3. Analysis of El Dorado Hills Air Samples _____________
3.1 Reported Concentrations
A large number of mineral particles were reported by Lab/Cor. Table 3-1 shows the number of particles counted in each activity that were identified as asbestos fibers (count sheet codes "F", "CF", and "MF").[4]
Lab/Cor reported concentrations for several different size classifications of mineral particles. Of interest are three classifications:
Primary Structures: any structure that contains (at a minimum) a mineral particle that is at least 0.5 µm long with a minimum aspect ratio of 3:1. PCME Asbestos Structures: particles that are phase contrast microscopy equivalent (PCME) in size: longer than 5 µm, at least 3:1 aspect ratio, and wider than 0.25 µm.
Protocol Structures 5 : particles longer than 5 µm and thinner than 0.5 µm. The Protocol structures are further divided into chrysotile and amphibole categories, as well as into two length classifications: 5 µm – 10 µm and =10 µm.
The first group is simply the total number of particles counted. The other two groups are size classes used (or proposed for use) in risk analyses.
Table 3-2 shows the reported concentrations for all mineral particles in for these three classifications (these samples exclude field blanks, filter blanks, performance samples, and quality control analyses). Samples with no reported fibers were calculated using "0" as the concentration in accordance with statistical theory.6 The median concentrations were 0.0040 s/cc (primary structures), 0.0010 f/cc (PCME structures), and 0.0 f/cc (protocol structures). Asbestos fibers longer than 5 µm have historically been related to fibrogenicity and carcinogenicity.[7] A statistically insignificant number of protocol fibers were counted by Lab/Cor. Table 3- 3 summarizes these counts.
Statistical comparisons for each activity comparing personal, area, and reference area (background) protocol fiber concentrations show no difference among the types of samples or between personal and reference concentrations. Protocol structures (those mineral fibers that are 10 µm and longer and thinner than 0.5 µm) have been shown to be useful in cancer risk estimation. When these concentrations are examined on an activity basis, Tables 3-4 and 3-5, there are no statistically significant differences in concentrations. Protocol fibers were detected during six activities on the personal samples and in only one set of reference samples; however there is no significant difference in concentration, regardless of whether the comparisons are made on an activity-basis or over the combined data.
The median concentrations for these size classifications are not significantly different than background concentrations. One study from 1984 8 indicates the national average to be 0.0004 f/cc for fibers longer than 5 µm. The Health Effects Institute-Asbestos Research report 9 indicates the background airborne concentrations for PCME fibers to range from 0 – 0.008 f/cc. These numbers are not statistically different than the median for the El Dorado Study (0.0010 f/cc).
3.2 Mineral Particle Identification
Large numbers of mineral particles were enumerated on the laboratory count sheets. Including quality control test samples, 6873 mineral particles 10 were enumerated, of which 5624 particles were counted during the original sample analyses, 779 during quality assurance testing, and 470 on five samples labeled as "performance" samples. There were 3948 amphibole and 2925 chrysotile structures counted in the original analyses.
The principle amphibole particle reported was actinolite (a very small number of other amphibole mineral particles were reported). Nearly seventy-two percent (72%) of [11] actinolite particles (those with some information on the chemical composition of the particle) contained aluminum. This is a very significant finding because aluminum is a minor component of actinolite. Examples of the observed chemistries are shown in Figure 3-1.
The mineral actinolite has a defined chemical composition that may contain only a very small amount of aluminum. Changes to the chemical composition will distort the crystal structure of the mineral, eventually (with sufficient chemical substitution) resulting in a different mineral structure. According to Leake et al 12 and Deer, Howie and Zussman 13 , the aluminum content of asbestiform actinolite is restricted to < 0.3 aluminum atoms pfu (per formula unit)14 . Actinolite with aluminum compositions above 0.3 Al pfu have aluminum concentrations too high to form asbestos fibers. As noted by Deer Howie and Zussman (page 182), "specimens that contain more than a very small amount of aluminum do not have an asbestiform habit". Verkouteren found a maximum of 1.5 percent Al2O3 or 0.26 pfu in a recent analysis of 34 actinolite asbestos samples, but estimated that under some circumstances, when aluminum substitutes for silicon in the tetrahedral site, and certain other cations are present, non-asbestos actinolite could contain as much as 1.0 pfu and still conform to the Leaky nomenclature.15,16 Dorling 17 examined one actinolite asbestos sample as part of a general evaluation of the characteristics of calcic amphiboles. The aluminum content of the sole asbestiform actinolite sample was 0.25 percent Al2O3.
Using the Verkouteren observation of 1.5 percent, based on the reported chemistries, sixty-three percent (63%) of the reported actinolite particles have been misclassified as asbestos. Table 3-6 summarizes the EDXA data reported for the actinolite particles. There are 341 particles that were identified as actinolite on the basis of the chemical composition identified by the EDXA. As shown in the Table, the aluminum content of the reported actinolite particles ranged from no aluminum up to 8.7 percent Al2O3. Only seven percent (7%) of the spectra report no aluminum; the median aluminum content is two percent Al2O3 (2%), and sixty-three percent (63%) exceeded 1.5 percent Al2O3. Sixty-three percent (63%) of the reported actinolite particles cannot be asbestos fibers due to the excessive amounts of aluminum reported by the laboratory. The EDXA compositional data for the reported actinolite particles are shown in Figure 3-2. In this graph, the iron (Fe), magnesium (Mg), and aluminum (Al) from each of the analyses are shown (the aluminum scale has been expanded for clarification). Sixty-three percent exceed 1.5 percent Al2O3.
3.3 Mineral Particle Size Analysis
There are 2386 amphibole particles identified as single amphibole asbestos fibers (codes "F", "MF", or "CF") on the original analyses. These "fibers" represent the amphibole minerals that were counted as single entities (not bundles of fibers or other complex structures) and represent the basic structures that were observed in the airborne samples. The average dimensions of these particles (of all lengths), shown in Table 3-7, indicate they are 5.3 µm long with a width of 0.8 µm and an aspect ratio of about 6:1. These dimensions demonstrate the amphibole particles to be a population of non-asbestos particles. This distribution is shown in Figure 3-3 which is a plot of particle length vs. particle width. The plot shows a general trend toward longer and thicker particles (lower left to upper right in the graph), an indicator of a non-asbestos particle distribution.
Analysis of the laboratory data shows that thirty-five percent (35%) of all amphibole particles that the El Dorado Hills Study identified as amphibole asbestos fibers have aspect ratios of less than 5:1 and do not, even under the general ISO 10312 standard, meet the definition of an "asbestos fiber."
The dimensions of the particles 5 µm and longer do not conform to the recommended EPA [18] definition of asbestos which says the average aspect ratio is 20:1 or more. Table 3-8 shows the dimensions for these "fibers". With an average aspect ratio of only 6.3:1, these dimensions are not indicative of an asbestos population as defined by the US Environmental Protection Agency.19,20
When compared with true asbestos fibers, the particles from El Dorado Hills are much wider. Figure 3-4 shows the Jamestown 21 amphibole asbestos to be much thinner than the El Dorado Hills amphibole particles.
3.4 The Amphibole Particles are not From an Asbestos Population
As noted in the definition of asbestos presented in an EPA analytical procedure, an asbestiform population is characterized by: a) mean aspect ratios of 20:1 or greater for fibers 5 µm and longer; b) fiber widths less than 0.5 µm; and c) parallel fiber occurring in bundles. The reported particles fail to meet the aspect ratio and width specifications for asbestiform populations as shown in Table 3-8. In fact, only 36 particles =5 µm in length (three percent of particles =5 µm) have aspect ratios greater than 20:1 and only 50 of the particles =5 µm in length (four percent of the particles =5 µm in length) are thinner than 0.5 µm. Only seven fibers (0.3 percent of all amphibole fibers) are =10 µm in length and thinner than 0.5 µm, a class of fibers used in recent risk models (the Berman-Crump model).
Within the data set, there were 85 amphibole bundles counted during the original analyses. The ISO 10312 [22] analytical method used by Lab/Cor defines a "bundle" as "a grouping of apparently attached parallel fibres". Thus a bundle is composed of (at a minimum) of two (2) fibers, however there are usually more. The average dimensions of the bundles are shown in Table 3-9. The reported width of a bundle is the overall width of the bundle and not the width of the component fibers.
Fiber bundles are a basic characteristic of asbestos. The dimensions of the bundles can be used to estimate the maximum widths of the asbestos fibers in this study. Assuming a bundle contains a minimum of two (2) fibers, the maximum width of the asbestos amphibole fiber dislodged from a bundle would be 0.4 µm. This is an extremely conservative estimate since asbestos bundles typically contain more than two fibers. In the EPA data, the vast majority of particles (eighty percent, 80%) are greater than 0.5 µm in width. If there was a significant asbestos population, there would be significantly more thin fibers than what the sample data indicate.
The vast majority of the amphibole particles counted in this study are non-asbestiform. True amphibole asbestos fibers are characterized by widths of 0.2 to 0.4 µm; amphibole particles wider than 0.5 µm are non-asbestos particles. There are 1901 amphibole particles (excluding bundles) that are 0.5 µm and wider or eighty percent (80%) of the amphibole particles. For particles 5 µm and longer, 1273 (ninety-six percent, 96%) of these particles are =0.5 µm wide. Figure 3-5 shows some of the particles observed by Lab/Cor that are non-asbestos in habit but were reported as asbestos. None of these particles illustrate asbestos characteristics: parallel sides, high aspect ratio, and proper termination (ends of the fibers).
The diffraction pattern analyses performed by Lab/Cor support the labeling of these particles as non-asbestos. Lab/Cor produced evaluations of the selected area electron diffraction (SAED) data that provides information related to the crystal structure of a mineral. Zone axis SAED patterns are one indicator for whether a particle is asbestiform or not. Ring [23] has shown asbestos fibers are generally associated with lower order zone axis patterns (such as [0 X X] or [1 X X]), while non-asbestiform particles generally have higher order zone axis patterns (usually [=3 X X]).24,25 The higher order zones are also those that have a possibility of matching other minerals due to the (relatively) large error associated with the measurements. The produced data listed the mineral identification, the matching crystal zone index (the zone that best fit the SAED pattern), and the reported zone index for the pattern. Table 3-10 summarizes the number of patterns for each identified mineral and the number of times the matched zone was reported as the zone axis for the pattern. When the reported zone axis is not the match zone, the analyst has incorrectly analyzed the pattern. The majority of the identified zone axis SAED patterns are higher order patterns, indicative of non-asbestos minerals.
3.5 Quality Assurance Testing
Lab/Cor re-analyzed a number of samples as part of an overall quality assurance test program. Of interest are the re-analyses that were performed on the same grid openings as were originally analyzed. The comparison of these two sets of count sheets permits an estimation of the accuracy of the original analysis. There were two groups of these analyses: 1) the original and one quality assurance test; and 2) the original and two quality assurance analyses. The first group permits an estimation of the overall accuracy of the counting; the second group provides information on the cause of the different counts.
Paired Analyses: Within the produced data, 16 samples had a second analysis performed on the same grid openings as were analyzed in the original analysis, permitting an estimate of the true counting rate by the laboratory analysts. Because there is no way to independently verify the actual analysis, it was assumed that a reported mineral particle is a true count when reported by both the original and quality assurance analyses. The remaining structures represent miscounts by either the original analyst (that is, a particle was observed in the original analyses but not reported in the quality assurance analysis) or by the quality assurance analyst (a fiber not reported in the original analysis but reported in the quality assurance analysis). Table 3-12 summarizes these data; the 16 sets of data are attached in Appendix 1. For this analysis, the counts are based on the number of primary structures counted. The data shown in Table 3-11 indicate the original mineral particle counts are inflated, on average, about seventy-eight percent (78%) above the agreed upon number of particles on each sample. The only explanation for the complete failure to verify the presence of a fiber in the original analyses on two of these samples (RHB-H2-3FD-100304 and SRA-R02-100604) is that the analysts must have counted different grid openings (even though the count sheets indicate otherwise).
Original and Two Quality Assurance Analyses: Three samples (NRA-R02-101104, SRA-R05-100604, and CC2-H8-1CT-100304) were analyzed three times on the same grid openings. The third analysis provides an opportunity to estimate the False Positive and False Negative percentages for an analysis. For this evaluation, a True Positive is defined as two out of the three analyses reporting a particle. A False Positive is only one of the three analyses reporting a particle. A False Negative was not reporting a structure observed in the other two analyses of the grid opening. Table 3-12 summarizes these data. As a reference for evaluating True Positives, False Negatives, and False Positives, proficient analysts at NVLAP accredited laboratories are expected to have rates in excess of eighty percent (80%) for True Positives, less than twenty percent (20%) for False Negatives, and less than ten percent (10%) for False Positives.[26]
The reported data had an average False Positive Rate of thirty-five percent (35%), far exceeding the NVLAP guideline of less than ten percent (10%).
All of these "same grid opening" analyses were incorrectly determined to be of acceptable quality by Lab/Cor. Lab/Cor determined whether an analysis was acceptable or not by comparing the number of particles counted in the second analysis to the Poisson confidence interval for the original count.27 The use of Poisson counting statistics is acceptable when comparing the data from different areas of the filter, but it is not an acceptable procedure when the same grid openings are examined. Poisson statistics are used to account for the distribution of fibers on the filter. However, when the same areas of the filter are analyzed multiple times, the issue of variable fiber distribution is no longer in question, rather whether the same fibers are counted or not. Further evaluation of the QA data show that the total number of QA analyses were less than generally accepted (eight percent [8%] versus ten percent [10%]) and that the QA was not even performed during much of the project (thirty-eight percent [38%] of the sample data had no QA analyses associated with the samples).
3.6 Blank Samples
Within each set of data, several samples were reported as either a "Field Blank" or a "Filter Blank". All of these filters were shown to be free of mineral particles. However, with few exceptions, all of these "Blank" samples were reported to have a volume of air sampled. The EPA report did not describe the purpose of samples labeled as "Blank"; it cannot be determined what air was sampled by these filters. By definition, a blank has no air filtered through the sample. This traditional definition is reflected in the various analytical methods (such as ISO 10312, AHERA, and NIOSH 7400) used for asbestos exposure analyses.
A summary of the "blank" filters is shown in Table 3-13.
3.7 Air Sample Volumes
For many of the samples, the volume of sampled air exceeded 2500 L. Prior studies (such as EPA/560/5-88-002, Assessing Asbestos Exposure in Public Buildings) have shown that sampling high volumes of air resulted in filters with excessive particulate that prevented precise and accurate counts of the asbestos fibers. Because high particulate loads may bias the analytical results, the National Voluntary Laboratory Accreditation Program (NVLAP) has restricted acceptable particle loadings to less than ten percent (10%) coverage on the filter's surface.28 Because the samples in the El Dorado study were collected outdoors, it is highly likely that the particulate loading on the samples with air volumes in excess of 2500 L exceed the accepted ten percent (10%) limit. The overall particle loading on a filter was not reported by Lab/Cor.
3.8 Transcription Discrepancies in Laboratory Data Sheets
While reviewing the laboratory data sheets, a number of transcription errors were noted, leading to a six percent (6%) error rate. The following Table 3-14 lists these errors. The majority of these errors appear to result from incorrect transcription of written data to computer format, however without the written count sheets this conclusion cannot be confirmed.
There were 351 analyses (excluding quality assurance tests and pending results) reported in the May report. With 22 observed transcription errors (Table 3-15), this amounts to a six percent (6%) error rate.
There was also a discrepancy in the reported PCM-equivalent structure concentration for sample CC2-L6-3CC-100304. In the EPA report, Table 5-9 shows a concentration of 0.00491 s/cc while Table 5-10 shows 0.00393 s/cc for the same sample. The correct concentration is 0.00393 f/cc.
28 The NVLAP guideline may be found at http://ts.nist.gov/ts/htdocs/210/214/docs/lb_7_2002.pdf.
[4] Chrysotile fibers were observed in air samples generally associated with activities on the ball fields. The median chrysotile concentration is <0.0001 for all three size classifications, indicating the majority of samples contained no chrysotile. The highest chrysotile concentrations were associated with the activities at the Community Park south baseball field. More than seventy percent (70%) of samples with statistically significant chrysotile counts were associated with baseball field activities. Less than two percent of the chrysotile structures were longer than 5 µm. The focus of this report will be on the reported amphibole particles. 5 D. Wayne Berman and K. S. Crump (2003). "Final Draft: Technical support document for a protocol to assess asbestos-related risk," U.S. Environmental Protection Agency, Peer-reviewed consultation held in San Francisco on February 25-26, 2003. 6 Oehlert, G. A; Lee, R. J.; and Van Orden, D. R. (1995). "Statistical Analysis of Asbestos Fibre Counts", Environmetrics, 6, p. 115 - 116.
[7] Agency for Toxic Substances and Disease Registry (2002). "Expert Panel on Health Effects of Asbestos and Synthetic Vitreous Fibers (SVF): The Influence of Fiber Length; Premeeting Comments", October 29-30, 2002, New York, NY. 8 National Research Council (1984). Asbestiform Fibers: Nonoccupational Health Risks, National Academy Press. 9 HEI-AR (1991). Asbestos in Public and Commercial Buildings: A Literature Review and Synthesis of Current Knowledge, p. 4-38 to 4-39. 10 Each line of data on the count sheets has been counted as a separate entry for the overall number of particles counted.
[11] There were 3071 reported actinolite particles with some information on the particle's chemical composition. 12 B. E. Leake et al (1997). :Nomenclature of Amphiboles: Report of the Subcommittee on Amphiboles of the International Mineralogical Association, Commission on New Minerals and Mineral Names", American Mineralogist, 82, p. 1019-1037. 13 W. A. Deer, R. A. Howie, and J. Zussman (1997). Rock-Forming Minerals: Double-chain silicates, Vol 2, second edition, p 137 – 145. 14 0.3 aluminum atoms pfu is equivalent to < 2% Al2O3 when the content is reported as oxide compounds. 15 J. R. Verkouteren and A. G. Wylie (2000). "The tremolite-actinolite-ferro–actinolite series: Systematic relationships among cell parameters, composition, optical properties, and habit, and evidence of discontinuities", American Mineralogist, 85, p. 1239 – 1254. 16 The maximum reported Al2O3 content of the 103 samples reported by Verkouteren was 4.3% found in a byssolitic (non-asbestos) actinolite sample. 17 M. Dorling and J. Zussman (1987). "Characteristics of asbestiform and non-asbestiform calcic amphiboles", Lithos, 20, p. 469 – 489.
[18] U.S. Environmental Protection Agency (1993). "Test Method: Method for the Determination of Asbestos in Bulk Building Materials", EPA/600/R-93/116, p. A-1. 19 M. E. Beard (1992). Letter to Sally Sasnett, November 3, 1992. Asbestos fibers have mean aspect ratios "ranging from 20:1 to 100:1" for fibers longer than 5 µm. 20 U.S. Environmental Protection Agency (1993). "Method for the Determination of Asbestos in Bulk Building Materials", EPA 600/R-93/116. 21 J. M. G. Davis, J. Addison, C. McIntosh, B. G. Miller, and K. Niven (1991). "Variations in the Carcinogenicity of Tremolite Dust Samples of Differing Morphology", Annals of New York Academy of Sciences, 643, p. 473 – 490.
[22] International Organization for Standardization (1995). "Ambient Air – Determination of asbestos fibres – Direct-transfer transmission electron microscopy method", ISO 10312.
[23] S. J. Ring (1980). "Identification of Amphibole Fibers, Including Asbestos Using Common Electron Diffraction Patterns", draft report dated March 31, 1980. 24 R. J. Lee, J. S. Lally, and R. M. Fisher (1978). "Identification and Counting of Mineral Fragments", in Proceedings of the Workshop on Asbestos: Definitions and Measurement Methods, National Bureau of Standards, July 18 – 20, 1977, Special Publication 506, p. 387-402. 25 A. M. Langer, R. P. Nolan, J. Addison (1991). "Distinguishing Between Amphibole Asbestos Fibers and Elongate Cleavage Fragments of Their Non-Asbestos Analogues", in Mechanisms in Fibre Carcinogenesis, p. 253-267.
[26] NVLAP (1995). Airborne Asbestos Analysis, NIST Handbook 150-13, item 3.7.d, page C-6. 27 A summary page, dated 1/4/2005, of the original and quality control analyses was included in the materials received on September 6, 2005.
Summary -- Intro -- Data Source -- Analysis of El Dorado Hills Air Samples -- Analysis of El Dorado Hills Soil Samples
