Document zz0ow9VwYQGVeBnv5agJdQeDR

PLAINTIFF'S EXHIBIT A3 A/C Pipe Producers Association \ Executive Committee T0 International Affairs Committee ----------. fCzzJ J.F. Welch, Vice President --i. -- Inferno! Correspondence date September 7, 1982 subject u.S. Environmental Protection Agency (EPA) - Epidemiological Study ACTION REQUIRED: Review for information Enclosed is the study, "Cancer Incidence in Relation to Asbestos in Drinking Water in the Puget Sound Region," as published in the August, 1982 issue of American Journal of Epidemiology. Like the ICanarek studies, an eeologio design was used in which the observational units were census tracts. Thus, the findings, however favorable, are subject to the same shortcomings. The study is somewhat less positive than the original *work, "A Study of the Effects of Asbestos in Drinking Water on Cancer Incidence in the Puget Sound Region," by Richard Keith Severson of the University of Washington. In his Master of Science thesis, Severson concluded that imbibed asbestos in the Puget Sound Region was not related to alimentary tract cancer. The enclosure concludes: Results of this study and prior studies of cancer in relation to waterborne asbestos are inconsistent, and provide little evidence that asbestos in community water supplies has altered the risk of any cancer. # # sje Results of the present study confirm none of these findings (Ed. note: relationship between asbestos in water and neoplasms of rectum, stomach, pancreas, lung, peritoneum, esophagus and pleura)...and most of the findings in the various published studies probably represent chance phenomena resulting from large numbers of comparisons. #** From the data we have analysed, v;s conclude that the cancer risk from asbestos in drinking water in the Sultan River area is, at most, very lew, although there is suggestive evidence for some anatomic sites. The sites to which the author refers arc the small bowel and pancreas. The study observed consistent, but not statistically significant, associations between small bowel cancers and asbestos in drinking water. After noting those associations, the investigators give three reasons "...to doubt that the observed relationship is causal." CAPCO JEN 0012282 Polissar also suggests a possible link between pancreatic cancer end waterborne asbestos, referring to five previous studies that showed some association in one or both sexes. But again, he goes back to middle ground and states, "...our results neither confirm nor refute prior findings." As an aside, Staff is not aware of data demonstrating increased incidence of pancreatic cancer in highly exposed occupational groups. In the absence of such data, Polissar's suggestion of a waterborne asbestos-pancreatic cancer link is no more than speculation. Polissar et al also are conducting a case control epidemiological study based on these exposure and cancer incidence/mortality data. A preliminary report will be finished for the EPA Summary Workshop on Ingested Asbestos. Staff Analysis This study is noteworthy in three respects. First, the findings are consistent with all other epidemiological studies (except Kanarek) of cancer and asbestos in drinking water. Second, putting aside the variable of population mobility, the exposure period (more than 50 years) is longer than the latency period for asbestos-related disease. Third, the so-called "signal tumor" for asbestos exposure (mesothelioma) did not even appear in sufficient numbers for analysis. Although the scientific caveats of the Polissar study give it less positive impact than the Meigs or Wigle studies, the industry should be encouraged by EPA's summarization of the findings in the July 28, 1982 Drinking Water Research Progress Report: The descriptive epidemiology study does not show any relationship between asbestos in water and increased cancer risk. Also enclosed is a news clipping from the Seattle Times that includes some very positive quotes about the Polissar study...moreso than the study itself. If you have any questions, please do not hesitate to call. JFW/ccw Enclosure cc: A. Kahn, Esq. AIA AIA/NA W. Smith, M.D. Brian Commins, Ph.D. Tim Hardy, Esq. copies: Executive Committee L. Ambler L. Taylor J. Cran 0172090710 International Affairs Committee E. van der Rest S. Al-Tarkait R. Dorner M. Peleourt L. Giannitrapani J. Cuvelier P. Hart B. Giboin A. Saoulis G. Zaviezo R. Jalan J. Schmaus V. Pattabhi J. Rodrigues C. Barton Dubuc K. Hudson Dimatit t> CAPCO JEN 0012283 American Journal ok Epidemiology ^ Copyright -0 106- by The Johns Hopkins University School of Hygiene and Public Health Ail rights reserved Vol. 116. NV 2 Printii{ in U.S A. CANCER INCIDENCE IN RELATION TO ASBESTOS IN DRINKING WATER IN THE PUGET SOUND REGION LINCOLN POLISSAR.'-5 RICHARD K. SEVERSON.'- EDWIN S. BOATMAN3 and DAVID B. THOMAS'- Polissar, L. (Fred Hutchinson Cancer Research Center, Seattle, V/A 98104), R. K. Severson, E. S. Boatman and D. B. Thomas. Cancer incidence in relation to asbestos in drinking water in the Puget Sound region. Am J Epidemiol 1982;116:314-28. Population-based and proportional odds ratios for various cancers, based on incidence data from 1974-1977 and mortality data from 1955-1975 for west ern Washington St3te, were calculated in relation to three measures of expo sure to asbestos in community water supplies. Six odds ratios were calcu lated for each neoplasm that occurred in sufficient numbers in each sex. About half of the 332 odds ratios calculated were above unity and half were below unity, and no more of them differed significantly from unity at the 5% level than would be expected by chance. Odds ratios for tumors of the small intestine were consistently elevated in both sexes, as were those for neo plasms of the thyroid, eye, testis, and prostate in males; however, odds ratios for brain tumors and leukemia were consistently lass than one in both sexes. Chance istha most likely explanation for these findings. Results of this study and prior studies of cancer in relation to waterborne asbestos are inconsis tent, and provide little evidence that asbestos in community water supplies has altered the risk of any cancer. However, all investigations conducted to date are correlational studies which have an inherently high probability of fail ing to detect actual increased risks associated with imbibed asbestos, and additional studies of individual exposures are warranted. Neoplasms of the pancreas and small intestine should be included in such studies. asbestos; environmental exposure; environmental pollutants; neoplasms; water pollutants; water supply Only within the last 11 years have in vestigators begun to check thoroughly for the presence of asbestos fibers in drinking water. In 1971, Cunningham and Pontef ract (1) found asbestos fibers in tap water Received for publication June 4. 1981, and in final form February 13. 1982. ' Fred Hutchinson Cancer Research Comer. 1124 Columbia St., S-'oitle. \VA 95104 (address reprint requests in Dr. Pilissan. = School of Public Health and Community Medi cine. C. of W.i.-hinqton. Seattle. WA. Supported by Grant No. KS0o84(iQ30 iron: the United States Environmental Protection Agency. The authors wish to thank Elaine E'crid';;: for edi torial assistance, the Health Data Section of the Washington State Dept, of Social and Health Ser vices fur essential mortality data, and-Joy Hogsarth. Put Mueller and Judy Keogh for technical assistance. samples from eight different areas of Canada, including Ottawa. Toronto and Montreal. In 1973. Duluth, Minnesota, public water supplies derived from Lake Superior were found to be contaminated with amphibole asbestos fibers (2'. Since then, asbestos has been found ir. a number of other public water supplies (3 - Despite the known carcinogenic effect of inhaled asbestos, few studies have ex amined the health effects of waterborne asbestos. In 1974. Mason et al. (4> com pared the 1950 - 19fi9 mortality rates *>f 21 neoplasms for Duluth to rates for Hen nepin County (including Minneapolis! and to rates for ail of Minnesota. Rc-ciai cancer showed the strongest and most' 314 consistent reinti sure from water- observed for sic: pancreatic canct pared Duluth's incidence rates fc for Minneapolis and did not con served excess r Duluth. Howeve cancer in males, in both sexes, wa Wigle (6i grou Quebec into thret tos concentration known high, pos; lowconcentrati expected mortali: and 1970-1972 each exposure g: with known high using asbestos-be* 1850. which sugg had passed to sh associations exist cant increased r male stomach car. though these obse s 5 due to occupation 1 An excess of fet: }j rates also was ofc tent with results f was no increase ir fJ1 rates. Harrington et a i J4 dence rates of ca turn, and stomacl i in which water 1ll asbestos-cement them to incident' i other types of pip- i fates in the form ; were observed. h based on truth ipl- posure in the sane 3 ,* additional types o: ; ao consistent as \ somewhat consist I worn Duluth and CAPCO JEN 0012284 Vo I Ilfi. Nn. 2 Pruitt/ /'! c .v.-\. \ drinkint; JiAVIM II. THOMAS'- WA 98104), ce in relation J Epidemiol r.cers, based = 975 for westores of expo. were caicuin each sex. :nd half were ity at the 5% : of the small ose for neo- odds ratios both sexes, of this study -ire inconsis;ter supplies ;onducted to ability of failibestos, and :asms of the neoplasms; \ n'ferent areas of iwa. Toronto and .lulh, Minnesota, .drived from Lake be contaminated s fibers (2). Since een found in a ater supplies (3). arcinogenic effect studies have excts of waterborne on et al. 14) cornmortality rates of a to rates for Hen.r.g Minneapolis) Minnesota. Rectal ongest and most CANCER AND ASBESTOS IN DRINKING WATER 315 consistent relationship to asbestos expo sure from water; lesser associations were observed for stomach cancer and female pancreatic cancer. Levy et al. fo) com pared Duluth's gastrointestinal cancer incidence rates for 1969--1971 with those for Minneapolis and St. Paul combined, and did not confirm the previously ob served excess risk of rectal cancer in Duluth. However, an excess of stomach cancer in males, and of pancreatic cancer in both sexes, was observed in Duluth. Wigle (6) grouped 22 municipalities in Quebec into three classes based on asbes tos concentrations in their water supplies: known high, possible high, and probable low concentrations. The observed-toexpected mortality ratios for 1965-1967 and 1970 -- 1972 were then calculated for each exposure group. The communities with known high concentrations had been using asbestos-bearing water since about 1S-S0, which suggested that enough time had passed to show whether any strong associations existed. Statistically signifi cant increased rates were observed for male stomach cancer and lung cancer, al though these observations may have been due to occupational exposure to asbestos. An excess of female pancreatic cancer rates also was observed, which is consis tent with results for Duluth, but there was no increase in male pancreatic cancer rates. Harrington et al. (7) computed the inci dence rates of cancers of the colon; rec tum, and stomach in Connecticut towns in which water is delivered through asbestos-cement pipes and compared them to incidence rates in towns with other types of pipes. No consistent excess rates in the former type of communities were observed. More refined analyses, based on multiple indices of asbestos ex posure in the same towns, and on rates for additional types of cancer (S), also showed no consistent associations. However, somewhat consistent with observations from Duluth and Quebec, of the many possible associations considered, more positive associations were found for pan creatic cancer in males than fur cancers of any other site. Conforti et al. f9) related 1959-1974 cancer incidence rates in census tracts of the San Francisco-Oakland Bay Area, California, to chrysotiie asbestos concen trations in municipal water supplies. Sta tistically significant associations, or results that showed the same trends in both sexes, were found for cancers of the stomach, esophagus, and pancreas. An association also was found for neoplasms of the pleura and peritoneum in women but not in men. Although the studies conducted to date have yielded inconsistent results, in the aggregate the findings are sufficiently disconcerting to be cause for concern. In each of the four areas in which studies have been conducted (Minnesota, Quebec, Connecticut, and California), some evi dence has been published suggesting an association between asbestos in water and pancreatic cancer in at least one of the sexes; and from all of the areas except Connecticut, there are similar findings for stomach cancer, which is consistent with the results of some studies of indus trial exposure to airborne asbestos (10). Further investigation obviously is war ranted. The Puget Sound region of western Washington presents an excellent setting in which to continue this effort. Its three largest metropolitan areas generally have used the same river sources of water since Table 1 Study areas with high and tow concentrations of asbestos in drinking water. Puget Sound region. Washington Asbestos concert* tration Source of water Major cities Counties High Low Sultan River Everett Snohomish' Cedar River, Seattle, King, Toll River. Tacoma Pierce Green River. Lakewood Wells 4 CAPCO JEN 0012285 CAPCO JEN 0012286 r; a a ctj ri 9 zi 9n--i *90M! 90^: m99\ 0o*nt 9 W9 9 9?l N0 OI 9j Ci tc r--^* 9r ctj 9; C(l OI 9: 9 00 9 i.*2 9C r3 9'Jl 9 9 9 Cl I: i: 3 .= 5 ~ c rr ^A. .9-- CANCEU AN*) ASBESTOS IN DRINKING V/ATEft 317 9 9 C9l 9 I 5j "i ^ 2 i.o oS2 ^ JS, r* iL9OO r=. cT g.* w*-9 y. S 7 - > i 5 <3 jH 1 r* o r hr r9*" --.Ss5P moa,, * 5-*5 ,s"ir c=si".. o ?f S2 c ~ 3 <f ~ = 3 2^* 9w*9Js oi " -. _f; 2P* t9 S93 V=. t--? c zz O .r -^wL. Orrj .oc>` .-n9 - 5. ii4 Sr _? w. .s o h: e " ~ psiso li Oq .1 I 55 o w ^*1i da a tsn* m" < 5g # + +fc w the early part of the century, and each receives water from a different river sys tem. There are large variations in levels of asbestos among waters of the three sys tems. The asbestos derives from geologic features in each river watershed, and the concentrations have probably changed lit tle over the past 60 years. This paper presents results of a correlational study of the relationship of variations in cancer incidence and mortality rates in the Puget Sound region to temporal and spa-' cial variations in exposure to waterborne asbestos. Materials and methods Study population. Cancer incidence and mortality risk ratios were calculated for the population in the Seattle-EverettTacoma metropolitan area served by several water sources with varying levels of asbestos. Only those I960 and 1970 cen sus tracts having the Sultan River, Cedar River, Tolt River, Green River or Lakewood Wells as the source of their drink ing water were included in the analyses. The specific census tract numbers are shown in the Appendix. The census tracts were grouped by level of asbestos concentration, as shown.in table 1. The two geographic areas indi cated in table 1, the Sultan River, which has high asbestos concentration, and the other water sources, which have low as bestos concentrations, are referred to as "Sultan" and "Other" in succeeding ta bles. Cancer incidence. Data on incident can cers occurring between 1974 and 1977 were obtained from the Cancer Surveil lance System, a population-based cancer registry that covers the study area. All cancer sites were coded using the International Classification of Diseases for On cology- classification scheme (11) (table 2). Cancer Mortality. We obtained death certificate information for all residents of the study area who died of cancer as a primary cause of death between 1955 and . j.' ~-j. i:. i [j.. U" . :. ; .! ' ' -; CAPCO JEN 0012287 318 POUSSAR. SEVERSON. BOATMAN ET Al. 1975 and who were ages 35-79 years al the time of death. Causes of death had been coded to the Sixth, Seventh and Eighth Revisions of the International Classification of Diseases (12) by the Washington State Bureau of Vital Statis tics. For each revision, the mortality codes were grouped in the same categories as for cancer incidence. As an example, the groups for the Eighth Revision are shown in table 2. Asbestos concentration. We obtained 95 grab samples of tap water from the study area. Samples were identified only by number, and the source of the water re mained unidentified until all water analyses were completed. The method used for identification and quantification of asbestos fibers was in accordance with the procedure of Chatfield et al. (13). The procedure involved the filtration of water samples through 47 mm diameter, 0.1 pm pore size Nuclepore membrane filters (Nuclepore Co., Pleasanton, CA). The filters were carboncoated and portions were placed on copper electron microscopy grids. The filter ma trix was dissolved by use of the Jaffe wick procedure using chloroform as a solvent. Examination of electron microscopy grids was conducted by use of a JEOL1-' 100S transmission electron microscope (Japanese Electron Optics, Ltd., Medford, MA) operating at 100 kV at a magnifica tion of 21.000X. The identification of as bestos fibers was based on size, shape, and appearance, and on crystal structure de fined by selected area electron diffraction. Control water samples consisting of membrane-filtered distilled water were examined at appropriate intervals, along with a UICC (Union Internationale Contre le Cancer, Canada) standard chrysotile preparation. A number of characteristics were repor ted for each water sample, including chrvsotile and amphibole fiber concentra tion and fiber size. Census variables. The i960 and 1970 US Censuses of Population and Housing (14, 15) provided census tract-level data that included population by age and sex, per cent high school graduates, median family income, per cent married, per cent working in construction and manufactur ing (where asbestos exposure is likely), and per cent with income at least three times above the poverty level. In analyzing the 1955-1975 mortality data, we assumed that the 1960 popula tion counts and characteristics would be close to average figures for 1955--1964 and that the 1970 data would be dose to average for 1935-1975. In analyzing the 1974--1977 incidence data, we estimated midperiod population counts but assumed that other census trace characteristics, such as socioeconomic status, were simi lar to those recorded ir. the 1970 census. The average annual population at risk used in the various analyses is shown in tables 3 and 4. Duration of exposure. In some analyses of data only from the high exposure Sul tan River area, we calculated an esti mated duration of exposure by two meth ods. First, using annually published dty directories (e.g., 16), which are available from the end of the 19th century, we checked the length of residence in the Sultan area city of Everett during the previous 40 years for cancer cases di agnosed among Everett residents in 1974--1977. Over 90 per cent of the cases were listed in one or more of'the direc tories. We checked the directories at five-year intervals preceding diagnosis and divided the cases into long-term and short-term exposure groups: >30 vs. <30 years appearance in the directories. Second, for each of the water district areas receiving the Sultan River water, we obtained information on the year water distribution from the Sultan River began and imputed an indirect duration of exposure based on these data. We used exposure groups of >30 vs. <30 years of Sultan River water supply. I ! I i ) | .t mi 1 CAPCO JEN 0012288 and Housing ct-level data agu and sex, .ites, median vied, per cent manufacturre is likely), t least three t1i. 75 mortality i960 populaics would be 1955-1964 .d be close to nalyzing the ve estimated but assumed iracteristics, :, were simi:970 census, ation at risk is shown in .me analyses xposure Sul.ted an estiby two methublished city re available ontury, we euce in the during the or cases di ssidents in of the cases of the direcrectories at g diagnosis ng-term and 2=30 vs. <30 ctories. ater district River water, on the year Sultan River -ect duration .ita. We used <30 years of CANCER AND ASBESTOS IN DRINKING WATER 319 Table 3 dumber ofcancers and population at risk for concur from waterborne asbestos used for unolyses in table 7, motes, Puget Sound region, Washington, 1955-1977 Site Sultan va. all ctheri Incidence* 11974-1977). Mortality* (1955-1975/ Sultan Other Sultan Ocher Buccal cavity and pharynx 42 399 Esophagus 7 105 Stomach Small intestine 32 .254 2 16 Colon 87 792 Rectum 55 445 Liver 5 64 Gallbladder 4 43 Pancreas 27 237 Retrop-ritoneum 08 Larynx . 20 1SS Respiratory system Bones and joints 212 1792 2 21 Soft tissue 10 50 Melanoma 24 180 Prostate 202 1566 Testis 19 105 Male genital, residual 8 31 Bladder 66 667 Kidney 18 171 Eye and orbit 4 10 Brain, CNS1 16 164 Thyroid 7 51 Hodgkin's disease 14 82 Non-Hodgkin's lymphoma 2S 234 Multiple myeloma Leukemia 25 93 20 253 All sites 1003 8420 Average annual population at risk 77.462 556.077 33 472 23 326 68 910 6 33 90 1068 37 473 16 269 10 106 50 S02 2 31 9 193 294 1273 7 49 2 67 9 109 69 SOI 4 31 1 16 37 437 29 318 2 18 26 443 4 25 10 176 31 390 11 181 37 493 950 13030 22,487 221.844 Sultan older vs. newer water districts: Incidence^ (1974-15771 Older districts Newer districts Sultan long-term vs. short-tem . Everett Incidence? (1974-1977) Long- Short* term term 30 14 9 7 5 410 24 11 8 4 2 020 58 44 2S 9 40 IS 9 S 5 001 3 100 19 13 7 2 0 000 18 5 6 2 152 S4 50 40 2 000 9 221 17 S i 5 147 70 58 22 16 5 2 1 8 011 48 22 10 6 12 8 3 3 1 410 11 10 1 3 4 310 7 610 21 S 2 3 IS 7 3 4 .12 9 4 5 723 374 220 133 - 56,157 35.717 c * Table 7, cols. 1 and 3. t Table 7, cols. 2 and 4. t Table 7, col. 5. Table 7, coL 6. 1CNS. central nervous system. I Analysis is not population-based. Statistical analysis. We calculated six different pooled odds ratios for each sex and site combination, using a case-control approach and the method, of Mantel and Haenszel (17) for pooling. The six odds ratios differed in the type of data used (in cidence or mortality), the definition of case and control groups, and the defini tion of exposure. These definitions are presented in table 5. For each odds ratio a specific cancer, such as stomach, was used as the case group. For three of the six pooled odds ratios, all- persons in the gen eral population constituted the control - i : ; > - :- CAPCO JEN 0012289 t A*'vMfr.W/5 .sfAfe . .* -4 'i ..v .-., ?*>*>.A.-f *-V-. vr.x**,'. :^^gl. - -.ZA&km:.'.. a&gr&s}.j; SSllI:; [jSrK'fii; -'' oK-V-Si ; 0*M':PHif ='>jrv<W?^ty * ; T`. .-jr'.fcCKSyS i *, fWvVtf.s.-'vf* *?tS?%M<c&5.s53 320 POL1S.SAK, SEVERSON. BOATMAN ET AL. Tadlk 4 fiutnher of cancers and population at risk for cancer from tcuterborr.e asbestos used for analyses in table 7, females, Paget S'Otnd region, Washington. 1355-1077 Suhan older vs. newer Site Sultan all others water districts: Incidence* lncid "P.Ce* Mor:a 0974-1977} (1974.-1977) 0955- 1975) Older Newer Sultan Other Sultan Other districts districts Buccal cavity and pharynx 27 Esophagus 6 Stomach Small intestine 11 2 Colon 101 Rectum 33 Liver 3 Gallbladder 5 Pancreas 30 Retroperitoneum 2 Larynx 7 Respiratory- system 62 Bones and joints 1 Soft tissue 7 Melanoma 26 Breast 245 Cervix 281 Corpus Uterus, NOS- 125 2 Ovary 43 Female genital, residual 14 Bladder 22 Kidney 6 Eve and orbit 6 Brain, CNS= 13 Thyroid S Hodgkin's lymphoma 10 Non-Hodgkin's lymphoma 24 Multiple myeloma 4 Leukemia 14 All sites 1201 Average annual population at risk 7S.637 219 45 169 13 943 399 32 65 205 18 40 734 17 60 194 2631 1653 1328 10 3S6 166 255 84 16 140 118 62 227 110 201 10910 588,791 9 6 30 4 94 30 13 14 39 3 2 53 3 2 3 19S 4S 21 12 64 11 14 16 1 17 3 3 27 23 24 SIS 22.7S5 199 115 464 27 1138 2S1 190 170 523 41 41 926 40 46 S2 250S 647 259 160 913 61 142 153 19 295 41 110 324 142 341 10811 236,363 19 6 8 2 72 29 3 4 25 1 4 63 1 6 22 179 207 92 1 33 7 16 3 T4 8 7 S 19 4 10 915 57.33S 10 0 8 0 41 4 0 9 8 1 4 31 0 1 7 91 120 44 2 16 9 9 4 2 7 5 2 8 0 7 469 34.912 * Tabfe 7, cols. 1 and 3. * Table 7, cols. 2 and 4. t Table 7, col. 5. Table 7, col. 6. 5 NOS, not otherwise specified; CNS, central nervous system, t Analysis is not population-based. Sultan longterm vs. shorMcrm * Everett residence: Incidence? (1974-19771 Long- Short term term 43 21 52 10 24 18 S6 00 01 74 00 21 17 14 00 2"0 03 56 31 7 20 23 24 10 10 12 14 32 24 20 13 00 00 3 21 10 194 171 c group (population-based odds ratios), and, for the other three odds ratios, all other cancers besides the designated site were the control group (proportional-based odds ratios). The population-based and proportionalbased odds ratios are calculated in exactly the same manner once the stratified 2 x 2 tables (case or control vs. dichotomous exposure) are set up. ( | i CAPCO JEN 0012290 <i ~ in Cable 7, Sultan ng-term ;h,*ft-term Everett .s.idence: .cidencel 74 -1077) * Short term 3 1 2 0 13 6 0 1 4 0 1 14 0 0 3 31 20 24 0 12 4 2 4 0 3 0 0 4 At 0 271 .'rtional-i exactly Tied 2 * .ntomous < i il :f 4 I { i i i 11I t i 24 r CANCER AND ASBESTOS IN DRINKING WATER Table 5 Definitions for odds ratio calculations used in table 7 Typn of Data Definition controls* Definition of exp^ure groups High Low Incidence Mortality Incidence Mortality Incidence Incidence General population General population Other cancer sites Other cancer sites General population Sultan River area Sultan P.iverarea Sultan River area Sultan River area Sultan River area, older water districts Other cancer sites Sultan River area, long-term Everett residents All other areas All other areas All other areas All other areas Sultan River area, newer water districts Sultan River area, short-term Everett residents * Case-specific cancer site. 321 Column in taM* 7 id " (2.i (31 (4) fa) (6) -- . - rC .*:> ' *," v? The population-based odds ratios have the advantage of describing the cancer risk in the entire population studied, but also have the disadvantage of being sub ject to estimation errors in the size of the population at risk. Our study includes es timated intercensal populations, and er rors in these estimates can bias the odds -ratios. Proportional-based odds ratios are free of this kind of bias because they represent risk for a given cancer relative to the risk for all other cancers combined and do not require population estimates-. The proportional-based odds ratios express a concept similar to standardized propor tional mortality ratios. The two statistics involve different but roughly parallel cal culations. A disadvantage of the propor tional-based odds ratio is that cancer for many sites or for all sites combined may be linked to the exposure of interest, and. if so, the odds ratios for one site relative to other sites will be lower than the cor responding population-based odds ratio. In the present study, potential errors in the population estimation may bias the population-based odds ratios while a mul ti-site effect of asbestos may bias the proportional-based odds ratios. There is some evidence that our population esti mates are in error since the populationbased odds ratios for most cancers tend to be either generally higher or generally lower than the corresponding propor tional-based odds ratios, though the dif ferences are small. Due to the systematic difference, it seems appropriate to present both types of odds ratios. All cases and controls were, stratified into a series of 2 x 2 tables (case-control status vs. dichotomous exposure). Each stratum was defined by a combination of specific levels of age and potentially con founding variables. The stratifying vari ables we used were age (four to five lev els), and, as other variables which were census tract averages: per cent high Table 6 Asbestos concentration and length ofasbestos fibers in drinking water, Puget Sound region, Washington, 1978--1979 Tap water source No. of samples Sultan River Other areas 22 73 Chrysntile Concentration (10* fibersliter) Mean Standard deviation Median Minimum Maximum 206.5 7.3 162.2 12.4 142.8 2.0 37.2 0.0 556.0 77.6 ?c < l^m 55.5 82.5 % < 5 fim 99.9 99.4 - k'-M `jkm i - S3 : rig. :m `jrd 4 CAPCO JEN 0012291 CAPCO JEN 0012292 . ... * i '/) --o dd 7 j '1 --O^1 O--O* C--O d _94 . j ' . 11 ------ -'X0r -0a0m .. 5 y1j dKOCOOo~SM)H3 tdoot'dOot'e- i% ? 1f 1 *. 41 4) >\x . .... ? x t~) to n ~ d - c*i -* 1 I I i1I + 1111 till! . | I J.l~.------- ; iii iii r/3 O 2 - x s i >> a - J; i! 5 ayyDO CANCER ASD ASBESTOS IS DtilSKISC WATER 323 ?5 i 1 1 5) fl , -- o^ ,1 T` v ^ M- - Ji o i > | Oa Mr o' o-- ` ccoc ac ra Mr---7 wo * i 1 O^ (^--3 C^* cd ?* o' Oca ` 0--5 OM --d X 3 CO ^ 1 | aCO oCO ctOi odd r- d --o o' ora ^ ora oca a> 1 1 U---i). 3o) oM - o NMOt r--ti Xcj*y. o 4 00 ^o 1 I * -- So CO oos o^ c*- h-o oo{ nei on >a oo *?oi ^ wm ^''r W 13 T ^MO * do -*oo-- *~o TCoaOH^.p-eo^ni,';-} J rHt CHO OO^MN NCOl <N3 O^^d^O *2 W os 53 = cC--*4j*O-eC4)McOodC0daCcc-'3cidO^C-H*0'-ta0oOo<tn0"*Ciod0 oW 2 t- o 1/5 CO { GO M 0^0 0 0^0-^0000 -'co-oNKjTrrine.oa o CHOHCOeOioOC'dJOrjCdO-CHO-HHOeOidO ---4' c 2 -S en 5 O -j w g | 2 o O o . w&-JS .*c"* o . e ` n u - c. O *j l8 c s-Scs n *P2 Oc 05 a OiJ? co ^- ^a g~ o o c .O "e Oo II . " U t- *cj *e o2 ** 1t5a s -2 a M to .3 I|c *55 -S S c 2cn .2 J-x,2*o S 3 = a= ^ I S 3 V ? Z 53 3 =. cu t- s e 8 ~ ZL Oc uoo o-* % 2 a 3 odSSS VV i i <a7 .+2+ 42- school graduates as a socioeconomic indi cator (two levels), per cent of the pop ulation in industries possibly associated with asbestos exposure (two levels), median family income (two levels), and per cent married (two levels). The mor tality data covered 22 years, hence we also stratified these cases and controls by calendar year (two to four levels). For pooled odds ratios based on length of residence in the Sultan area city of Everett, we used the per cent of census tract population with income at least three times above poverty level as a stratifying variable in place of the median family income. Also, due to the smaller number of cases available for this analysis, the per cent married was not used for stratification. We did not use census tract population density in the analysis. Only urban areas were included in the study, hence popula tion density of tracts would not be an in dicator of urban vs. rural lifestyle or ex posure. Results Asbestos concentration. Table 6 shows some characteristics of the tap water from the Sultan River and other areas. Only one amphibole fiber was found in all of the water samples, hence exposure was characterized by chrysotile fiber concen trations alone. Mean chrysotile fiber con centration in the Sultan River area differs from that in the other areas by a factor of more than .25. We tested to see if the asbestos expo sure varied geographically within the Sultan River distribution area. We found that there was no significant clustering of exposures, which allowed us to treat the population in the Sultan River area as re ceiving a uniform exposure to asbestos. Cancer risk. A summary of numbers of cases and populations at risk is shown in tables 3 and 4 for males and females, re spectively. Table 7 shows the pooled odds ratios of cancer risk by sex and site. For each cancer site, the population-based . lit# * *** iSg ' j-Si . t f-*?r :5H l-.vsi ' * ?<y CAPCO JEN 0012293 324 rOUSSAK. SEVERSON. BOATMAN ET AL. and proportional incidence ratios are based on the same cases, as are the population-based and proportional mor tality ratios. Therefore, the odds ratios in columns one and three, and those in col umns two and four, of table 7 are not to tally independent. For most cancer sites, the odds ratios in the methodologically re lated columns have about the same mag nitude, which provides some grounds for confidence in the validity of the results. Since a large number of comparisons are being made, it is likely that some of the results are significant by chance alone. Altogether, 332 odds ratios are presented, not counting those for '"all sites" and for sites where the pooled odds ratio was indeterminate due to insuffi cient cases or controls. The 332 odds ratios are about evenly split above and below unity, suggesting a chance mechanism for most of the results. Among 332 indepen dent odds ratios, 16.6 would be expected to differ significantly from unity at the conventional 5 per cent significance level. Thirty of the 332 odds ratios actually dif fered from unity at that level, which is not an excessive number because the odds ratios based on incidence and propor tional incidence, and those based on mor tality and proportional mortality, are not independent (same numerators). Fur thermore, only six of the 30 "significant" odds ratios were greater than one, and there is no known biologic reason to ex pect that the 24 that were less than unity represent a protective effect of asbestos. Consistent results among the various analyses and populations shown in table 7 may be more likely to represent real al terations in risk than single statistically significant odds ratios. There appears to be a consistent, or nearly consistent, posi tive association with asbestos exposure in all analyses for neoplasms of the small in testine in both males and females. On the other hand, cancers of the brain and cen tral nervous system and leukemia show a reduced risk in both sexes in the high- Table 3 Mcnleldiaenseel punted, incidence odds radon for exposure to Sultan fhuer drinking water vs. all other source^ of water for census tracts with tower migration rates, Puget Sound region, Washington, 1974 -- 1577 Site Buccal cavity anti pharynx Esophagus Stomach Small intestine Colon Rectum Liver Gallbladder Pancreas Retroperitoneum Larynx Respiratory system Bones and joints Soft tissue Melanoma Breast Cervix Corpus Uterus. NOS Ovary Female genital, residual Prostate Testis Male genital, residual Bladder Kidney Eye and orbit Brain, CNS Thyroid Hodgkin's lymphoma Non-Hodgkin's lymphoma Multiple myeloma Leukemia All sites' * p < 0.05. * p < 0.01. Males Females Total Odds Total Odds cases ratio cases ratio 106 0.S2 53 0.26 133 0.98 40 307 1.17 213 0.97 30 0.42 20 1.53 1128 0.97 0 33 0.53 60S MO 1$ O.So 34 3.4i- 91 1.05 -- ___ ---- -- ___ ___ -- ---- ---- S23 1.43* 51 1.16 15 4.53" 332 0.90 97 0.96 80 79 0.91 22. 3.70 31 1.78 131 O.SO 55 1.55 141 0.S7 111 23 S2 5 442 193 IS 23 101 5 19 334 5 35 50 1316 S34 720 6 191 75 -- -- -- 130 47 11 67 49 35 107 55 99 1.25 0.70 0.35 0 1.23 1.03 0 0 1.25 0 1.50 1.17 5.59 2.51 1.05 0.91 1.04 0.94 5.17 1.41 0.75 -- -- -- 1.11 0.81 7.750.37 0.23 1.17 0.45 0.50 0.63 1290 1.14- 5477 0.99 exposure area. In men, odds ratios tend to be consistently elevated for cancers of the prostate, testis, eye and orbit, and thyroid. Although consistent in direction, most of the odds ratios for these sites are not statistically significant. To determine the possible effect of mi gration on our results, we calculated population-based odds ratios for cancer incidence using the census tracts with lower migration rates as recorded in the 1970 census. This analysis, which in cluded about half the cases in the Sultan River distribution area, used tracts with at least 73 per cent of the population over 1 1 i i ! i i I i | CAPCO JEN 0012294 >" ratio* for *r f:. all other :*h lower Females Odds ca*es raiict ui 1.25 23 0.70 82 0.35 50 442 M3 193 1.08 IS 0 33 0 301 1.25 50 19 1.50 3S4 1.17 3 5 S3 ' 35 1.5! 90 1.05 1316 0.31 334 1.04 720 0.94 6 5.17 131 1.41 75 0.75 ---- ' --- -- 130 l.II 47 0.S1 11 7.75* 67 037 49 0.28 3S 1.17 101 045 55 0.50 23 O.oS 5477 0.39 1 ; * X i c f * ! ; %r -- * i ? i i i i 1 i i tins tend to. :cers of the rbit, and I direction. -u sites are (feet of micalculated for cancer :racts with :t:ed in the which inthc Sultan tracts with nation over J CANCER AND ASBESTOS IN DRINKING WATER 325 Tahiti 9 Summary ofstudies ofcanter risk in relation to asbestos in water supply by site ofneoplasm' First author (reference no.) Colon Site of neoplasm Kuctum Stomach Pancreas Lung Perito neum Eh>pK**kus Duluth Mason (4) 00 MF MF Levy (5) 00 00 MO Quebec Wigle (6) 00 00 M0 Connecticut Harrington (7) 00 00 00 Meigs (Sj 00 00 00 California Conforti (9) 00 00 MF Present study 00 00 00 OF MF OF MO 00 B 00 MF 00 00 00 OF 00 00 MF 00 Area First author (reference no.) Site of neoplasm Gall. bladder Pleura Small intestine Brain Leukemia Thy roid Eye Testis tate Duluth Quebec Connecticut California Present study Mason (4) Levy (5) Wigle (6) Harrington (7) Meigs (8) Conforti (9) 00 00 - - - 00 00 - - - - -- - 00 - - - - - 00 00 - - -- -- - -- OF 00 00 -- 00 00 - MF MF* MF* M0 M0 - - M - - 00 M * M, association in males; F, association in females; B, association in both sexes combined; 0, r.o associa tion; not studied. * Inverse association with asbestos levels. .age five years living in the same house both in 1965 and 1970. Thus, a subset of the incident cases and population used to calculate the odds ratios in column one of table 7 was used. Results are shown in table 8. There were no cases of small in testine cancers in the low-migration cen sus tracts with high levels of asbestos, so the odds ratios were reduced to zero. This also occurred for cancer of the eye and orbit in males. Odds ratios for leukemia, and for cancers of the prostate and male thyroid, .were not altered appreciably. Of the 59 sex-specific odds ratios in table S, 27 were larger than those in column one of table 7, and 29.5 would have been ex pected to be so due to chance. Except to help rule out associations between asbes tos exposure and neoplasms of the small intestine, and of the eye and orbit, the estimated odds ratios in low-migration tracts do not materially alter the conclu sions based on all tracts. Discussion' Table 9 presents a summary of results from this and previously published studies (4--9). Our interpretations of the data from these studies are somewhat subjective and are not always the au thors'. Of the seven types of neoplasms possi bly related to asbestos in drinking water in this study, the associations were- most consistently observed for cancers of the small bowel--the only neoplasms among the seven types that arise from tissue that comes in direct contact with imbibed as bestos. However, there are three reasons to doubt that the observed relationship is causal; 1) the increased risk was not con fined to a single histologic type of small bowel tumor; 2) the association was not observed when analyses were confined to census tracts with relatively low migra tion rates; and 3) this relationship was `V- .. i-i'i. # y'r< CAPCO JEN 0012295 326 POI.JSSAR. SEVERSON. BOATMAN ET AL.' not found in the two previous studies (5, 9) that examined tiiis site. Nevertheless, due to the consistency of our findings, ad ditional studies of the relationship of neo plasms of the small bowel to asbestos ex posure may be warranted. Risks of cancers of the brain and central nervous system, and of leukemia, were inversely related to levels of asbestos in water. These probably represent chance observations. There is no known reason to expect asbestos to exert a protective effect against these neoplasms, and two previ ous studies (6, 9) found neither to be re lated to waterborne asbestos. The observed increase in risks of tumors of the eye and thyroid also are most likely spurious findings. The odds ratios were elevated only in males; those for ocular tumors were not elevated when analyses were confined to census tracts with relatively low migration rates, and the observed association for thyroid tumors is not consistent with previous re sults from California (9). Risk of testicular tumors has not previ ously been studied in relation to asbestos in water. Because the histologic types of testicular tumors did not differ in areas with relatively high and low water asbes tos levels, this association also probably has no biologic significance. An increased risk of prostate cancer was found in this study, as well as in Wigle (6), one of the two previous studies that included this site. The distribution of our cases by stage of disease at diagnosis was very similar in the high and low as bestos exposure areas, so the finding ap pears not to be an artifact due to a greater rate of urologic procedures (which tend to reveal small tumors) in the high exposure area. As shown in table 9, one or more previ ous studies have shown some relationship to asbestos in water for neoplasms of the following sites: rectum (4), stomach (4-6, 9), pancreas (4-6, 8, 9), lung (6), peri toneum (9), esophagus (9) and pleura (9). Results of the present study confirm none of these findings (there were too few pleural and peritoneal tumors for anal ysis), and most of the findings in the vari ous published studies probably represent chance phenomena resulting from large numbers of comparisons. An exception to this might be pancre atic cancer. All of the five previous studies that included this neoplasm '4-6, 8, 9) showed some association between water asbestos levels and risk in one or both sexes. In the present study, results of the various analyses presented in tables 7 and 8 are not consistent for pancreatic cancer. Some odds ratios are above and some are below unity, and our results neither confirm nor refute prior findings. Because there are so few leads as to the cause of pancreatic cancer, this possible association with imbibed asbestos ob served in five out of six studies that in cluded the site should be the subject of a more intensive study. In addition to chance, the discrepant re sults in table 9 may, in part, be due to differences in characteristics of asbestos exposure in the various study popula tions. These differences are summarized in table 10. The paucity of positive find ings from Connecticut (7, 8) could be due both to the low concentration of asbestos in the water and the relatively shore du ration of exposure in some areas of the state (IS). Short duration of exposure also could be an explanation for some of the nonassociations in Duluth (4, 5), as could asbestos type (amphibole), which is dif ferent from the type of asbestos in other study areas (chrysotile). In addition, the concentration of other possibly car cinogenic contaminants of water, as well as asbestos fiber length, may vary among the various studies cited. Another limitation of all of the studies to date is that specific asbestos exposures are imputed to the population in an entire geographic region. There may be a con founding effect from other factors that CAPCO JEN 0012296 r-cudy confirm were too few mors for ami. igs in the variably represent ng from large ~ht be pancrefive previous <i.oplasrn (4-6, ation between risk in one or tody, results of ;ted in tables 7 for pancreatic are above and id our results prior findings, .'eads as to the \ this possible asbestos ob: tidies that inhe subject of a ' discrepant remrt, be due to ics of asbestos study popula te summarized ~ positive find- 1 be due asbestos short du' areas of the "exposure also >r some of the 4, 5), as could which is difestos in other i addition, the possibly carwater, as well uy vary among . of the studies stas exposures on in an entire nay be a conr factors that CANCF.Il ANI) ASBESTOS IN DRINKING WATER 327 Tabix 10 Chr.rr.clurislics of asbestos in drinking tenter in tenons study populations - Study population (reference no./ Duluth (4,5) Quebec (61 Connecticut (7, 8. 17i California (91 Present study * BDL, below detectable limits. Type of asbestos Amphibole Chrysotile Chrysotile Chrysotile Chrysotile tenge of no. of fiberjrfiter 1-30 x 10' 1.1-1300 x JO' BDL* -0.7 x 10' 0.02-35 x 10' 7-207 x 10' Duraliun of community exposure (years) <20 > 50 > 15 > 50 >'50 vary geographically, and there may be misclassification of exposures due to re cent in-migration into the study area (19). Exposure to chrysotile asbestos is of long duration (>50 years) in the present study, and in Quebec (6) and California (9). These three study areas differ in that the concentrations of asbestos are lowest in California, intermediate in the present study area, and highest in Quebec. How ever, the associations summarized in table 9 do not appear to van' in accor dance with the differences in asbestos concentration among the three study areas, which suggests that the observed associations do not have a biologic basis. Prom the data we have analyzed, we conclude that the cancer risk from asbes tos in drinking water in the Sultan River area is, at most, very low, although there is suggestive evidence for some anatomic sites. AH of the studies to date share an un known but probably substantial misclassification of exposures, which causes a de crease in the power to detect excess risks. To remedy this shortcoming, we are cur rently conducting a case-control study of "cancer risk in relation to imbibed asbestos in which exposures will be determined from in-person interviews. We anticipate that this technique will greatly reduce the amount of exposure misclassification and will allow a more accurate determi nation of cancer risk from waterborne asbestos. RereRE.vces 1. Cunningham HM, Pontefract R. Asbestos fibres in beverages and drinking water. Nature 1971;232:332-3. 2. Cook PM, Glass GE. Tucker JH. Asbestiform amphibole minerals: detection and measure ment of high concentrations in municipal water supplies. Science 1974:155:853-5. 3. Millette JR, Clark PJ, Pansir.g MF, et al. Con centration and size of asbestos in water sup plies. Environ Health Perspect 1980:34:13-25. 4. Mason TJ, McKay R\V, Miller R1V. Asbestos like fibers in Duluth water supply: relation to cancer mortality. JAMA 1974;225:1019-20. 5. Levy BS, Sigurdson E, Mandel J, et al. Investi gation of possible effects of asbestos in city water: surveillance of gastrointestinal cancer incidence in Duluth, Minnesota. Am J Epidemiol 1976:103:362-81 6. Wigle DT. Cancer mortality in relation to asbes tos in municipal water supplies. Arch Environ Health 1977;32:185-90. 7. Harrington JM. Craun GF, Meigs JW, et al. An investigation of the use of asbestos cement pipe for public watersupply and the incidence ofgas trointestinal cancer in Connecticut, 1935-1973. Am J Epidemiol 197S;107:96-103. 8. Meigs JW, Walter SD. Heston JF, et al. Asbes tos cement pipe and cancer in Connecticut 1955-1974. J Environ Health 19S0;42:1S7-19I. 9. Conforti PM, Kanarek MS. Jackson LA, et at. Asbestos in drinking water and cancer inci dence in the San Francisco bay area: 1969-1974. J Chronic Dis 1981:34:211-24. 10. Miller AB. Asbestos fibre dust and gastro intestinal malignancies. Review of literature with regard to cause/effect relationship. J Chronic Dis 1978:31:22-3. 11. World Health Organization. International clas sification of diseases for oncology. Geneva: World Health Organization 1976. 12. US Department of Health, Education and Wel fare. International classification ofdiseases. Sth ed. Washington, DC: 196S iPHS publication no. 16931. 13. Chatfield EJ, Glass RW, Dillon MJ. Preparation of water samples for asbestos fibers counting by tv: CAPCO JEN 0012297 328 POUSSAK. SEVERSON, BOATMAN ET At. electron microscope. El'A-ORD Research Re port, f"EPA G00-4-7S-O11. 1978. Springfield, VA: National Technical Information Service, 1976. 14. US Bur-nu of the Census. U.S. census of popula tion and housing. I960, PHCil 1-142, census tracts in Senttle-Everelt, Washington, PHCilJ153. census tracts in Tacoma, Washington. Washington. DC: 1962. 15. US Eureau of the Census. Census of population ar.d housing: 1970, census tracts, final reports PHCUM95, Seattle-Everett, Washington SMSA, PfiC(l)-2l0, Tacoma, Washington SMSA. Wasington, DC: 1972. 16. RL Polk and Co. 1930 Everett City directory. Kansas City: 1980. 17. Mantel N, Haensccl W. Statistical aspects of the analysis of data from retrospective studies of disease. JNC1 1959:22:719-48. IS. Craun C. Miilette J, V.'oodhui! R. el al. The Connecticut study. Proceedings of the .American Waterworks Association 97th Annual Confer ence. 1977. Anaheim, CA. Denver. CO: Ameri can Waterworks Association, 1977. 19. Polissar L. The effect of migration on compari son of disease rates in geographic studies in the United States. Am J Epidegiioi 1950:111: 175-82. Appendix Census tracts included in study ofcancer and waterborne asbestos, by year and county. Puget Sound region, Washington Year County Census tracts 1970 1960 King Snohomish Pierce King Snohomish Pierce 1-121, 201-22S, 230-250, 260-276, 273-291, 293, 300, 323 401-415, 417-420, 501-520* 601-635, 718.01-722, 723.01*, 723.02T. 735 Al-5, Bl-6, BUI-3, Cl--1, Dl-12. El-4. F1A-2, Gl-S. Hi-3, 11-3, Jl-3, Kl-5, KC4, KC11, KC23. ICC31-112. Kll, Ll-5. Ml-5, Nl-4, NP109, 01-4B, Pl-3, Ql-3, RlA-53. 51A-3, Tl-23, TU41, Ul-2 EMI, EV2-12, LYl, MT1, SC113-119* 1-35. PC1S-22. PC23*. PC35 * 1970 tracts 505-509 and 1960 tract SC11S received Tolt or Cedar River water (low asbestos concentra tion). t Included in mortality odds ratios only. 'ifaftsi I1S1IPS CAPCO JEN 0012298 jL, ' ft *J f.S rSf**'*f!?y>. ,3 t, aiz=i) 3 jSaUJrt' a js.^a $ T*jp? 5^ 1 yi It means Seattleites "can and drink all they want.*' That was the reaction of John Courchene, Seattle's .director of water quality, to a study showing local water does not contain dan gerous levels of asbestos as had. been feared. Investigators .for the Fred Hutchinson Cancer Research Cen ter have announced that, a fouryear study concluded "asbestos in drinking water has not led to an increased cancer incidence or., death rate for people living in the Puget Sound region." The study of the water was commissioned by the Environmen tal Protection Agency after natu rally occurring asbestos, was found in the Sultan- River, which supplies water to Everett and other South Snohomish communities, and the Toll River, which supplies a large part of Seattle's water. Dr. David B. Thomas, an asso ciate member cf the Hutchinson - Center and a professor of epide miology at the University of Wash ington. said the study compared cancer rates ir. people served by the Sultan River,-which' has the highest asbestos content, with rates in others ir. the Puget Sound region who have fewer or no traces of asltescos in their drinking water. "V/e didn'c find any significant differences." he said. " Nearly 45,000 cancer cases ir. the Seauln-Tacoma-Everett area .were studied during the research, ' headed by Lincoln Polissar, an assistant member of the Hutchin son Center and a research associ ate professor of biostatistics at the UW. The study's results have little bearing on a $30 million waterfiltration project under construc tion by Everett whose water comes from the Sultan River. Clair Olivers, an Everett utility engineer supervising the flitretion project, said it was turbidity -- suspended panicles of clay that promote bacterial growth ir. the water -- not asbestos that, created the need for filtering. Although the study does not resolve the argument about how much asbestos is safe. Thomas said, the doses are sufficiently low not to create a measurable health hazard; The study ends 7i/2 years cf concern about asbestos in the water. Courchene said. "This is the best piece of good r.ews we've had. in a long time." S', a-dte to Ui UJ <*> CO cogg Ec*22 cv co cn CAPCO JEN 0012299