more appropriate control for general changes over time, and it is already a matching factor. It remains unclear how start date would affect the risk of developing melanoma independently of its LLNL-related exposures.
Given the analysis problems described above, we do not view the findings of this study as providing evidence either to support or discount the potential relationship between melanoma and occupational exposures in the LLNL workforce.
Others conducted reviews of the studies by RCE in an effort to confirm results and understand if some systematic bias or error had influenced those results. These are summarized below.
Independent Review by the State of California's Legislative Audit Committee of RCE Report #1 (37,38) (Joint Legislative Audit Committee Report: "The Cancer Incidence Among Workers at the Lawrence Livermore Laboratory": A Synthesis of Expert Reviews of the Study (54))
This review was conducted in 1980 in response to the original study by RCE (Report #1), which found increased melanoma among LLNL employees (37). A panel of expert scientists and physicians was asked to review the study and provide comments and recommendations to the California Legislature's Joint Legislative Audit Committee. This broad-based group of experts concluded that the findings of an unusually high melanoma rate appeared correct. The group opined that no conclusions could be drawn about possible links to the LLNL environment, but that the findings were extremely significant and required further study and attempt at explanation.
Review by Committee Convened by U.S. Department of Energy (Report of the Ad Hoc Advisory Board on Melanoma (55))
In 1980, also in response to the RCE Report #1 (37), the Department of Energy appointed an ad hoc committee of experts to review the study and its findings. The Board agreed with the study findings that the rate of malignant melanoma among employees of LLNL was 3 to 4 times higher than the general population. It maintained that preliminary efforts to explain the excess had not yet succeeded, and recommended that the Department of Energy should support further epidemiological investigations by CDHS.
Synthesis of Additional Reviews (Malignant Melanoma at a Scientific Laboratory: A Synthesis of Reviewer's [sic] Comments on the Austin and Reynolds' Study of Employees at the Lawrence Livermore National Laboratory (56))
This document produced in 1985 was commissioned by LLNL to synthesize the comments of seven experts that LLNL had commissioned to review RCE Reports 1, 2, and 3 (37,38,39,49,50). Because the authors of this document sought to have the experts' comments anonymous, this report reflects the opinions of the authors and uses selected quotes from the reviewers.
The commissioned document addressed two questions: (1) if there was a real excess of melanoma among employees; and (2) if the observed excess is causally related to factors in the occupational environment of LLNL. The authors of the synthesis report stated that, in reference to the two questions, "The majority, but by no means unanimous, opinion of the seven reviewers is that two problems remain unresolved...." The authors reiterated the concept that possible spontaneous regression of melanomas was responsible (i.e., early stage cancers disappearing rather than progressing). Another possible explanation, early harvesting (identifying melanomas at earlier stages that would otherwise have been found later when they were thicker), was dismissed as unlikely because a decrease in rates was not subsequently found. Most reviewers also remained unconvinced about the likelihood of occupational factors being a cause, citing the lack of a biological plausibility being previously established and the lack of a dose-response gradient.
However, in our view, the specific establishment of biological plausibility is outside the scope of an epidemiological review, but cannot be ruled out (see earlier discussion of biological plausibility in the background section). Similarly, the lack of a dose-response relationship, given the data limitations (e.g., badges measure only specific kinds of radiation; unrecorded exposures), does not rule out a relationship either. The case-control study (Report #3) was the first systematic examination of the relationship between radiation and melanoma, and the validity of the relationship withstood extensive testing in several re-analyses (summarized below).
Independent Review of RCE Report #3 (49,50) (A Case-Control Study of Malignant Melanoma Among Lawrence Livermore National Laboratory Employees: a Critical Review (57))
In 1987, at the request of LLNL, the University of North Carolina (UNC) produced this independent review of RCE Report #3 (issued in 1984). The researchers systematically validated the study (49,50) by performing several steps, including checking the accuracy of the original data; validating the control selection; confirming the statistical calculations; and examining the occupational factors using additional statistical models. As part of the effort to validate the study, the reviewers also used data on individual radiation exposure as measured by badges worn by LLNL employees (this information was not available to RCE).
Using data from the radiation badges, they found no association between exposure to ionizing radiation and melanoma. They also used radiation badge information to classify job codes into those that were radiation-related and those that were not. Using this criterion, they found that the melanoma cases were significantly more likely than controls to have worked at jobs that were deemed radiation-related.
The study authors suggest several possible reasons for the lack of agreement between the findings based on job codes vs. badge measurements. The accuracy of the badge data was investigated by safety personnel at LLNL for one worker, finding that the radiation received by that worker could have been considerably greater than what was recorded on the badge. It is also possible that badges are accurate, but that they were not worn during the time of exposure. For example, exposures during off-site nuclear testing would not necessarily be recorded or included in the LLNL file. Badge data may not have included all relevant exposures, as only data for the previous 10 years was available. Also, there were known problems in the computerized retrieval system for dosimetry, which was relatively new at the time.
The analysis found that after rigorous analytical validation and testing, three occupational factors remained associated with malignant melanoma at LLNL:
- working around radioactive materials;
- presence at Site 300; and
- working around volatile photographic chemicals.
The reviewers concluded the study was generally well conceived, well conducted, and well analyzed, although they considered the causal nature of the factors to be "overstated" in the RCE report. They suggested this because the factors had been revealed in a hypothesis-generating study, and they would have liked substantially more evidence from other studies before concluding that the factors were "true risk factors." The reviewers then recommended further investigation regarding the potential relationship between these exposures and risk for malignant melanoma.
Independent Review of RCE Report #3 (49,50) (Exposure to Ionizing Radiation and Risk of Cutaneous Malignant Melanoma: Search for Error and Bias (58))
In an extension of the previous UNC review (57), the independent experts investigated the inconsistency between reported radiation exposure and individual dosimetry readings, using dosimetry readings to create an exposure index. The findings were published in a journal article in 1994. The analysis used a number of strategies to rigorously test the previous results.
The RCE Report #3 results were consistently confirmed. Melanoma case status was significantly associated with risk of exposure based on an occupation-specific radiation index, as represented by an odds ratio of 10.8 (95% confidence interval: 1.4-85.1). This association was confirmed in numerous multivariate models; a multivariate model is a statistical method for taking into account several risk factors at the same time. This is interpreted as: persons with melanoma were more than 10 times more likely to have had radiation exposure (as measured by the occupational index) than persons without melanoma, after accounting for other potential confounders. The robust nature of the results suggested to the authors that the odds ratio for reported employment in proximity to radiation may be valid.
This section reviews cancer incidence studies among Livermore community residents. Cancer incidence rate ratios from these health studies are shown in Table 3 in Appendix B. This table will be discussed later in this report.
Initially, concern about exposure focused exclusively on identification of factors in the workplace, lifestyle, or constitutional factors that might have resulted in increased melanoma rates among LLNL employees. In the 1980s, research in the United Kingdom at a nuclear establishment found a higher than expected number of cases of leukemia among children of male workers who were exposed to radiation during a pre-conception period (59). In an attempt to see if a similar effect could be seen at another nuclear facility, researchers undertook a review of cancer rates among children and young adults in the vicinity of LLNL.
Investigation of Cancer Rates in Children and Young Adults in Livermore (Cancer Incidence Among Children and Young Adults in Livermore, California: 1960-1991 (60))
This study was focused on leukemia and non-Hodgkin's lymphoma, the cancers found to be elevated in the United Kingdom study. Cancer incidence was examined in two cohorts: (1) white children and young adults (to age 24) residing in Livermore, California, and diagnosed between 1960-1991; and (2) white children and young adults (to age 24) born in Livermore, California, between 1960-1990, and diagnosed between 1960-1991.
The overall rate for all types of cancer combined was 1.2, with a 95% confidence interval of 1.0-1.4, indicating that the rate was not statistically different from 1.0 (indicating no elevation). Rates for the other cancer groups evaluated were not found to be in excess: leukemia 1.2 (95% CI=0.7-1.7); non-Hodgkin's lymphoma 0.7 (0.2-1.7); Hodgkin's disease 0.7 (0.3-1.5); and brain cancer 1.3 (0.8-2.1). The standardized incidence ratio for brain cancer among children and young adult Livermore residents was elevated at a level that reached statistical significance (3.0; 95% CI=1.4-5.8) during the earliest decade studied, 1960-1969, but then gradually diminished over the next two decades. Brain cancer has been noted to be relatively sensitive to the effects of ionizing radiation (32).
However, the study found an excess of melanoma in young community residents. Among children and young adults residing in Livermore, melanoma rates were 2.4 times higher than expected, and among persons 24 years of age and under who had been born in Livermore, the rate was 6.4 times higher than expected. Both results were statistically significant.
Investigation of Community Cancer Rates in Livermore (Cancer Incidence in Livermore: 1988-1993 (61))
Subsequently, a review of cancer rates in eight census tracts in Livermore was undertaken by the Cancer Surveillance Section (formerly RCE) of CDHS. The years covered (1988-1993) were more recent than the earlier investigation. The Cancer Surveillance Section examined cancers of all anatomical sites (as one group), plus separately those cancer types well-known to be associated with exposure to radioactivity. Those sites included: bone, female breast, brain, leukemia (excluding chronic lymphocytic leukemia, which has not been associated with ionizing radiation), lung, prostate, and thyroid. Melanoma was also reviewed, although it was not historically considered a radiation-induced cancer. The review included all ages and was based on residence, and did not contain any information on whether cases were LLNL employees or family members of employees. One census tract in the southeastern section of Livermore (tract 4515) was of particular concern to residents.
In a cancer rate review, it is typical to calculate a confidence interval around the observed number of cases to assess the statistical likelihood of a result (the number of observed cases of cancer) occurring by chance. Consistent with the standard California cancer registry procedures, a 99% confidence interval was used for this purpose; any difference between the observed and expected numbers is not statistically distinguishable from chance if the expected number of cancers falls within the confidence interval for the observed number of cancers.
The review found that the incidence of cancer among residents of the eight census tracts was similar to the Bay Area as a whole: a total of 1295 cases were expected during this time period, and 1214 cases occurred. The number of melanomas in the community was not elevated (51 expected vs. 56 observed.) Similarly, in the census tract next to LLNL (4515), the overall number of cancers was not elevated: a total of 285 cancer cases were expected, and 257 actually occurred. In census tract 4515, 18 melanomas were diagnosed: seven in women (compared to five expected) and 11 in men (compared to six expected). Although the number of melanomas was greater than the number expected, it fell within a range that could have occurred by chance.
Table 3: Cancer and melanoma among community members
Table 3 in Appendix B shows the incidence ratios of cancer among different groups of community members studied. Confidence intervals are presented as well. (Please see discussion of confidence intervals presented as part of the summary of Report #1, "A study of cancer incidence in Lawrence Livermore Laboratory Employees.") Blank areas of the table reflect no data analyzed within that study. The table reflects the general lack of cancer elevation among the community, with the notable exception of melanoma. In addition to the main and consistent findings from the RCE/CDHS study of melanoma elevation across time and within both cohorts evaluated, a smaller, non-statistically elevation was found among dependents of LLNL employees who were KPMCP members.
(Birth Defects Around Livermore: 1983-1989 (62))
The California Birth Defects Monitoring Program of CDHS conducted a review of birth defects in the zip codes around the Livermore area. The review covered the years 1983-1989, the only years the program operated in Alameda County. The following data were reviewed: (1) the total number of birth defects; (2) the number of birth defects that have been associated with ionizing radiation (on the basis of what has been published in scientific literature), including chromosomal abnormalities, microcephaly, and autosomal dominant mutations; and (3) number of other major birth defects. The overall rate of birth defects was very similar to the statewide rate (2.5 per 100 live births in Livermore compared to 2.9 per 100 across the state). The numbers of specific birth defects were similar to or lower than statewide rates, and the number of other major birth defects was not significantly greater than expected in Livermore. (Specific numbers are not shown in order to protect confidentiality when the observed number of cases is less than six.)
In general, the incidence of cancer among LLNL employees was not higher than expected. However, a number of studies from the 1970s through the mid-1980s consistently found melanoma rates among employees of the LLNL to be approximately 3 times higher than expected.
Overall rates of cancer in community residents were not elevated over three decades (1960-1990). However, among children and young adults residing in Livermore, the melanoma rate was over 2 times higher than expected; and among children and young adults who had been born in Livermore, the rate was over 6 times higher than expected. More recently, cancer rates among Livermore residents have been found to be similar to the Bay Area as a whole, according to a report pending publication by LLNL. The number of melanoma cases occurring in a census tract bordering LLNL was greater than expected, but statistically within a range that could have occurred by chance.
In 1984, LLNL instituted an on-site medical surveillance program for melanoma. Early detection and treatment would be expected to prevent melanomas, as precancerous lesions can be treated before they develop into melanoma. Also, if exposures to radiation and chemicals at LLNL have decreased over time (3), risks associated with those activities would be commensurately reduced. According to LLNL, melanoma rates in the workforce have fallen to Bay Area averages since 1985.
Although overall cancer rates do not appear to be elevated among the LLNL workforce or community residents and no elevation of birth defects was noted, there appears to have been an excess of melanoma in both populations from before 1970 until the mid-1980s. A portion of the excess of melanoma might be in part due to earlier diagnosis of LLNL employees compared to community members, as publicity about melanoma possibly prompted heightened awareness in later years. However, as the elevation in melanoma exceeded the expected numbers among both early stage and invasive tumors, heightened awareness alone would not be expected to account for the increase.
Within the LLNL workforce, the search for workplace risk factors for the melanoma excess found several associated factors, including work around sources of ionizing radiation. Although LLNL's study findings did not include this association, independent researchers confirmed the findings through a re-analysis of the original data.
In the scientific literature, studies of radiation have had inconsistent results regarding melanoma. This may be due, in part, to the rare nature of melanoma compared to other skin cancers, or that melanoma may be less radiosensitive. More recent studies examining various sources of radiation (e.g., x-ray technicians, pilots and flight attendants [who are exposed to gamma radiation], and those with exposures from x-rays and radon) now include a number of large and well-conducted studies noting an increased melanoma risk.
Sources of exposure that would have affected community members are unclear, although it is known that tritium has been released over time, with most of the large releases from two accidents in the 1965 and 1970. This type of release would affect both LLNL employees and community members. Another possible exposure pathway to community members is potential contact with plutonium-contaminated sewage sludge distributed in the community, although the levels have been estimated to be well below a level required to cause health problems.
Because of the consistent historic elevations of melanoma among the LLNL workforce and the associations found between exposure to radiation and melanoma in this population, it would be useful for future epidemiological studies of radiation exposure (not specifically at LLNL) to consider addressing melanoma directly.
Because the previously elevated rates of melanoma in the community and LLNL have been described as decreasing more recently, data should be periodically reviewed to check the rates of melanoma to see if this trend is continuing.
The Public Health Action Plan contains a description of actions taken, to be taken, or under consideration by ATSDR and CDHS regarding this site. The purpose is to ensure that this health consultation not only identifies public hazards, but also provides a plan of action designed to mitigate and prevent adverse human health effects resulting from exposures to hazardous substances in the environment. CDHS and ATSDR will follow up on this plan to ensure that the actions are carried out.
- In 1999, CDHS prepared a public health assessment addressing the plutonium contamination in Big Trees Park (4).
- In 2002, ATSDR prepared two documents addressing tritium releases from LLNL (7,9).
- In 2001, CDHS prepared a health consultation addressing the historic distribution of plutonium-contaminated sewage sludge (63).
- In 2003, CDHS prepared a health consultation documenting community concerns relating to the LLNL Site (12).
- In 2003, CDHS attended a public meeting in the LLNL community and presented the public comment version of this health consultation.
Sumi Hoshiko, MPH
Environmental Health Investigations Branch
California Department of Health Services
Marilyn C. Underwood, PhD, Chief
Site Assessment Section
Environmental Health Investigations Branch
California Department of Health Services
COMMUNITY RELATIONS COORDINATOR
Gina Margillo, MA
Impact Assessment, Contractor to the
Environmental Health Investigations Branch
California Department of Health Services
ATSDR REGIONAL REPRESENTATIVES
William Q. Nelson
Regional Representatives, Region IX
Agency for Toxic Substances and Disease Registry
ATSDR TECHNICAL PROJECT OFFICER
Tammie McRae, MS
Environmental Health Scientist
Division of Health Assessment and Consultation
Superfund Site Assessment Branch, State Programs Section
The California Department of Health Services under a cooperative agreement with the Agency for Toxic Substances and Disease Registry (ATSDR) prepared the "Review of Health Studies Relevant to Lawrence Livermore National Laboratory and the Surrounding Community" health consultation. The document is in accordance with approved methodology and procedures existing at the time the health consultation was prepared.
Tammie McRae, MS
Technical Project Officer, SPS, SSAB, DHAC
The Division of Health Assessment and Consultation, ATSDR, has reviewed this public health consultation and concurs with the findings.
Chief, State Program Section, DHAC, ATSDR
1 CDHS requested the report this information is based on, but it was not available, per LLNL.
2 Using a chi-square test for independence, we found that although a difference between the two time periods (1974-1979 and 1980-1985) might exist, it was not statistically significant. This was true for LLNL as well as in the community. The chi-square test is interpreted using a p-value, which indicates how likely it is that the results occurred by chance. A p-value of less than 0.05 means that the likelihood of a result occurring by chance alone is less than 5%. The chi-square for LLNL resulted in a p-value=0.46, and for the community, p-value=0.32. Both values are well over 0.05, suggesting that the results could have occurred by chance (i.e., there was no difference in lesion thickness between the two time periods, or any difference could have occurred by chance). On the second question regarding lesion thickness at LLNL compared to the community, again, although a difference might exist, statistical testing did not detect a difference in thickness between LLNL lesions and community lesions in either time period (earlier period, p=0.19; later period, p=0.50).
3 The author of this report contacted the current LLNL Medical Director, Dr. James Seward, to obtain information on the original algorithm and why and how it was changed. The Medical Director reported that he contacted as many authors of the report as possible, but none could recall the original algorithm nor why it was changed, but that all persons queried agreed it was a minor change.