Document 6Yeg3kakpy5p3bX2e1pqkRmE

I REVIEW OF "CHRYSOTILE ASBESTOS FIBERS IN DRINKING WATER FROM ASBESTOS-CEMENT PIPE" By 1 oo T. A. Scripps , J. P. Leineweber , H. L. Olson and J. F. Welch4 * Statistician, Research Department, Manville Corporation, Denver, Colorado 2 Technical Director, Health, Safety and Environment Department,' Manville Corporation, Denver, Colorado O Manager, Environmental and Engineering Development, Pipe Division, Johns-Manville Corporation, Denver, Colorado 4 Vice President, A/C Pipe Producers Association, Arlington, Virginia CAPCO JEN 0010501 TABLE OF CONTENTS I. Introduction II. Assumptions About the Aggressiveness Index III. Experimental Design and Data Collection IV. Statistical Analysis V. Confounding Factors VI. Agreement with OtherData VII. Conclusions REFERENCES LIST OF TABLES Table I Table 2 - Paired Before and After Results - Chrysotile Fibers Per Liter from Cooper et al. System X - Comparison of Asbestos Levels in Water Before and After Conveyance Through A/C Pipe > Page 1 2 5 9 10 13 15 16 8 13 CAP CO JEN 0010502 INTRODUCTION The paper, "Chrysotile Asbestos Fibers in Drinking Water from Asbestos-Cement Pipe" by Kanarek et al *is a supplementary analysis of data taken fronrfa larger two part study 2>3>4,5,6 t^at examine(j the presence of asbestos fibers in drinking water and disease incidence in the San Francisco (California) Bay Area. In the paper, the authors question "the validity of aggressiveness indexes for prediction of fiber release from asbestos-cement pipes." Their Arguments are based on a purported "demonstration of substantial increases in asbestos fiber counts in nonaggressive drinking water" flowing through asbestos-cement (A/C) pipe within selected distribution systems. This review will comment on the Aggressiveness Index (AI) as a measure of asbestos release from asbestos-cement pipe, the assumptions and inferences made by the investigators, experimental design, sampling and data analysis techniques, and possible explanations for the study's observations. / -1CAPCO JEN 0010503 tr V* n. ASSUMPTIONS ABOUT THE AGGRESSIVENESS INDEX The underlying assumption of the paper is, in the authors' words, " ... that there is a positive correlation between fiber release from the walls of A/C pipe and the 'aggressiveness' or corrosiveness of water." It is further assumed that this correlation is quantitative, consistent and predictive. Both these assumptions are incorrect. Prior to development of the Aggressiveness Index (AI), manufacturers recommended that the use of A/C pipe in water with a Langlier Index of -2.0 (AI=12.0) or less would be inconsistent with maintaining the 50-100 year design lifetime of water systems. In the early 1970's, engineering consultants requested that American Water Works Association (AWWA) and American Society for Testing and Materials (ASTM) specifications for A/C pipe further define conditions of product use in so-called aggressive and non-aggressive waters.' It should be emphasized that the manufacturers' recommendations and AWWA/ASTM guidelines7 'ft were developed solely for the purpose of ensuring long-term, structural integrity of A/C pressure pipe. The AI was not conceived to be a predictor of asbestos fiber release. As the U.S. Environmental Protection Agency (EPA) Drinking Water Research Division has stated on this matter: The use of the AI as a predictor of the condition of the interior surface of the pipe in contact with the drinking water and its tendency to retain or release asbestos fibers has come about in spite of the fact that the original intention of the index was to outline water conditions that might cause structural failure of the pipe. The difference in application is very significant. . (emphasis in original) -2- CAPCO JEN 0010504 There is a substantial body of scientific evidence supporting the view that there is neither a quantifiable nor a predictive relationship between any corrosion index and levels of asbestos released from A/C pipe into drinking water. In 1980, an EPA researcher testified: Further work has shown that the aggressive index is not the perfect indicator of whether a water will corrode an asbestoscement pipe and release fibers. The calculation of the aggressiveness index alone is not always sufficient to predict actual behavior of A/C pipe.10 And in 1981, EPA again underlined the imprecision of the AI, almost flatly rejecting the Al-fiber release hypothesis: Substantial field and pilot plant data, combined with comprehensive evaluation of many of the important chemical complexation and precipitation reactions occurring in drinking water,' give compelling evidence that the AI has immense shortcomings with regard to use as a predictor of fiber release and interior surficial pipe condition that enjoin against its use under most circumstances.11 The Agency's most recent published work on the Aggressiveness Index states: There are several major avenues by which the AI is open to criticism as an indicator of corrosivity and fiber release, the manner in which has most frequently been used. The EPA paper goes on to detail the deficiencies of the AI in reliably predicting fiber release, concluding: -3- CAPCO JEN 0010505 The Aggressiveness Index as traditionally conceived is not theoretically sound from a chemical standpoint to predict fiber release and degradation of the interior pipe surface in systems that are undersaturated with respect to calcite. Fortunately, it has generally proved to predict falsely pipe deterioration more often than to predict falsely pipe stability in field situations. ** ** Observations of A/C pipe sections from many localities conclusively demonstrate that natural inhibitory factors, such as those mentioned, are commonly the pipe protection mechanism, not calcite saturation. The AI is a technical guideline governing the selection of the proper type of A/C pipe for waters of differing aggressiveness. It is based on empirical observations and physical testing of specimens from A/C pipe systems. Based on published analyses1^, it can be concluded that the AI is comfortably conservative in assuring long-term structural,integrity of the product. In summary then, the basic assumption of the Kanarek et al paper i.e. that there is a valid correlation between the AI and fiber release, simply is incorrect. The AI was never meant to correlate with fiber release from A/C pipe and in fact, extensive studies by the U.S. Environmental Protection Agency demonstrate this rather conclusively. -4CAPCO JEN 0010506 m. EXPERIMENTAL DESIGN AND DATA COLLECTION The study design and collection of data for the paper gave little or no consideration to selecting a sample population (representative distribution systems) in such a way that a valid generalization to a larger population (all distribution systems in the San Francisco area, for example) could be made. The data suggests no purposeful sampling technique to determine whether the A/C V distribution systems were contributing asbestos to water conveyed through them. The authors explain this away by stating the collection of "before" and "after" A/C pipe samples " ... was only one of several factors involved in this overall assessment of asbestos in drinking water supplies." The facts are (depending on which of the original studies are read) that the study of fiber release from A/C pipe was a matter of tertiary importance to the authors. By its own admission, the study is of limited value in its relevance to A/C pipe: Samples were taken before and after lengths of asbestoscement pipe (See Table 18). Although, in some cases there were apparently substantial increases as a result of the water passing through the pipe, the data appear inconsistent. This study did not include the amount of asbestos-cement pipe in the various distribution systems as a variable in the analysis because of its questionable significance and the difficulty of determining its location, age, and dimensions. 4 (emphasis added). An EPA funded report on the asbestos sampling plan for the San Francisco Bay area study has stated: The contribution (if any) of asbestos-cement pipe to the level of asbestos found in drinking water has yet to be determined. The Cooper study was not specifically designed to measure the risk that might be attributed to the use of asbestos-cement pipe. 5 -5- CAPCO JEN 0010507 The harshest criticism of the study's design and data colleetion/analysis comes from the Environmental Protection Agency, which funded the original project: The only suggestion of asbestos cement pipe being attacked by water with an aggressiveness index greater than 12 was in conjunction with an epidemiological study done in the Bay Area of California. The data on the pipe were not gathered specifically for the purpose of studying pipe corrosion but were pulled from the overall study data set. Some information about the systems such as previous tapping procedures and whether the pipe had been deteriorated from som^ other cause were not available. None of the pipe had been dug up for inspection, unfortunately. I did not feel after reviewing the data that I could consider this a complete picture of what is happening in these systems. A more complete study is necessary before I can conclude pipe corrosion occurs with water of an aggressiveness index of greater than 12. (emphasis added). With these facts in mind, any statistics generated from this sampling data base and any conclusions drawn from those statistics, must be considered of dubious value. A closer examination of the samples themselves reveals equally serious compromises in scientific principles. The authors purport to have thirty-nine pairs of "before" and "after" water samples corresponding to "before" and "after" flowing through A/C pipe. There are not thirty-nine paired samples. Firstly, consider a day on which a single "before" sample is taken from a water treatment plant discharge and four "after" samples are taken from residential taps after flowing through varying lengths of A/C pipe. This sampling plan is not generating four independent "before/after" pairs. The four "before" samples are artificial -- they are not independent of each other, they are the same observation because only one sample was .actually taken. In the paper, nineteen of the thirty-nine pairs -6- CAPCO JEN 0010508 of samples were selected in this manner. Table 1 (page 8) lists the original data, indicating which pairs are from identical "before" samples. No attempt is made by the authors to account for this dependence. This is particularly ironic in light of the authors' statement that various taps were sampled "to determine the continuity of values throughout the distribution area." So, although the paper acknowledges the need for a supplementary analysis to account for this dependency, it was never undertaken. Second, consider a given section of A/C pipe sampled' on different days. Each repeated sample contributes information regarding the change in fiber counts within that section. Only to the extent that the given section of pipe is better characterized does each repeated sample contribute information with respect to characterizing the distribution system as a whole. These repeated samples logically should not be allowed to have as much influence in an analysis of the total system as a pair of measurements on different sections of pipe within the system. In the paper, fifteen of the thirty-nine samples are from sections of pipe with repeated measures. Again, no attempt was made by the authors to account for these two different sources of variation in fiber counts. It should be mentioned that these "within-section" repeated measurements are desirable. They supply information concerning the variability of fiber counts over time. This information can be used to establish limits on the uncertainty associated with the expected number of counts for a given sample. Unfortunately, as seen above, the authors do not use the data to this end. -7- CAPCO JEN 0010509 Table 1 -8CAPCO JEN 0010510 1, (A) < The fiber count is below the lowest detectable limit listed. ` One 'before' sample for multiple 'after' samples. RX - Replication of same section of pipe. .* i c[eC<pre /" 'Kanarek, M.S., P.M. Conforti, L.A. Jackson. "Chrysotile :^Asbestos Fibers in Drinking Water'from Asbestos-Cement . Pipe", Environmental Science ancl Technology ,' v. i5, no. 8, p. 923, August 1981. ^ . Cooper,' R.C., et al., "Asbestos in Domestic Water Supplies in Five California Counties", U.S. * i ." -- Environmental Protection Agency Contract No. -- ,R804366-02. . Snedecor, G.W. , 'and W.G. Cochran, Statistical Methods' Sixth Edition Iowa State University Press, Ames, Iowa ' -(1967). '. ... CAP CO JEN 0010511 table 1 PAIRED BEFORE AMD AFTER RESULTS CHRYSOTILE FIBERS PER LITER FROM COOPER ET AL. Pair Number 1 2 3- 4 5 6 7 89 ... 10 ` 11 I?. : 13 . 14 * 15 16 .. 17 18 19 ... ,, , 20 21 . 22 ` 23 24 ` 25 26 . '27 28 29 30 - 31 ' 32 (33 34 (35 ' (36 . 37 38 * 39 . Svstem A A A .. A A' A A A A A A A A A B B B B B B B B B B B B C C C C -- -- C c c c c - - Before A/C Pi <1.0x104 (A) 6.0x104 <2.5x104 <2.0x104 ' '2.0x105 . 1.0x105 . <1.0x104 <2.0x104 i.0xl05 <2.0x104 4.0x105 ' 5.0x104 5.0x104 5.0x104 * ` <5.0x103 <5.0x103 1.6x105 . <2.5x104 <2.5x104 <2.5x104 <2.5x104 ' i.8xl05 ' 7.6x105 7.6x105 . - 2.6x105 2.5x104 5.0x104 5.0x1045.0x104 5.0x104 2.5x104 <2.5x104 2.0x105. -2.0x105 <2.5x104 <2.5x104 <2.5x104 4.0x104 <2.5x104 .After. AtC. Pipe , <2.oxio4 pa vcj , <2.0x104 R1 <i.Oxi'o5 riv-. <2.0x104 R2 * 1.5x105 R2-.- 3.0x105 R2 .. <2.0x104 R3` 1.4x105 R3 * 2.0x106 R3 <2.0xl04`\ /r 1.0x105 - 5.0x104] (B) 3.5x105 5.0x105 .. <7.0xi03' < 2.5x104] 5.4x105 r4 . 1.5x105 r4 1.9xi06`R5 5.4x104. . <2.5x104 ' . 3.8x104 R5 2.0x105 . 4.0x105] 2.0x105 '2.0x105 6.2x106' 3.4x107 2.5x106 <1.0x106 2-.0xi05 3.0x105 6.0x106 6.4x106 2.0x107 1.5x106. 5.4x104 4.0x104 1.6x106 CAP CO JEN 0010512 IV. STATISTICAL ANALYSIS Although the shortcomings of experimental design and data collection are sufficient to invalidate this research, there also exists erroneous data analysis. Assume for discussion purposes that there are indeed thirty-nine pairs of "before/after" samples with the results as reported: nineteen increases, eleven ties, nine decreases. The authors claim a two-sided significance level of 0.039 V from a sign test corrected for continuity. This is not obtained by the standard formula X2 = (/a-b/ -l)2 n = (/24.5 - 14.5/ - l)2 39 = 2.08 where each tie is treated as one-half an increase and one-half a decrease. The significance level of the above chi-square value is 0.150 -- not considered to be a significant result. So, not only does the study fail to employ sound scientific principles in data collection, it incorrectly calculated the significance associated with the most important test statistic. -9CAPCO JEN 0010513 V. CONFOUNDING FACTORS The paper concludes that in comparison to the "before" water in one system (System C) " ... an increase of 115 times the number of fibers" was observed after passage through sections of A/C pipe. The authors go on to hypothesize a number of possible explanations for these apparent increases, including: .v breakup of naturally-occurring fiber after floxv through A/C pipe size of pipe water flow rate alternate wetting and drying of (pipe) surfaces quality of joint materials used corrosive cleaners in use to flush sections of pipe tapping of pipe the possibility that hydrants or "dead ends" could have acted as a "reservoir" for fibers, producing artificially elevated fiber counts. The paper then dismisses out of hand, without any supporting justification or substantiating data, the possibility that any of the above factors accounted for the apparent asbestos releases in System C. There is no explanation as to whether these factors might explain smaller apparent increases in Systems A and B. Conspicuous by its absence in the list of potentially confounding factors is perhaps the most important -- variations in levels of naturally-occurring asbestos in the source waters. Many of the source waters in the study have fiber concentrations -10- CAPCO JEN 0010514 in excess of one million fibers per liter (mfl) and some over 20 mfl. Other investigators have pointed out the ubiquitous nature of asbestos in Bay Area drinking waters: As can be seen from the charts (representative asbestos values in Bay Area water systems), none of the major water districts servicing the San Francisco Bay Area are free of asbestos fibers. Individual asbestos measurements range from 10' fibers per liter to 1.8 x 108 fibers per liter of finished water. 6 To demonstrate the possibility (if not probability)' of significant asbestos variations in source waters, consider the sources for System C. The Hetch Hetchy Aqueduct delivers water directly to the system's customers. The Hetch Hetchy system (0.13 mfl - 0.24 mfl) is fed by the Calaveras Reservoir (110 mfl) and the San Antonio Reservoir (0.46 mfl). Portions of System C are also supplied by three treatment plants (0.03-2.1 mfl) and local runoff and wellfields whose asbestos content was not measured. The well and run off water receives no further treatment which would reduce its asbestos content. Given the natural asbestos content of the different source waters, the variability in asbestos content in the treated waters, and the distinct possibility that intermittent "slugs" of untreated high asbestos content water may have been delivered directly to System C by any number of sources, any asbestos levels observed in that system clearly are subject to wide variation regardless of the pipe material. The inability to control for such important confounding factors clearly renders meaningless any conclusions regarding the role of A/C pipe in contributing asbestos to the drinking water. -11- CAPCO JEN 0010515 Further, the limited number of water samples also raises serious questions about the paper's observations. An EPA funded report on asbestos sampling in the San Francisco Bay Area comments on the problems of such limited sampling, as well as the previously mentioned likelihood of asbestos level variability: The individual asbestos fiber concentrations listed in previous sections of this report (Kanarek et al samples) are the values for a single grab sample of tap water taken from a census tract. Water sampling sites were chosen on the basis of 1) being representative of a specific water distribution system and 2) convenience of sampling. In order to. more fully characterize the asbestos content of these water systems, it may be necessary to sample a small proportion of a very large flow of water. This would minimize the problems of utilizing grab samples and reduce the number of water samples to be analyzed. This would also take into account the intermittent and non-ujiiforrn occurrence of asbestos in water from various sources, (emphasis added). Finally, in the 1950's and 1960's substantial amounts of asbestos caulking rope were used with lead caulked valves and fittings. This rope is in contact with the water and could conveivably contribute asbestos to water in the distribution system. -12- CAPCO JEN 0010516 Vn. AGREEMENT WITH OTHER DATA Interpretation of data should have coherence i.e. the results of a study are suspect if they conflict with findings from other studies. In this regard, the paper is in serious want of corroboration. Nothing in the extensive body of literature produced on the subject of potential fiber release from A/C pipe makes findings comparable to those made in this paper. V The results of an independent study conducted by one of the Bay Area's largest water utilities actually contradict the paper's findings and hypothesis. TABLE 2 COMPARISON OF ASBESTOS LEVELS IN WATER BEFORE AND AFTER CONVEYANCE THROUGH A/C PIPE Sample 1-Treated Water Tap Water1 2 2-Treated Water Tap Waterz Alkalinity (mg/L. CaCO^) 9.38 20 9.38 19 8.09 103 8.08 103 Calcium Hardness (mg/L CaCO^) 8.6 8.6 35.9 38.7 Chrysotile AI Asbestos (f/L.) 12.21 BDL 12.20 BDL 12.14 BDL 12.16 BDL 1 Conveyed through approximately 1700 linear feet A/C pipe. 2 Conveyed through approximately 2000 linear feet A/C pipe. -13CAPCO JEN 0010517 The paper discusses the "unusual" water chemistry of System C and theorizes that the high pH may be unduly and falsely weighting the AI towards nonaggressiveness. The authors also note the low calcium content of the water implicitly suggesting that this may account for the apparent asbestos increases. A comparison of Sample 1 (Table 2) with the mean water quality values for System C follows: Sample Alkalinity (mg/L. CaCO^) Calcium Hardness (mg/L. CaCO?) AI Before After System C Sample 1 9.2 9.38 53.2 20 23.05 8.6 11.93 12.21 12.54 12.20 Table 2 shows that both the alkalinity and the calcium hardness of Sample 1 are considerably lower than that of System C -- a worst case scenario, if you will. Sample 1 demonstrates no asbestos pick up and no increase in AI as the water traverses the A/C pipe system. Although this is a limited study, both these observations dispute the hypothesis that the low calcium content of the water in System C explains the purported asbestos increases. -14- CAPCO JEN 0010518 vn. CONCLUSIONS The paper suffers from numerous weaknesses and errors. Its underlying assumption is fallacious -- there is not, nor was there ever intended" to be a positive correlation between the Aggressiveness Index (AI) and fiber release from A/C pipe. The study's original design, but its own admission, is of limited relevance to A/C pipe. The sampling procedure was extremely limited and did not generate independent "before/after" pairs. The most important test statistic is incorrectly calculated. Important confounding variables are overlooked or ignored -- others are dismissed out of hand without explanation. Finally, the results have not been replicated or corroborated by other, studies and they are inconsistent with more contemporary knowledge on the subject. -15CAPCO JEN 0010519 REFERENCES 1. Kanarek, M. S., Conforti, P. M., and Jackson, L. A., "Chrysotile 'Asbestos Fibers in Drinking Water from Asbestos-Cement Pipe," Environmental Science and Technology, Vol. 15, No. 8, August 1981. 2. Cooper, R. C., Kanarek, M., Murchio, J., Conforti, P., Jackson, L., Callard, R., and Lysmer, D., "Asbestos in Domestic Water Supplies in Five California Counties," Progress Report for Period April 25, 1977 to June 30, 1978. U.S. Environmental Protection Agency Contract No. R 804366-02, EHS Publication No. 78-2, 1978. 3. Kanarek, M. S., "Asbestos in Drinking Water and Cancer Incidence," Ph.D. dissertation, Department of Epidemiology, University of California at Berkeley, 1978. 4. Cooper, R. C., Kanarek, M., Murchio, J., Conforti, P., Jackson, L., Callard, R., and Lysmer, D., "Asbestos in Domestic Water Supplies in Five California Counties, Part n, 1969-1974, Population and Tumor Data Base," U.S. Environmental Protection Agency, Contract No. R 804366-02, EHS Publication No. 79-1, 1979. 5. Kanarek, M. S., Conforti, P. M., Jackson, L., Cooper, R. C., and Murchio, J. C., "Asbestos in Drinking Water and Cancer Incidence in the San Francisco Bay Area." American Journal of Epidemiology, 112:54-72, 1980. -16- CAPCO JEN 0010520 5. Conforti, P. M., Kanarek, M. S., Jackson, L. A., Cooper, R. C., and Murchio, J. C., ''Asbestos in Drinking Water and Cancer Incidence in the San Francisco Bay Area: 1969-1974.'' J. Chron. Pi. 34:211-224, 1981. 7. American Water Works Association Standard for Asbestos Cement Distribution Pipe, 4 In. Through 16 In., For Water and Other Liquids, AWWA C-400-77, 1977. V 8. American Society of Testing and Materials, Standard Method of Testing Asbestos-Cement Pipe, ASTM C-500-76, 1976. 9. Schock, M. R., Logsdon, G. S., and Clark, P. J., "Evaluation and Control of Asbestos-Cement Pipe Corrosion," paper presented at the International Corrosion Forum, National Association of Corrosion Engineers, 1981. 10. Millette, J. R., Direct Testimony at Public Hearing on Proposed Addition to Section 19-13-B102 of the Connecticut Public Health Code, September 16, 1980. 11. Op. Cit. (Schock) 12. Schock, M. R. and Buelow, R. W., "The Behavior of Asbestos-Cement Pipe Under Various Water Quality Conditions: Part 2, Theoretical Considerations," Journal American Water Works Association, December, 1981. -17- CAPCO JEN 0010521 13. Houck, D. H., "Structural Performance of Asbestos-Cement Pipe in Corrosive Potable Water Environments," paper presented at the Internationl Corrosion Forum, National Association of Corrosion Engineers, 1981. 14. Op. Cit. (Kanarek, 1978) 15. Tarter, M. E., and Leong, C. J., "Asbestos Sampling Plan for the San Francisco Bay Area, California," U.S. Environmental Protection Agency, Order No. C3253NAET, July, 1980. 16. Snedecor, G. -W. and Cochran, W. G., Statistical Methods - Sixth Edition, Iowa State University Press, Ames, Iowa, 1967. 17. Op. Cit. (Tarter). -18CAPCO JEN 0010522 -3During this ten year period, the following facts formed the basis of industry's position on the safe manufacture and use of A/C pipe: o asbestos is unique and irreplaceable in the manufacture of A/C pipe o the risk, if any, from occupational exposures to asbestos have been minimized through manufacturer's compliance with OSHA standards and use of industry-recommended work practices for A/C pipe field operations o there is no evidence that current environmental discharges from A/C pipe plants, if they occur at all, result in unreasonable risk o a ban on A/C pipe is not supported by medical/scientific evidence o prohibiting the use of A/C pipe would adversely affect water utilities by reducing competition and increasing costs of piping materials. These same facts are the basis of industry's position today. The A/C pipe industry cannot support and indeed, will strongly oppose EPA's newest theory that A/C pipe may present an unreasonable risk to health and the environment. CAPCO JEN 0010523