Temporal Variability of Tungsten and Cobalt in Fallon, Nevada

Background Since 1997, Fallon, Nevada, has experienced a cluster of childhood leukemia that has been declared “one of the most unique clusters of childhood cancer ever reported.” Multiple environmental studies have shown airborne tungsten and cobalt to be elevated within Fallon, but the question remains: Have these metals changed through time in correspondence with the onset of the leukemia cluster? Methods We used dendrochemistry, the study of element concentrations through time in tree rings, in Fallon to assess temporal variability of airborne tungsten and cobalt since the late 1980s. The techniques used in Fallon were also tested in a different town (Sweet Home, OR) that has airborne tungsten from a known source. Results The Sweet Home test case confirms the accuracy of dendrochemistry for showing temporal variability of environmental tungsten. Given that dendrochemistry works for tungsten, tree-ring chemistry shows that tungsten increased in Fallon relative to nearby comparison towns beginning by the mid-1990s, slightly before the onset of the cluster, and cobalt has been high throughout the last ~ 15 years. Other metals do not show trends through time in Fallon. Discussion Results in Fallon suggest a temporal correspondence between the onset of excessive childhood leukemia and elevated levels of tungsten and cobalt. Although environmental data alone cannot directly link childhood leukemia with exposure to metals, research by others has shown that combined exposure to tungsten and cobalt can be carcinogenic to humans. Conclusion Continued biomedical research is warranted to directly test for linkage between childhood leukemia and tungsten and cobalt.

We assessed recent temporal variability in environmental tungsten and cobalt in Fallon, Nevada ( Figure 1A), where 16 cases of child hood leukemia were diagnosed from 1997 to 2002(Expert Panel 2004 and an addi tional case was announced in December 2004 (Nevada State Health Division 2004). All cases but one were acute lymphocytic leu kemia. As of the 2000 Census, Fallon has 7,536 residents (U.S. Census Bureau 2000), and its pediatric population up to 19 years of age is approximately 2,400 children. Counting all 17 cases in the time span of 8 years, the rate of childhood leukemia in Fallon is many times higher than the expected rate of 4.3 cases per 100,000 children (0-19 years of age) per year (National Cancer Institute 2007). This cluster has a very small likelihood of being a random event (Expert Panel 2004), and Fallon has been declared "one of the most unique clusters of childhood cancer ever reported" (Steinmaus et al. 2004).
Extensive research has been conducted in Fallon to determine if an environmental cause might be playing a role in its child hood leukemia [Agency for Toxic Substances and Disease Registry (ATSDR) 2002, 2003a, 2003b, 2003c; Centers for Disease Control and Prevention (CDC) 2003a, 2003b; Moore et al. 2002;Seiler 2004;Seiler et al. 2005]. A theory known as population mixing has also been proposed for Fallon (Kinlen 2004; Kinlen and Doll 2004). Among other environ mental findings, a consensus has emerged that the heavy metal tungsten is elevated in Fallon (CDC 2003a;Sheppard et al. 2006aSheppard et al. , 2007bSheppard et al. , 2007c. Cobalt also is elevated in Fallon (ATSDR 2003a;Sheppard et al. 2006aSheppard et al. , 2007bSheppard et al. , 2007c. The temporal variability of tungsten and cobalt in Fallon over the last several years is not known because the environmental moni toring techniques used thus far cannot resolve changes through time. Dendrochemistrythe meas urement and interpretation of ele ment concentrations in tree rings (Amato 1988)-can document temporal variability of elements in the environment with up to annual resolution. Dendrochemistry has been used in studies of temporal patterns of vari ous heavy metals in the environment, includ ing lead (Hagemeyer and Weinand 1996), nickel (Yanosky and Vroblesky 1992), cad mium (Guyette et al. 1991), and mercury (Li et al. 1995). Dendrochemical measure ments are typically used to evaluate relative changes through time in environmental avail ability of elements as well as to compare their absolute concentrations across different trees or different sites (Lewis 1995). Accordingly, dendrochemistry was used in Fallon to assess temporal variability of tungsten and cobalt since the late 1980s-that is, since before the onset of the cluster of childhood leukemia.

Materials and Methods
Fallon, Nevada. We selected cottonwoods (Populus sp.) in Fallon for analysis. Sampling was targeted at an area near the center of town, just northwest of the intersection of the two main highways ( Figure 1B), which has been identified as the source area of airborne tungsten (Sheppard et al. 2007b). Trees were selected from around an industrial facility specializing in hardmetal metallurgy, which uses tungsten carbide and cobalt to harden steel (Harris and Humphreys 1983). The Nevada Division of Environmental Protection has considered this facility to be a candidate source of tungsten in Fallon (Mullen 2003). For comparison data, we sampled cotton woods and elms (Ulmus sp.) in the towns of Lovelock, Fernley, and Yerington ( Figure  1A). We selected four time periods of rings to meas ure for concentrations of multiple ele ments. Two periods predate the 1997 onset of excessive childhood leukemia in Fallon (1989-1992 and 1993-1996) and two peri ods postdate it (1997-2000 and 2001-2003 or 2001-2004, depending on the last ring available for meas urement).

Independent test case.
To independently test the accuracy of dendrochemistry specifi cally for tungsten, we repeated this exper iment in a different small town that has a known source of airborne tungsten. Sweet Home, Oregon (Figure 2A), has a tung stenpowder industry that was established in November 2000. Spatial environmental techniques have confirmed that tungsten is elevated in the area immediately surrounding this known industrial source compared with the rest of Sweet Home, with other towns, and with outlying open areas (Sheppard et al. 2007a). Douglasfirs (Pseudotsuga menziesii) and cottonwoods near the tungsten indus try were sampled ( Figure 2B). For com parison data, Douglasfirs were sampled at a rural location just outside of Crawfordsville, about 10 km from Sweet Home (Figure 2A). Approximately the same four time periods of rings that were measured in the Nevada trees were selected in the Oregon trees for measure ment of concentrations of multiple elements.
Field sampling and sample preparation. Field sampling and sample preparation meth ods followed standard protocols for dendro chemical research. We collected increment cores using a 5.15mm diameter Haglof borer (Forestry Suppliers, Inc., Jackson, MS). The borer was cleaned after each use with 70% isopropyl alcohol. In most cases, only one core per tree was collected to maximize the number of trees sampled rather than the number of cores within trees (McClenahen et al. 1989).
To see ring growth more clearly, we cut a minimal surface on one transverse side of each core using a stainlesssteel razor blade. Growth rings were identified visually using standard anatomic features that occur in rings (Kramer and Kozlowski 1979). Contamination of the core samples with tungsten and other met als from the increment borer itself is possible because borers are made of hardened steel. To eliminate this potential contamination, the outer surface of the cores was removed by laser trimming, yielding inner cores that had never been touched by metal tools (Sheppard and Witten 2005). Inner cores were then bro ken into the time periods using a nonmetallic, ceramic knife.
ICP-MS measurements. The wood of rings was chemically digested and then analyzed by inductively coupled plasma mass spectroscopy (ICPMS). Before analysis, samples were freezedried to a constant weight and weighed into precleaned, preweighed, trace metal-free polypropylene centrifuge tubes. For every 25 mg of sample, 1 mL concentrated Optima grade nitric acid was added to the tube. The samples were allowed to sit at room tempera ture for 2 days and then were digested at 70°C in an ultrasonic bath for 3 hr. Following diges tion, the sample tubes containing the digestate were reweighed to calculate dilution factors. An aliquot of digestate (∼ 0.25 g) was gra vimetrically diluted by a factor of approxi mately 20 with ultrapure 18.2megaOhm/cm water and spiked with three internal standards: beryllium (20 ppb), indium (10 ppb), and bismuth (5 ppb).
To calibrate the ICPMS data, we pre pared linearity standards from multielement calibration standards obtained from High Purity Standards (Charleston, SC). Beryllium, indium, and bismuth internal standards were added to the linearity standards at approxi mately 20 ppb (for beryllium), 10 ppb (for indium), and 5 ppb (for bismuth). We used four standard points to calibrate the instru ment for all elements of interest. We calcu lated the exact concentrations for all standards, and these data were used to create the linear calibration curve of instrument response versus concentration for each analyte. The linearity standards were reanalyzed repeatedly during the analytical run to ensure continuous correct instrument response. Solutions were measured for lithium, aluminum, manganese, cobalt, nickel, copper, zinc, strontium, molybdenum, silver, cadmium, tin, antimony, cesium, tan talum, tungsten, thallium, lead, and uranium. Limits of detection were mostly ≤ 10 ppb. Sample values less than the limit of detection were considered missing values.
Statistical analysis. As a conservative quan titative analysis, we calculated medians for each metal and time period. The median is insensitive to outlier values, which can be an issue when sample size is small (Sokal and Rohlf 1981). Samples were compared statistically using the onetailed MannWhitney test of differences in cumulative distribution functions. The null hypothesis of no difference between samples applied to all tests, but the alternative hypoth eses differed depending on the samples being tested: a) tungsten and cobalt increase through time in Fallon or Sweet Home; b) tungsten and cobalt are higher in Fallon or Sweet Home than in comparison areas; or c) temporal patterns for tungsten and cobalt are different from those of other metals.  This reflects the fact that the sampled area in Sweet Home is industrial ( Figure 2B) and therefore generally elevated with metals, whereas the forest outside of Crawfordsville is relatively removed from point sources of pollution. Median treering tungsten in Sweet Home does not vary through the first three time periods, but it increases in the last period-the only period that fully postdates the establishment of the tungsten industry in Sweet Home ( Figure 3A). Median treering tungsten also increases during the last period in trees outside of Sweet Home, but not by as much as in Sweet Home. Considering all sampled trees within Sweet Home, the tung sten increase through time is borderline sig nificant (Table 1). Looking more closely in Sweet Home, temporal variability of tungsten is higher in the cottonwoods than in the Douglasfirs ( Figure 3B). The tungsten increase through time in just the cottonwoods within Sweet Home is significant (Table 1). Temporal smoothing of environmental signals can be an issue for dendrochemistry (Hagemeyer 1993), partly because of tree physiologic rea sons (Smith and Shortle 1996). The damped temporal variability in the Douglasfirs might be an example of this effect, which appears not to be so strong in the cottonwoods. Additional research is merited to determine why cotton woods express more temporal variability.

Independent test case.
Other representative trace metals, includ ing cobalt, do not increase significantly through time within Sweet Home (Table 1). This independent test case confirms the accu racy of dendrochemistry for showing temporal variability of environmental tungsten, espe cially when using cottonwoods, the principal species used in Nevada.
Fallon, Nevada. For the earliest time period (1989)(1990)(1991)(1992), before the onset of excessive childhood leukemia in Fallon, median treering tungsten in Fallon is not statistically different from that of comparison towns ( Figure 4A). However, for the next three time periods, median treering tungsten in Fallon increases whereas that of comparison towns remains relatively constant. For these three periods, Fallon medians are higher than those of comparison towns, and this tungsten increase through time in Fallon is significant (Table 1).
Median treering cobalt in Fallon is higher than in comparison towns for all peri ods ( Figure 4B), but there is no significant increase in cobalt through time within Fallon (Table 1). Other representative trace metals are not consistently higher in Fallon than in comparison towns ( Figure 4C-E); the signifi cant differences for cadmium are attributed to the medians from the other towns going down ( Figure 4D). These other representative trace metals also do not increase consistently    (Table 1). From this dendrochemical assessment, tungsten is unique in Fallon by its increase since the mid 1990s-that is, since slightly before the onset of excessive childhood leukemia there. Cobalt is also notable for being high within Fallon throughout the last ~ 15 years.

Discussion
Fallon is distinctive spatially by its elevated airborne tungsten and cobalt relative to comparison towns and outlying desert areas (ATSDR 2003a;CDC 2003b;Sheppard et al. 2006aSheppard et al. , 2007c. Now, based on replicated treering chronologies of multiple metals and backed up with an independent test of dendro chemistry of tungsten around a known source of tungsten, Fallon is also distinctive tem porally by its increase in tungsten beginning by the mid1990s as well as by its elevated cobalt since at least the early 1990s. Although environ mental data alone cannot directly link childhood leukemia with exposure to metals, the temporal cooccurrence of these metals with excessive childhood leukemia beginning by 1997 reinforces previous conclusions that continued biomedical research is warranted to directly test for linkage between childhood leukemia and exposure to tungsten and cobalt (CDC 2003a;Sheppard et al. 2006aSheppard et al. , 2006bSheppard et al. , 2006cSheppard et al. , 2007bSheppard et al. , 2007c. Sweet Home differs from Fallon in that it does not have excessive cases of child hood leukemia or other cancers (Sherman and Pliska 2005), raising the question: What might be causing this apparent inconsistency in the temporal cooccurrence of increasing airborne tungsten with or without excessive childhood leukemia? On the environmen tal side, the areal extents of airborne tung sten in these two towns differ substantially. In Fallon, elevated airborne tungsten extends out from the identified source area (Sheppard et al. 2007b) for up to 3 km (Sheppard et al. 2006a), corresponding to an area that includes residences and schools (Sheppard et al. 2006b). By contrast, in Sweet Home elevated airborne tungsten extends out from the known source at most for only 0.5 km, corresponding to an area that is mostly indus trial and that includes few residences and no schools (Sheppard et al. 2007a). If nonoccu pational exposure to elevated airborne tung sten were related to childhood leukemia, then variability in areal extent of exposure could be a consideration for explaining different rates of disease occurrence.
Little research on tungsten and cobalt with cancer has been published, but the few stud ies that do exist are suggestive. Simultaneous exposure to tungsten and cobalt has converted human osteoblastlike cells into the tumori genic phenotype (Miller et al. 2001), and it has activated the expression of genes related to cancer (Miller et al. 2004). Simultaneous exposure to cobalt and tungsten carbide, which might occur as a byproduct of hard metal metallurgy (Lombaert et al. 2004), appears to have a synergistic carcinogenic effect (Lasfargues et al. 1992;Lison and Lauwerys 1992;Van Goethem et al. 1997). The International Agency for Research on Cancer (IARC 2003) has declared cobalt and tungsten carbide together to be a prob able carcinogen to humans based on sufficient evidence. This allows for a possible linkage between childhood leukemia and concurrent exposure to both tungsten and cobalt, but research directed more specifically at child hood leukemia is needed to evaluate the role of these metals. In one example, tungsten ore administered to preexisting human leukemia cells in the laboratory increased their growth by 170% compared with control samples over a 72hr culture period (Sun et al. 2003).

Conclusion
Additional research has been called for to explain the high levels of tungsten in urine of residents of Fallon (Expert Panel 2004), and toxicologic study of tungsten has been requested (ATSDR 2004). We concur with these calls for more research to evaluate the potential link between childhood leukemia and exposure to both tungsten and cobalt. We also encourage continued environmental research in Fallon to confirm current and past airborne exposures and to definitively identify their source.