Document b59QNkNYEzBrRoQ1NkQ8bEOJk

American Journal of Epidemiology Copyright 1997 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Analysis of Current Trends in United States Mesothelioma Incidence Vol. 145, No. 3 Printed in U.S.A interest (fat, | terest, or the | Bertram Price Mesothelioma incidence often is interpreted as an index of past exposure to airborne asbestos. The mesothelioma rate for US males exhibits an increasing trend throughout the 1970s and early 1980s. The trend has been attributed to occupational exposure in the shipbuilding industry during World War II, in manufactaring, and in building construction. Incidence data (1973-1992) from the Surveillance, Epidemiology, and End Results Program were used to investigate current trends in age-adjusted and age-specific mesothelioma rates. An age and birth-cohort model was used to project both lifetime probabilities of mesothelioma by cohort and the annual number of cases expected over the next 70 years. The current trend in female rates is flat (age-adjusted rate = 0.30 per 100,000). The estimated lifetime risk for females is 2.5 X 10"4, independent of birth cohort. The projected average annual number of female cases is 500. For males, the age-adjusted mesothelioma rate is increasing solely due to the age group 75 years and over, albeit at a declining growth rate. Lifetime risk for males peaks at 2 X 10~3 for the 1925-1929 birth cohort, then decreases to 5 x 10-4 for the 1955-1959 birth cohort. The pattern of rates reflected in the age and birth-cohort model suggests a peak in the annual number of mesothelioma cases for males at 2,300 before the year 2000. The number of male cases then will drop during the next 50-60 years toward 500. These trends minor the US trend in raw asbestos consumption and a reduction in workplace airborne asbestos levels. Am J Epidemiol 1997;145:211-18. 'j > \ { ' . 1 asbestos; incidence; mesothelioma 1 Mesothelioma incidence often is interpreted as an the United States (3, 6-8) and in the United Kingdom index of past exposure to airborne asbestos. The inci (9). Spirtas et al. (3), using incidence data from New dence of mesothelioma for United States males exhib York State; Los Angeles County, California; and the its a generally increasing trend throughout the 1970s Surveillance, Epidemiology, and End Results (SEER) and early 1980s (1-3). The trend has been attributed to Program demonstrated a statistically significant in i I occupational exposure to asbestos, which, for some workers, was substantial from the 1930s through the crease for males over time by comparing data for 1977-1980 with data for 1973-1976. Peto et al. (9) 1960s (1, 4). Occupational exposure in the United analyzed death rates from mesothelioma registries in States during this time period occurred in the ship England, Wales, and Scotland. They reported an in building industry during World War II, in manufac creasing trend in the 1970s and 1980s and a continu turing, and during building construction (1, 2). Cur ation of the trend for men now under age 50 years, } rently, exposure potential exists for asbestos removal most of whom began work in the mid-1960s or later. workers; workers conducting renovations in buildings They concluded that exposure in the United Kingdom with asbestos-containing material; and maintenance, was greater around 1970 than in any previous period repair, and custodial workers in buildings with and that mesothelioma rates will continue to increase asbestos-containing materials. These exposures, how as that generation ages. ever, are orders of magnitude lower than historical Mesothelioma data from the SEER Program data occupational exposures (5). Trends in mesothelioma incidence rates have been base (10) for 1973-1992 were used to analyze current trends in age-adjusted and age-specific US mesotheli studied using various databases and methods both in oma rates and to project lifetime probabilities of con tracting mesothelioma for birth cohorts beginning with Received for publication May 13,1996. and in final form October 30, 1996. Abbreviations: SEER, Surveillance, Epidemiology, and End Re sults; SMSA, Standard Metropolitan Statistical Area. From Price Associates, )nc., Washington, DC. Reprint requests to Dr. Bertram Price, Price Associates, Inc.. 1800 K Street, N.W., Suite 71B, Washington, DC 20006. the 1885-1889 cohort and continuing through the 1955-1959 cohort. The results of the analysis show the downward direction of mesothelioma incidence in the United States. The pattern of mesothelioma inci dence mirrors the US trend in raw asbestos consump tion, which approached peak levels during World War HWBUI0010417 Price a II, as well as the timing and impact of government regulations that address asbestos exposure. MATERIALS AND METHODS Mesothelioma (pleural plus peritoneal) incidence rates were developed from the SEER database, which 'represents 9.5 percent of the US population (11). The database is organized by case. Each case is identified by age, sex, race, date of diagnosis, and other infor mation characterizing the cancer type. Data on malig nant mesothelioma of the pleura and peritoneum were compiled by selecting cases with International Clas sification of Diseases, Ninth Revision, morphology code 905 and topography code 163.9 for pleural me sothelioma and codes 158.8/158.9 for peritoneal me sothelioma. These data were used to develop rates for 5-year age groups in each diagnosis year. Three sets of rates were analyzed: 1) age-adjusted rates; 2) agespecific rates for 10-year age groups; and 3) 5-year age-specific rates for 5-year birth cohorts beginning with the 1885-1889 cohort and continuing through the 1955-1959 cohort. Trends in age-adjusted and age-specific rates were investigated by fitting a three-parameter logistic growth curve (12) to the data. The logistic curve smooths fluctuations in the observed incidence rates and highlights trends. The logistic growth curve equa tion is: Y = Ml + ft X exp[ -ft X X]) where Y -- age-adjusted or age-specific mesothelioma rate; X -- diagnosis year; and j3 = growth curve parameters. The distribution of male and female mesothelioma incidence rales by age and birth cohort were analyzed by Poisson regression (13, 14) to obtain models that could be used to project future mesothelioma inci dence and lifetime mesothelioma risk. Mesothelioma counts extracted from the SEER database by age and year of diagnosis were translated into 5-year age group counts for each 5-year birth cohort. Person-years for each age and birth-cohort group were developed in a similar manner using population counts recorded in the SEER database. These data were used to fit a standard age and birth-cohort model: u,j = a; X cj where fiy is the incidence rate for the ith age category and y'th birth cohort; a, is the age-specific incidence rate for the ith age category; and c, is a measure of the cohort effect for the jth birth cohort. The model was fit by the maximum likelihood method using GAUSS (Aptech Systems, Inc., Maple Valley, Washington). The cohort effects (i.e., {cy}) were normalized by forcing c9, the effect for the 1925-1929 birth cohort, to take the value J.O. The j other cohort parameter values, therefore, may be in- S terpreted as estimates of relative risk (i.e., relative to j the 1925-1929 cohort). For this parameterization of : the age and birth-cohort model, {a,} represent age- ! specific mesothelioma incidence rates for the 19251929 birth cohort. Predicted mesothelioma counts ; were obtained for the ith age group and jib birth cohort group by multiplying the incidence rate (w,y = a; X Cj) by the person-years for that group. DEVCAN (15), the National Cancer Institute life table analysis procedure for estimating lifetime cancer probabilities, was used to project the lifetime proba bility (also referred to as lifetime risk) of contracting 1 mesothelioma for each birth cohort. DEVCAN re- . quires age-specific mesothelioma incidence rates, agespecific mesothelioma death rates, and agespecific death rates for ail causes. For each cohort, the mesothelioma incidence rates used with DEVCAN were the observed rates calculated directly from SEER data, where available, or the predicted rates from the Poisson regression otherwise (i.e., <3, X cj for the (zj)th age and birth-cohort group). Because mesothe lioma is considered to be fatal within 1 or 2 years after diagnosis, the mesothelioma death rate was set equal to the incidence rate. Male and female death rates for all causes were based on US mortality statistics for 1990. Approximate 95 percent confidence intervals for lifetime, risk were determined by Monte Carlo simu lation. Monte Carlo replicates of lifetime risk were developed by using the observed mesothelioma counts for each age and birth-cohort group as expected values for the Poisson distribution. Based on these expected values, a new set of Poisson distributed mesothelioma counts were generated (16), the parameters of the age and birth-cohort model were estimated from these counts, and DEVCAN was applied to compute life time risk for each cohort. This procedure was repli cated 100 times. A confidence interval was calculated for each cohort from the mean and standard deviation of the 100 replicates. The Poisson regression models and DEVCAN also were used to project the number of mesothelioma cases for males and females in the future. Historical counts of male and female births were used as starting populations through the 1990-1994 birth cohort. Fu ture birth cohort populations were assumed to be the same size as the 1990-1994 cohort. To project beyond the 1955-1959 cohort, the downward male trend was continued by using the 1885-1889 birth cohort inci dence rates for the 1960-1964 birth cohort and the average female rates for all subsequent cohorts. To project mesothelioma cases for females, the average female rate 1959 cohoi sis were th oma risk: ? incidence based on L results Trends in Trends were anal; ihree-parai logistic gr satisfactor small and val. An al oma rate 1 stant rate ' to the dat: 1982 is 0 0.30 per 1 Plots o show: 1) females ai for males: versus ye; growth r. females. 1 males fel percent fc FIGURE Am J Epidemiol Vo!. 145, No. 3, 1997 Am J Ef HWBUI0010418 ............................ . ' Trends in Mesothelioma Incidence 213 1-0. The !e in- ,pve to ization oj isent agethe 1925- la COUDK rth cohori >j ~ o, titute life Tie cancer ne proba>ntracting CAN reates, agend ageohort. the >EVCAN >m SEER from the 'j for the mesotheears after set equal rates for istics for for simuisk were ia counts :d values expected thelioma f the age m these ute lifeas repliilculated leviation AN also helioma istorical starting ion. Fud be the beyond ?nd was art inciand the )rts. To average female rate was applied to all cohorts after the 19551959 cohort. Other parameters required for the analys;s were the same as those used to project mesotheli oma risk: Mesothelioma death rates were set equal to incidence rates; and mortality from all causes was based on US 1990 mortality data. RESULTS Trends in age-adjusted rates Trends in age-adjusted mesothelioma incidence were analyzed for males and females by fitting a ihree-parameter logistic growth curve (figure 1). The logistic growth curve does not provide a completely satisfactory fit for the female rate (p = 0.13), which is small and virtually constant over the 1973-1992 inter val. An alternative model, with a constant mesotheli oma rate for 1973 through 1982 and a different con stant rate for 1983 through 1992, provides a better fit to the data (p = 0.03). The rate estimated for 1973-- 1982 is 0.25 per 100,000; the rate for 1983-1992 is 0.30 per 100,000. Plots of the observed and fitted rates (figure 1) show: 1) a consistently higher rate for males versus females across all years; 2) a positive trend versus year for males; and 3) a virtually constant rate for females versus year. Figure 1 also indicates a declining annual growth rate for males and a zero growth rate for females. Based on the fitted curves, the growth rate for males fell from 14.0 percent for 1973-1974 to 0.4 percent for 1991-1992. The growth rate for females is zero, with the exception of 1982-1983, when a one time increase in the level by 20 percent occurred. Trends in age-specific rates Trends in age-specific incidence rates for age groups 45-54,55-64,65-74, and 75 years or more are shown in figure 2. For age groups 55-64 and 65-74 years, the trend is flat after 1983. For the age group 75 or more, incidence still is increasing in 1992, however, at a lower growth rate than in previous years. For each age group, the growth rate is declining. With the exception of the age group 45-54 years, which shows a decline in the rate over time, the logistic growth curve was fit to each set of data (fitted curves not shown). Table 1 contains a summary and comparison 5 of growth rates for 1982 and 1992 estimated from the ' fitted curves. The age group 75 years or more has the highest growth rate in 1992, 3.0 percent, down from 7.7 percent in 1982. Growth rates in 1992 for the other age groups are virtually zero, and for the age group 45-54 years, the growth rate is negative. Predicted counts and lifetime probability (risk) of mesothelioma Predicted and observed mesothelioma counts for males on the basis of Poisson regression analysis of the SEER data are shown in table 2. The relatively small differences between observed and predicted ffi Male Observed $ Male Predicted (3-Parameter Logistic Growth) x Female Observed a Female Prcdicied (Constant: 2-levels) FIGURE 1, Age-adjusted mesothelioma incidence (pleural + peritoneal) - observed and predicted. Based on SEER data. November 1995. Am J Epidemiol Vol. 145, No. 3,1997 1 HWBUI0010419 li... . TABLE 1. Male age-specific mesothelioma incidence (pleural Projected number of mesothelioma cases + peritoneal) growth rates in 1982 and 1992 based on the three-parameter logistic growth curve Age group (years) Year 1982(%) 1992 (%) The projected numbers of mesothelioma cases are displayed in figure 4. On the basis of characteristics of mesothelioma incidence described above, the number of female cases will remain constant at approximately 45-54* 55-64 65--74f 275 -2.2 2.5 3.2 7.7 -2.8 0.1 0.1 3.0 * The logistic growth curve was not fit to these data because the incidence rate declines over time. The growth rates in the table were derived from a straight line tit (y = 1.413 - 0.026 x t, t= 1, 2, .... 20; R* = 0.172, p value = 0.069). t The data observation for 1989 was treated as an outlier and was not used to obtain the logistic fit (sea figure 2). 500 per year. The number of male mesothelioma cases is likely to peak before the year 2000 at approximately 2,300 cases and then decline to approximately 500 cases per year by 2055. DISCUSSION The analysis of mesothelioma incidence trends re ported here is based on data collected in the SEER program, which may be interpreted as a sample rep counts indicate that the model adequately captures the resenting the US population. The SEER Program cov mesothelioma incidence trends. ers five states: Connecticut, Iowa, New Mexico, Utah, DEVCAN life table analysis results are shown in and Hawaii, and four metropolitan areas: Detroit Stan figure 3. The two curves in figure 3 display lifetime dard Metropolitan Statistical Area (SMSA), Atlanta mesothelioma risk versus birth cohort for males and SMSA, San Francisco-Oakland SMSA, and Seattle- females. (The bars are approximate 95 percent confi Puget Sound. These regions include 9.5 percent of the dence limits obtained from Monte Carlo simulation.) US population (11). Walker et al. (17) suggested that The curve for males increases, reaching a maximum SEER data overestimate mesothelioma incidence be I risk of approximately 2 X 10-3 for the 1925-1929 cause of a disproportionate number of shipbuilding birth cohort, and then decreases. Statistical tests (not areas in the SEER regions. Nicholson (1) argued that ; shown) comparing the risk for the 1925-1929 cohort the SEER data underestimate mesothelioma incidence with risks for subsequent cohorts substantiate the because large urban areas where asbestos was used in downward trend for males. The risk curve for females manufacturing and construction are underrepresented. is essentially flat, at an average level approximately Spirtas et al. (3) compared the white male pleural f equal to 2.5 X 10"4. cancer mortality rate for the SEER regions with the Am J Epidemiol Vol. 145, No. 3, 1997 Am J HWBUI0010420 Trends in Mesothelioma Incidence 215 5D s w T- CM CO CO u> CO N CM P"* CO CM O Ov-TCM- oh*. CcDo iin- Uto*5 h05c0o hn* nCM ,g 0> U5 K K 05 88 8 88 !2? 8 ^5 00 CO xf f- C> ^ f^-cd xf tf> wd OO rv<-- t- u> u> 0c5o oo> coco r**. x* 3 c> o CM m N 050 O o N <D x <0 OO <N Oi xf CO CM C5 fs. - h* tf5 i-; CO CM W *" *i> COCO COCO N O xj- U"> 05 CO 005 05 0> T- CM 05 CM CM CO O s are iCS of mber m ately 500 r-. co o> <q ui xf CM lO S8 SS 05 xf> 88 f-<* hO CM OS f-x CO <> 05 cO'-' O 05 O " s re- EER repcovJtah, >tanlanta ittief the that be- ding that ;nce d in ited. ural the * ` 1 i * A 4, S. If ID O c as V Z A 9 CO | Sjf & cv o f i 5 So \J O o J(8* 2 S CM is* 6vg<N Ui co o o> o J 03 CO 05 CO co in O O W rf (O (D ui h- xj evi o co to o o co id xj- ur> o xt O O P) r g<OfI> Oo-J ?ionI (hoO: A" 5 - i go : gj. ;L' i" io 8 8 8" 05 X? x* in ;f 05 CM ^ CO i 5 J.% AS S 8 Am J Epidemiol Vol. 145, No. 3, 1997 HWBUI0010421 Birth Cohort Mid --Interval Year FIGURE 3. Lifetime probability (risk) of mesothelioma (pleural + peritoneal) and 95% confidence intervals. DEVCAN life table analysis based on mesothelioma incidence rates recorded in the SEER database, November 1995. FIGURE 4. Projected number of mesothelioma cases. Poisson regression and DEVCAN life table analysis used to project the number of mesothelioma cases into the future based on SEER data, November 1995. total US rate and concluded that the SEER data may overestimate the national incidence, but that analyses of trends based on these data would not be affected. With respect to selected demographic and epidemio logic factors, the SEER regions are reasonably repre. sentative of the US population (11). The trend in female rates is a baseline or background for evaluating mesothelioma incidence trends I in general (2). The trend in age-adjusted incidence for females is essentially flat, exhibiting a constant rate of 0.25 per 100,000 until 1982 and then a slight increase fto 0.30 per 100,000 from 1983 through 1992 (figure 1). The shift that occurs in the 1982-1983 time inter val is most likely a diagnostic effect, a consequence of Environmental Protection Agency regulatory activities during the 1980s (18-27). Connelly et al. (8) conclude that the diagnostic effect in US data is real, but note that any large impact due to diagnostic changes is unlikely. An alternative explanation of the small shift in the female rate, namely that environmental expo sure to airborne asbestos is increasing, is not supported by the data. The absence of a steadily increasing age-adjusted rate for females makes "increasing envi ronmental asbestos exposure" unlikely as an explana tion for the shift. The increasing trend in age-adjusted rates for males is due to the continuing upward trend in the age group 75 years or more. Growth rates for the age group 75 or Am J Epidemiol Vol. 145, No. 3, 1997 more, i rates f> negativ Thes specifn three-p model linear!} curve f ations i data an capture has not project data. The cline ir the pat lifetimi Male n in ship the yea imum; nifican 1930s. peaked decline 1929 1 peak a Wor cupatic Envirc The O has re< since 1 Agenc structi demol exposi worke worke with c ductin buildi The cline using mode! provit jectin. from erthel amou recen Am J i HWBUI0010422 V. is based ! mber of :ivities nclude it note ges is 1 shift expoported easing enviplanamales group 75 or Trends in Mesothelioma Incidence 217 more, although positive, are falling (table 1). Growth rates for all other age groups are near zero or are negative. These general trends in age-adjusted and agespecific rates (figures 1 and 2) are represented by the three-parameter logistic growth curve. The curve is a model for quantities with growth rates that decline linearly as the quantity increases (12). The logistic curve has been used in this analysis to smooth fluctu ations in incidence rates over the range of the observed data and thereby highlight trends. It cannot, however, capture downward trends in incidence. Therefore, it has not been used in this analysis to make quantitative projections of incidence beyond the range of observed data. The overall dynamics of long-term growth and de cline in male mesothelioma incidence are reflected in the pattern of lifetime risk (figure 3). The maximum lifetime risk appears for the 1925-1929 birth cohort. Male members of this cohort would have been at work in shipyards, manufacturing, and construction during the years 1930-1960, a period of increasing and max imum asbestos consumption in the United States. Sig nificant growth in the use of asbestos began in the 1930s, was halfway to peak consumption in 1940, peaked in 1950, where i( remained until 1970, and then declined precipitously (2, 28). Workers bom after 1929 have experienced fewer years of exposure at peak asbestos consumption levels. Workers born after 1929 also benefit from the Oc cupational Safety and Health Administration and the Environmental Protection Agency asbestos programs. The Occupational Safety and Health Administration has reduced its permissible exposure limit four times since 1971 (29-31), and the Environmental Protection Agency restricted the use of asbestos in building con struction and imposed work practices for building demolitions (32). Currently, the potential for asbestos exposure, albeit at very low levels relative to historical worker exposures, remains only for asbestos removal workers; workers conducting renovations in buildings with asbestos-containing material; and workers con ducting maintenance, repair, or custodial activities in buildings with asbestos-containing material. The overall dynamics of long-term growth and de cline in mesothelioma incidence have been analyzed using the standard age and birth-cohort model. From a modeling perspective, the age and birth-cohort model provides a more detailed and accurate basis for pro jecting mesothelioma trends than do trends derived from age-adjusted and age-specific aggregates. Nev ertheless, the model is an approximation, and the amount of data available for the earliest and most recent cohorts is sparse. However, the relatively small differences between observed and predicted mesothe- . lioma counts (table 2) indicate that the model ade- i quately captures the general trend in mesothelioma risk. The estimates of lifetime risk reflect a relatively large degree of statistical uncertainty for the most ; recent birth cohorts (refer to the 95 percent confidence intervals in figure 3), but not large enough to obscure i the overall downward trend. The projected number of mesothelioma cases for . future years is sensitive to assumptions concerning the incidence rates for birth cohorts after 1959. For fe males, the average historical mesothelioma rate was ; applied to all cohorts after the 1955-1959 cohort. This assumption is not controversial because the trend for : females has been virtually constant for the past 20 ' years. The projected numbers of male cases are based on stronger assumptions. To project beyond the 1955- i 1959 cohort, the downward male trend was continued, by substituting the 1885-1889 birth cohort incidence rates for the 1960-1964 birth cohort and the average female rates for all subsequent cohorts. Different as sumptions about future rates could either increase or decrease the number of cases projected for the future or extend the time it will take to reach background levels. However, independent of the particular as sumptions selected, the characteristics of the trend would not change, with a peak around the year 2000 at approximately 2,300 cases, followed by a decline to a constant background level. The analysis by Peto et al. (9) of mesothelioma in the United Kingdom indicates a peak around the year 2020. The difference between the United States and United Kingdom results is a consequence of the timing of maximum exposure. Peto et al. report that maxi mum exposure in the United Kingdom occurred around 1970, whereas maximum exposure in the United States occurred from the 1930s to the 1960s. Considering the time period of maximum exposure, the subsequent downward trend in exposure, and the currently low exposure levels experienced by workers (5), the assumptions used to project the number of future US cases are reasonable. The US peak in cases occurring approximately in the year 2000, and the decline during the next 50-60 years toward 500 re flects both the US trend in raw asbestos consumption and reductions in workplace asbestos exposure levels over time. REFERENCES 1. Nicholson WJ. Occupational exposure to asbestos: population jat risk and projected mortality 1980-2030. Am Ind Med 1982:3:259-311. 2. McDonald JC. Health implications of environmental exposure Am J Epidemiol Vol. 145, No. 3, 1997 HWBUI0010423 |HI8 Price to asbestos. Environ Health Persp 1985;62:319-28. 3. Spinas R, Beebe G, Connelly R, et al. Recent trends in mesothelioma in the United States. Am J Ind Med 1986;9: 397-407. 4. Dupre JS. Mustard JF, Uffen RJ. Report of the Royal Com mission on matters of health and safety arising from the use of asbestos in Ontario. Ontario, Canada: Ontario Ministry of the Attorney General, 1984. 5. Price B, Crump KS, Baird EC. Airborne asbestos levels in buildings: maintenance worker and occupant exposures. J Exp Anal Environ Epidemiol 1992;2:357-74. 6. Peto 3, Henderson BE, Pike MC. Trends in mesothelioma incidence in the United States and the forecast epidemic due to asbestos exposure during World War II. In: Peto R, Schneiderman M, eds. Brandbury Report 9, Quantification of Occupational Cancer. Cold Spring Harbor, NY; Cold Spring Harbor Laboratory, 1981. 7. Archer VE, Rom WN. Trend in mortality of diffuse malignant mesothelioma of the pleura. Lancet 1983;2:1112-13. 8. Connelly R, Spinas R, Myers M, et al. Demographic patterns for mesothelioma in the Unites States. J Nall Cancer Inst 1987;78:1053-60. 9. Peto I, Hodgson J, Matthews F, et al. Continuing increase in mesothelioma mortality in Britain. Lancet 1995;345:535-9. 10. National Cancer Institute. Surveillance, Epidemiology, and End Results cases diagnosed 1973-1992. Submission cancer public use database. Belhesda, MD; US Department of Health and Human Services, 1995. 11. Ries LAG, Miller BA, Hankey BF, et al., eds. SEER Cancer Statistics Review, 1973-1991: tables and graphs. Bethesda. MD: National Cancer institute, 1994. (NIH publication no. 94-2789). 12. Draper NR, Smith H. Applied regression analysis. 2nd ed. New York, NY: John Wiley & Sons, Inc., 1981. 13. Frome EL. The analysis of rates using Poisson regression. Biometrics 1983;39:665-74. 14. Frome EL, Checkoway H. Epidemiologic programs for com puters and calculators: use of Poisson regression models in estimating incidence rates and ratios. Am J Epidemiol 1985; 121:309-23. 15. Feuer EJ, Wun L, Boring CC. Probability of developing cancer, cancer statistics review 1973-1989. Bethesda, MD; National Cancer Institute, 1995. {NIH publication no. 92-2789). 16. Fishman GS. Concepts and methods in discrete event digital simulation. New York, NY: John Wiley & Sons, 1973:273. 17. Walker AM, Loughlin JE, Friedlander ER, el al. Projections of asbestos-related disease 1980-2009. J Occup Med J983;25: 409-25. 18. Final rule for friable asbestos-containing materials in schools; identification and notification. Federal Register. May 27. 1982;47:23360-89. 19. Nicholson WJ, Rohl AN, Weisman 1. Asbestos contamination of the air in public buildings. Research Triangle Park, NC; Environmental Protection Agency, 1975. (EPA publication no. EPA-450/3-76-004). 20. Nicholson W, Rohl AN, Sawyer RN, et al. Control of sprayed asbestos surfaces in school buildings: a feasibility study. (Pre pared for the National Institute of Environmental Health Sci ences). New York. NY: Environmental Sciences Laboratory. 1978. 21. Environmental Protection Agency. Asbestos-containing mate rials in school buildings: a guidance document (The Orange Book). Washington, DC: Environmental Protection Agency, 1979. (EPA publication no. EPA-C00090). 22. Notice announcing program for asbestos-containing materials in school buildings, school asbestos program. Federal Regis ter. March 23, 1979;44:17790-1. 23. Environmental Protection Agency. Guidance for controlling friable asbestos-containing materials in buildings (The Blue Book). Washington, DC; Environmental Protection Agency, 1983. (EPA publication no. EPA 560-5-83-002). 24. Environmental Protection Agency. Guidance for controlling asbestos-containing materials in buildings (The Purple Book). Washington. DC: Environmental Protection Agency, 1985. (EPA publication no. EPA 560-5-85-024). 25. Environmental Protection Agency. Communicating about risk: EPA and asbestos in schools. Final report of the internal task force. Washington. DC: Environmental Protection Agency, 1992. 26. McCormick K. Asbestos: the clock is ticking in your schools, and inaction could prove to be devastating. Am School Board J 1984;171:33-6. 27. Siavisky LP. State responsibility for the control of asbestos in the schools. J School Health 1982;August:358-64. 28. Bureau of Mines. Asbestos in 1992. Washington. DC: United States Bureau of Mines, 1993. (Bureau of Mines Annual Report-Mine Industry Surveys). 29. Occupational Safety and Health Standards; standard for expo sure to asbestos dust. Federal Register. June 7, 19?2;37: 11318-22. 30. Final rule for occupational exposure to asbestos, tremoiite. amhophyllile. and aclinolite. Federal Register. June 20. 1986; '51:22612. 31. Final rule of occupational exposure to asbestos. Federal Reg ister. August 10. 1994;59:40964-1165. 32. National emission standards for hazardous air pollutants; as bestos, beryllium, and mercury. Federal Register. April 6. 1973;38:8820-3. Am J Epidemiol Vol. 145, No. 3, 1997 Electr High-' Tore Tyi i i \ 1 i i i i i In 1rnortaVt with it years I amort? (2). Su ing ck pothes' frequet origin vided : magne hood \ feels arounc logic i In I addres and c; nested Rect Octobe Abbi ratio. 'Ca Reaean a No Repi lion Au HWBUI0010424