Document zdv4EgznOrjDqJNGRyBDm6wwm

ESStf i/s J'rAMtZj-Ci- tSaZuo i Estimating risk to human health \W:- Trichloroethylene in drinking water is used as the example C. Richard Cothern William A. Coniglio William L. Marcus Environmental Protection Agency Office ofDrinking Water Washington, D. C. 20460 The decision to regulate the maxi mum level of an environmental contaminant allowable in air, water, or food is predicated on the ability of the substance to alter life processes and on the degree of exposure that creates a significant risk of adverse effects to a sensitive population. For more than four centuries, scientists have tried to understand the adverse effects of exposure to environ mental contaminants. In general, any contaminant can cause damage if exposure to it creates high concentrations within a cell, and if accumulated cellular dam age can result in disease or death. Envi ronmental regulations are developed to reduce exposure to a level that pre cludes such health effects. Drinking-water regulations are being developed for many contaminants, some of which are known to cause can cer in laboratory animals. One group, the volatile organic compounds (VOCs), is currently being assessed for the development of national drinkingwater standards. Reference 1 contains an explanation of the risk assessment studies being used to evaluate VOCs. 4 This article not subject to U.S. copyright. Published 1986 American Chemical Society ti'#r^SL 03AA64J Environ. Set. lechnot.. MX. 20. No. 2, I960 ail , x. Levels of human exposure to VOCs found in drinking water and the level of cancer risk imposed by their presence have been estimated as part of the proc ess of developing regulations. Esti mates include the health benefits to be derived from different levels of national regulation. They couple the estimate of carcinogenic potency, derived from an extrapolation of a dose-response curve developed from high-level exposure of experimental animals, with estimates of the number of persons exposed to con taminated drinking water. The esti mates include the level of daily intake-- oral, respiratory, and dermal--resulting from the presence of these chemicals in water (Figure 1). Estimates of the number of cancers resulting from currently unregulated conditions and of those cancers that will be prevented or allowed by setting standards at different levels can be in cluded in the regulatory process. Cur rent knowledge about the health effects of contaminants has its limits: Much is known about some aspects; little or nothing is known about others. But there is uncertainty in all areas. The population risk estimate is an as sessment of the total expected excess number of cancers in the United Slates during a lifetime from a given contami nant. This is an estimate and not a state ment of perfect knowledge. Therefore, a level of uncertainty or a range of val ues will be included to describe the esti mate. If the frequency distribution of the contributing factors controlling ex posure and toxicity were known, the propagation of errors could be deter mined by use of standard methods. Fre quency distributions often are not known, however, and a range of values often must be used to approximate the situation. This article describes some of the data on which regulatory judgments are made. It also describes the assumptions introduced to complete the assess ments, the results, and their uncer tainty. We take as our example the VOC trichloroethylene (TCE). Recent research has shown that TCE FIGURE t Risk estimation process - -. . Individual risk rate (ingestion, V\ Occurrence in inhalation. `-v-.......... drinking water dermal) r 1 i Number of cancers 1 1 i averted by eliminating J ! concentrations in j '*""*! S2&. I-- j_ JPopulation risk >\ I . r' .- i i drinking water that lead to a nsk exceeding 10-4,10`s, and 10*` per lifetime, and the number of public drinking-water systems affected J { | | | '* * ' L __ 1 can be transformed in water to vinyl chloride, a known human carcinogen (2). The cancer risk associated with the presence of vinyl chloride in drinking water also has been assessed (1). TCE in drinking water There arc approximately 60.000 pub lic water supplies in the United States, serving some 220 million persons. Three major federally sponsored moni toring surveys have been conducted na tionwide to obtain information on the concentration of selected VOCs appear ing as pollutants in potable water pro vided to the nation's cities and towns. By using the survey data in conjunc tion with the multinomial approach for projecting national occurrence, it was estimated that 97% of groundwater sys tems of all sizes contain either no TCE or concentrations <0.5 pgIL (Table 1). It is not possible to estimate how many systems in this category are contami nated with TCE. Of the estimated 1632 systems ex pected to have levels >0.5pg/L, 421 (0.9% of the total number of groundwater systems) are expected to have concentrations >5.0pg/L, and 133 systems (0.3%) nationally have con centrations >50/ig/L. Although a greater percentage of the large systems is expected to be contaminated, because of the large number of small groundwater systems it is likely that in abso lute numbers more small systems are contaminated. In a similar analysis of surface water supplies, it was estimated that 96% ei ther have no TCE present or that levels are <0.5 pg/L. It is estimated that 498 surface water systems have levels a 0.5 pg/L (4.5% of total surface wa ter systems); 9 (<0.1%) are expected (population ervad) ystema --v-. tn U.S. <0.5* 0.5-5 >5-10 >10-20 >20-30 >30-40>40-50>50-60 >60-70 > 70-60 > 80-90 > 90-100 >100 . ,, . 25-100 ; 19,125 18,506 465 26 52 0 26 0 ' 26 - o :*/$- o 'o . 0 i'V 26 i;*;. 101-500 . 5,674 15,166 381 21 . 42 110 21 :i:o v. 21 --.. 0 .fi- 0 -- .0 o r/;. 21.&&. 501-1.000 C 4.877 4,719 1,001-2,500. , ^4,400 4,257 V,` 2.501-3,300 k't? 691 .. 862 3,301-5,000 V/1.065 1,031 118 7 13 6 0. ,;Iv7 107 6 12 ^.0 6 -v'i 0 'a'. 6 ,a: 22 i;.., 2 . 26. ; i. :. 3 0 0 ;,>J Vt <V.':,0- -,v?l 1 o 'J o o o ; o vr-:* 0 >: 0 . 0 0 : 0 ?;"o . 0 -; 0 :a." 0 iittTA ..;$"* J 5,501-10,000 , 1,168 1,130 28 2. ' 3 JO 2 0 ,i -'2 V-. 0 0 .0 9 .. 'i- 2 - ' 10.001-25,000 ` 835 -. 775 34 11 4 ,-r o 08 25.001-50,000 290 269 12 ; 4 V: :.i.r.o ":r-o" ;.3 0 *0 0 .'4 0 , ` 0 0 ' T ' 0' : ; 0 i 0 it ; 0 Y: . 50.001-75,000 64 59 . 3 1 75.001-100,000 13 : 0 0, 0. ,, 0 00 0 i o. 0, o ` `.0 0 0 0 0 0 'Q A ; >100,000 . Total ; ^ 55 41 14 0 46.458 46,828 1,211 80 00 132 Tv 0 '. 64 00 12 ' 64 000 0V I v"**- 641 S-W' Calculated as the difference between the number of systems expected to have % 0.5 TCE and the total number of systems in that else category. This group includes those having no contamination and those with levels <0.5 /ig/L , > `: ; -*yAZt.-r Ill Environ Sci Teehnol , Vol 20, No, 2.1986 SL 034465 to have levels > 5 /ig/L; none is esti mated to have TCE >40 rig/L. Using the combined data from sur face water and groundwater supplies, it was estimated that nearly 190 million persons (88.3% of the population served by public drinking-water sys tems) receive uncontaminated water or water contaminated with TCE at levels <0.5 /rg/L. Of the 25 million persons (11.7%) receiving water that contains TCE at levels >0.5 Mg-'L. an estimated 1.8 million (0.9%) are exposed to lev els >5 Mg/L. About 212.000 persons (0.1%) are probably exposed to levels >50/ig/L, and 42.000 (<0.1%) are estimated to be exposed to levels > 100 ng/L. Of the approximately 23 million per sons exposed to levels ranging from 0.5 pg/L to 5 fig/L, 76% obtain water from surface water supplies. However, of the 1.8 million persons exposed to levels >5 fig/L, 62% use groundwater sources. All exposure to TCE in drink ing water in levels >40 fig/L is pro jected to be from groundwater sources. Several assumptions have been made during the assessment of TCE contami nation of drinking water, among them that sites selected in surveys are repre sentative of drinking-water supplies across the nation and that grab samples are representative of water quality throughout the water supply system. Using scientific judgment, the uncer tainties in these estimates were ana lyzed. The uncertainties include those in the measurements themselves and those involved in projecting the existing data to estimate natural occurrence. In general, the overall uncertainty is less than an order of magnitude. Human exposure to contaminants The average adult in the United States uses 48-55 gal of water each day (3, 4). This means that a household of two children and two adults uses ap- proximately 255 gal of water each day (5). Water pollutants pass through the home in water that is used for launder ing clothing, washing dishes, removing wastes, bathing, cooking, and drinking. The level of exposure to any pollu tant present in drinking water is the result of many personal daily choices and several factors over which we have very little direct control. Where we live, what we eat and drink, and how old we are all have a profound influ ence on the magnitude of exposure. Also, the physical and chemical charac teristics of the pollutant govern the amount that stays in the water and that which is transported into indoor air. Table 2 provides an example of the potential differences in exposure of Expoaure route Formula-led '"'-'"-s*' ' Infant (4 kg) Preteen Adult female Adult male* (32 kg) - (60 kg) (70 kg) Fluid ingestion Inhalation of ` enriched indoor air Inhalation of '*- - enclosed shower air Dermal absorption . Bathing ; ' *-v - Swimming 80 , .150 . 200 10 7' .... mo .. ... : i, .? ^.--.300 .02-.06 -v 10-300 - , . .V . 200 50 * 200 ' , -- Total absorbed dose , 20 pg/kg/d 10-20 pg/kg/d 10 pg/kg/d 7 pgfkg/d V. *100 rig/L -iV'*;/-' `Showering adult male . . i "'A'TM . ,, fV ' *Vt ` '' 1 J-,- ^ ' ,U\' '*`>1 ' 'i,'.'/<** : ' 'il B** Vr ti _L family members to a VOC resulting from personal choices. This table has been prepared for a theoretical VOC, similar in chemical characteristics to TCE. We assumed that 100% of the compound is absorbed for all oral in take, 50% for respiratory intake, and a very small amount is absorbed in direct proportion to water flux across the skin. Cancer risks projected for TCE are based on assumed lifetime ingestion of 0.03 L of drinking water for each kilo gram of body weight and on the as sumption that 100% of the ingested chemical is absorbed. The factor for in gestion was developed using an intake standard of 2 L/70 kg adult weight. In dividuals may experience many times this intake rate. Formula-fed infants and young children, for example, have average intake rates that are as much as eight times greater than those of aver age adults. Adults in tropical areas may consume twice as much liquid as the average, as may athletically inclined adults when engaged in strenuous physical activity. Persons who are ill also may consume much more water than the average. So cial behavior ritualized around the drinking of tea or coffee may lead to increased water consumption, although boiled water may contain smaller amounts of contaminants. A recent Ca nadian study shows that children under 5 years old, 10% of the children 6-17 years old, and 2% of the adult populaton consume more than 0.03 L/d of drinking water for every kilogram of si 34466 Environ. Sci. Technol.. Vol. 20, No. 2,1986 113 body weight (6). Each of the VOCs considered, in cluding TCE, has been shown to trans fer from water into air if the water is heated or aerated. Monitoring data in dicate that this process is continuous within the home and leads to an imme diate enrichment of respirable air at the point of water use and to a diffusion throughout the home. Showers taken within an enclosed bathroom, using 20-30 gal of water, result in the libera tion of all or a portion of the VOCs into the air. A person who showers inhales air containing a highly enriched level of the chemical while inside the bath room. VOCs from other sources as well mix with general home air, resulting in higher levels of the pollutants being relased into the indoor air. Very few data are available on the enrichment of general indoor air with VOCs. However, research has been conducted on the radon enrichment of indoor air. and it is assumed that the fraction of radon released is about the same as that for VOCs. These data indi cate that an enrichment of indoor air (the amount of radon in a liter of air) is about 10J to the concentration found in a liter of drinking water (7-10). This model assumes an indoor air change rate of one per hour. For our analysis it has been assumed that respiratory exposure to VOCs in drinking water is roughly equivalent to that from ingestion. To compare the rel ative exposures from inhalation and in gestion, assume that for both pathways 100% of the VOCs that enter the body reach the bloodstream. Then if water with a concentration of X /*g/L is in gested, the daily intake to the blood stream is 2X Mg/d. For example, X fig/L X 2 L/d = 2X Mg/d (1) If water with a concentration of X tig/L transfers all of the VOC to indoor air and it all goes to the bloodstream, then the intake to the bloodstream is again 2XMg/d. For example. Xfig 10'* L (water) L L (air) 20 m3 (air) 106 cm3 (air) d m' (air) L (air) X 1000 cm5 = 2*Mg/d (2) Under these conditions exposure from ingestion and inhalation is equal. Individual risk rates Estimates of individual risk rates are based on bioassay data derived from animal experiments. These data are converted to a continuous human equivalent exposure for use in risk esti mates. The bioassay data selected to es timate carcinogenic potency of VOCs in drinking water are from experiments TABLE 3 - ' - V" T;-'" ' ^ Population risk estimates for current TCE levels in drinking water Mean drinking-water concentration _______ (i*g/L) Number of perrons ______ served______ Total lifetime Individual risk for the mean concentration* Low (problt) High (Welbull) 0.25 2.75 1.9 x 10* 2.3x 107 7.5 4.3 x 10* 15 , ! 1*; '. 2.1 x 105 35 -.W?7.4 x10s 45 .......>- 2.6 x10s 55 - V - 4.2 x 104 75 . . .. 1.3 x 10s . . 100 . . 4.2 x 104 Total6 ` - :' ' ' . <10-'0 ... v: -. 2.4X10-*. 7.3 x io-* - 1.3x10^ 1.7 x 10~* ' 7 x 10- ^ 2.3 x 10rS 1 .3 x 10-7 2.6 x io-* 4 x 10-7 - 2.B x 10- 6 x 10'7 3.2 x 10-* v 1.2 x 10-* 7 3.7 x IC* 'The total individual risk was determined by assuming that the risk due to inhalation Is equal to that due to Ingestion "Rounded to one significant figure ., SL 034467 114 Environ Sci Technol, Vol, 20. No, 2, 1986 s-s-'.-,rrir-- L:' Lifetime population risk 1-45,600 1-16,790 .<1-559 ...V;' <1-367 TTV;/< 1-1.702 . 7* <1-676 :;773T<i-ii7 1f ; > <1-70,000 '-y-wrj-.' - *" j.ti'L V`T` i reviewed and selected by the National Academy of Sciences and the EPA Car cinogen Assessment Group (77. 72). When the carcinogenic data were from inhalation experiments, they were con verted to ingestion values. The bioassay data are fit to four ana lytical models: logit, multistage, probit. and Weibull. These four models are chosen somewhat arbitrarily and are representative of models currently in use. It is well-known that other models could be used to fit the same data (7J16). There are no biologically based criteria for choosing one model over another. The models are in the form of analytical expressions that can be used to fit the high-dose data mathemati cally. The dose-response curves for the models are fit to equivalent human dose-response data converted using the relative surface area of animals and hu mans. The curses are extrapolated to the lower environmental levels. A loglog plot of the entire region shows the two nonzero data points for TCE (Fig ure 2). The solid lines are the point esti mates. and the error bars show the up per 95% confidence limits. The GLOBAL computer program was used for the multistage model, and the Krcwski computer program was used for the logit, probit, and Weibull models. The curves shown in Figure 2 are not the only models that could be fit to the data. Additionally, there is no assurance that the actual curve is in the range of those shown. It is possible, although highly unlikely, that the actual curve lies outside those shown. For example, the probit model generally decreases extremely quickly relative to dose, whereas the Weibull model decreases very slowly. Although the dose-response curves projected by each model in Figure 2 start at the same points, they diverge significantly at lower dose levels. At a concentration of 50 jig/L TCE in drink ing water, the Weibull model provides a risk estimate approximating 1 x 10'2, whereas the probit model provides an estimate of 10'l0. These estimates pro vide a range of uncertainty equivalent to not knowing whether one has enough money to buy a cup of coffee or pay off the national debt. Population risk calculations The population at risk of health dam age due to TCE in drinking water was estimated. The population risks are de termined by a product of the number of people affected and the individual risk rate (Table 3). In each case the concen tration is in the average of the range. The lowest concentration (<0.5 ng/L) represents the minimum reported con centrations for the national surveys and does not necessarily represent the de tection limit of the chemical in drinking water. Although we have used the esti mated average of 0.25 /rg/L for this category, it is possible that all of the persons in this category could be ex posed to 0/rg'L or that all of them could be exposed to the maximum of 0.5 jig/L. For each chemical the indi vidual lifetime risks were determined from the model fits. In each case the curves in Figure 2 giving the lowest and the highest indi vidual risks were used to determine the range of population risks. The uncer tainty in choice of model in Figure 2 is by far the largest uncertainty of any in volved in the estimates of risk. Thus it alone was used to calculate the range of risk. When calculations yielded less than one person, the number was listed as a whole number rather than as the fractional number of persons. The risk reduction analysis (Table 4) is a determination of the number of cancer cases averted if the standard is set corresponding to individual lifetime risk rates of 10"\ 1(H, or 10'6. For these individual risk rates, the corres ponding range of concentration and the cases averted are listed. The concentra tion range was determined from the graph shown earlier in Figure 2. The population risk estimates are listed as ranges; the range is indicative of the uncertainty in this estimate. For TCE, the lower estimate for each TABLE 4 Maximum allowable drinking-water concentration (afl/L) ... Approximate Individual rlak rate lor maximum " ' concentration ' Cumulative caaea ' averted* 100 45 ' " 7.5 2.75 - 1 X 10'r to 1 x 10"* 1 x 10-* to 1 x 10'* <1 x 10_,0to6 x 10"* <1 x 10-' to : - <1-200 " 'V"V.'<1-1.000 ' ^ 1-4,000 ' For limiting TCE concentration In drinking water -rNeuumboer otf cases averted ftor concentration shnouwn in iirsi column j, ;., v.-,, v.. concentration range is less than one. The sum of the upper population risk estimates is 70.000. when rounded off to one significant Figure. The largest portion of the population risk is con tributed by the lowest concentration category. This category represents cal culations below the detection limit and may be an artifact of the mathematical analysis technique used to estimate the population risk. The range is largely a reflection of the uncertainty in choice of the mathematical model selected to estimate individual risk. The contribu tion to uncertainty from other factors is negligible by comparison. If a drinking-water standard--a max imum contaminant level--for TCE were established at 7.5ng/L, nation wide compliance would result in the avoidance of anywhere from < 1 to 4000 cases. Explanation The development of regulations cov ering contaminants in drinking water requires the assembly of data from many different sources. One important piece of information needed is the esti mate of the expected number of health effects and how many of these can be prevented by regulation. Risk assess ment usually involves using skimpy data concerning the levels of a contami nant in drinking water and using health effects data from animal experiments conducted at exposure levels much higher than those usually found in the environment. The risk can be estimated using TCE in drinking water as the example, al though the range of uncertainty is quite large. If there were no data, the range of estimated health effects would be be tween none and all the people. Using the available data for TCE in drinking water, the estimated range of health ef fects lies between < 1 and 70,000. If all exposure to levels >7.5/tg/L were eliminated, as few as one or less of these health effects could be prevented, or perhaps as many as 4000. These esti mates are the best representation of the existing data for TCE in drinking water. Their value to the regulatory decision making process must be judged when placed alongside the other information used in that process. Acknowledgment Wc thank Joseph A. Cotruvo for his en couragement and helpful discussions on the work described here. The thoughts and ideas expressed in this article arc those of the authors and are not necessarily those of the EPA. References (I) Colhcrn. C. R.; Coniglio, W. A.: Marcus. W. L. "Techniques for the Assessment of Carcinogenic Risk to the U.S. Population Due to Exposure from Selected Volatile Or- Environ. Sci. Teehnol., Vol. 20, No. 2.1986 115 funic Compounds from Drinking Water via the Ingestion. Inhalation and Dermal Routes." EPA 570/9-85-001; EPA; Wash ington. D.C.. 1985. (2) Vogel, T. M.; McCarty. P. L. Appl. Envi ron. Microbiol. 1985, -49. 1080-83. (3) Watson. K. S.; Farrell. R. R; Anderson, J. S. J. Water Pollut. Control Fed. 1967, 50(12), 2039-54. (4) Lenaweaver. F, P.. Jr. In "A Study of Flow Reduction and Treatment of Wastewater from Households"; Bailey. -J.R., Ed.; Gen eral Dynamics report to the Federal Water Quality Administration: Washington, D.C., 1969. (5) Bailey, J. R. et al. "A Study of Flow Re duction and Treatment of Waste Water from Households," Water Pollution Control Re search Series II050FKE: General Dynamics report for the Federal Water Quality Admin istration; Department of the Interior: Cin cinnati. Ohio. 1969. (6) "Tap Water Consumption in Canada." 82EHD80; Canadian Environmental Health Directorate, Ministry of National Health and Welfare: Ottawa. Ont., 1981. (7) "Sources and Effects of Ionizing Radia tion," Report to the General Assembly of the United Nations by the United Nations Scientific Committee on the Effects of Atomic Radiation. 1977. (8) Hess. C. T.; Weiffenbach, C. V.; Norton, S. A. Environ. Ini. 1982, 8, 59. (9) Kahlos, H.; Asikamen, M. Health Phys. 1980,39. 108. (10) Nazaroff. W. W, et al. "Potable Water as a Source of Airborne Radon-222 in U.S. Dwellings: A Review and Assessment," LBL 18154; Lawrence Berkeley Labora tory: Berkeley. Calif.. 1985. (11) Drinking Water and Health; National Academy of Sciences: Washington. D.C.. 1982. (12) Anderson. E. L. et al. Risk Anal. 1983,3, 277-95. (13) Van Ryzin, J. J. Occup. Med. 1980, 22, 321-26. (14) Van Ryzin. J.; Rai, K. In The Scientific Basis of Toxicity Assessment; Witschi, H., Ed.; Norlh-Holiand Biomedical Press: New York. N.Y.. 1980; p. 273. (15) Krewski. D.; Van Ryzin, J. In Statistics and Related Topics; Csorgo, M. et al., Eds.; North-Holland: New York, N.Y., 198J; p. 145. (16) Munro. T. C.: Krewski, D. R. Food Cosmet. Toxicol. 1981,19. 549-60. t C. Richard Cothem is a physicist with EPA. He received his Ph.D. in physics from the University of Manitoba. He is in volved in setting standards for radionu clides in drinking water and in assessing risks that result from drinking-water con taminants. r William A. Coniglio (l.) is a biologist with EPA. He has an M.S. in economic biology from Rutgers University and received pub lic health training at the University ofAla bama. He develops background informa tion on the occurrence of and exposure to drinking-water contaminants. William L. Marcus (r.) is a toxicologist with EPA. He received his Ph.D. in toxi cology and pharmacology from Howard University. He reviews and analyzes toxi cological reports and data and advises on toxicological matters. CIVIL ENGINEERS ENVIRONMENTAL ENGINEERS CHEMICAL ENGINEERS HYDROGEOLOGISTS Dames & Moore, one of the world's largest firms in engineering and applied earth sciences, is continuing to expand its practice in waste management and water resources engineering. Our Subur ban Washington. D. C. office has several immediate vacancies: Civil Engineers--1 to 5 years experience in all aspects of civil engineering with a concentration in water resources and hazardous waste management; Chemical and Environmental Engineers--2 to 15 years ex perience in hazardous waste remedial investigations and al ternative technologies studies, RCRA Part B permitting, and advanced wastewater treatment systems development; Hydrogeologists--2 to 10 years experience in design and review of groundwater sampling and monitoring programs, groundwater modeling and analysis of contaminant trans port. All positions require excellent communication skills. Hydrogeo logy positions require a minimum of a BS in geology or hydrogeo logy. Engineering positions require a minimum of BS in Engineer ing and P.E. or eligibility. interested candidates should send their resumes to: Dames& Moore 7101 Wisconsin Avenue Bethesda, MD 20814 Attention: Personnel Dept. CEM EOE/AA SL 034469 TANK LEAK? LINE LEAK? NO LEAK? GET THE ANSWER FAST, WITHOUT EXCAVATION! ET 4i. i.f1- < rin.^--| g Ask us lo prove our "edge'm accuracy, speed, trained per- '4ji sonnel, economy; national -^ program capability, state*' 3 -;;5.: of-art technology/ Meets NFPA 329 standards. 4nimi / / * ENVIRONMENTAL SERVICES, INC. f-t 115 DEWALT AVE. N.W.. CANTON, OHIO 44702 $ ^.^PP.~S^2^7QJ^r^jiJof2i6^53^jp(j)i^ggg^ ESST bdcu&o Estimating risk to human health Trichloroethylene in drinking water is used as the example ....... .. C. Richard Cothern William A. Coniglio William L. Marcus Environmental Protection Agency Office ofDrinking Water Washington, D.C. 20460 The decision to regulate the maxi mum level of an environmental contaminant allowable in air, water, or food is predicated on the ability of the substance to alter life processes and on the degree of exposure that creates a significant risk of adverse effects to a sensitive population. For more than four centuries, scientists have tried to understand the adverse effects of exposure to environ mental contaminants. In general, any contaminant can cause damage if exposure to it creates high concentrations within a cell, and if accumulated cellular dam age can result in disease or death. Envi ronmental regulations are developed to reduce exposure to a level that pre cludes such health effects. Drinking-water regulations are being developed for many contaminants, some of which are known to cause can cer in laboratory animals. One group, the volatile organic compounds (VOCs), is currently being assessed for the development of national drinkingwater standards. Reference 1 contains an explanation of the risk assessment studies being used to evaluate VOCs. 4 This article not subject to U.S. copyright. Published 1986 American Chemical Society ''*** 3 SL Q3A470J Environ. ScL "lechnoi.. Uol. 20. No. 2,1986 Jit; Levels of human exposure to VOCs found in drinking water and the level of cancer risk imposed by their presence have been estimated as part of the proc ess of developing regulations. Esti mates include the health benefits to be derived from different levels of national regulation. They couple the estimate of carcinogenic potency, derived from an extrapolation of a dose-response curve developed from high-level exposure of experimental animals, with estimates of the number of persons exposed to con taminated drinking water. The esti mates include the level of daily intakeoral, respiratory, and dermal--resulting from the presence of these chemicals in water (Figure 1). Estimates of the number of cancers resulting from currently unregulated conditions and of those cancers that will be prevented or allowed by setting standards at different levels can be in cluded in the regulatory process. Cur rent knowledge about the health effects of contaminants has its limits: Much is known about some aspects; little or nothing is known about others. But there is uncertainty in all areas. The population n.sk estimate is an as sessment of the total expected excess number of cancers in the United States during a lifetime from a given contami nant. This is an estimate and not a state ment of perfect knowledge. Therefore, a level of uncertainty or a range of val ues will be included to describe the esti mate. If the frequency distribution of the contributing factors controlling ex posure and toxicity were known, the propagation of errors could be deter mined by use of standard methods. Fre quency distributions often are not known, however, and a range of values often must be used to approximate the situation. This article describes some of the data on which regulatory judgments are made. It also describes the assumptions introduced to complete the assess ments, the results, and their uncer tainty. We take as our example the VOC trichloroethylene (TCE). Recent research has shown that TCE FIGURE 1 Risk estimation process v- ' ``'v ; r ;' Occurrence in drinking water :. 1 'i i ;II . .-'.'Vi::.- sr. Individual risk rate (ingestion, inhalation, dermal) Population I risk r Number of cancers averted by eliminating concentrations in drinking water that lead to a risk exceeding 10*\ 10_i, and 10` per lifetime, and the number of public drinking-water systems affected can be transformed in water to vinyl chloride, a known human carcinogen (2). The cancer risk associated with the presence of vinyl chloride in drinking water also has been assessed (7). TCE in drinking water There arc approximately 60.000 pub lic water supplies in the United States, serving some 220 million persons. Three major federally sponsored moni toring surveys have been conducted na tionwide to obtain information on the concentration of selected VOCs appear ing as pollutants in potable water pro vided to the nation's cities and towns. By using the survey data in conjunc tion with the multinomial approach for projecting national occurrence, it was estimated that 97% of groundwater sys tems of all sizes contain either no TCE or concentrations <0.5 /tg/L (Table 1). It is not possible to estimate how many systems in this category are contami nated with TCE. Of the estimated 1632 systems ex pected to have levels >0,5 jig/L, 421 (0.9% of the total number of groundwater systems) are expected to have concentrations >5.0/rg/L, and 133 systems (0.3%) nationally have con centrations >50/rg/L. Although a greater percentage of the large systems is expected to be contaminated, because of the large number of small groundwater systems it is likely that in abso lute numbers more small systems are contaminated. In a similar analysis of surface water supplies, it was estimated that 96% ei ther have no TCE present or that levels are <0.5 pg/L. It is estimated that 498 surface water systems have levels 20.5 pgIL (4.5% of total surface wa ter systems); 9 (<0.1%) are expected (population served) systems hi U.S. <0.5* 0.5-5 >5-10 >10-20 >20-30 >30- 25-100 : 19,125 18,506 465 101-500 . 1.15,674 15,166 381 501-1,000 ^ 4,877 4,719 118 -1.001-2,500. 4,257 107 2,501-3,300 V* 891 . 862 22 - 3,301-5,000 1,065 1,031 26. ' 5,501-10,000 , 1,168 1,130 28 10.001-25,000 535 775 34 25.001-50,000 - 290 269 12 50.001-75,000 64 59 , 3 75.001-100,000 : " 14' 13 . / >100.000 55 41 14 Total . 46,458 46,828 1,211 26 . 52 r-s 0 21 42 ;;o 7 13 6 26 21 .7 ,21 .o0;:.-,p;' o 8 12 ^0 6 I'". 2. - ' 2 3 3 0 .'Vi - 0 o i,.1, ,.SS-3Srt' 2 11 , , 4 A-.*'' 0 `0 r-, r: 8 v*f: 2 0 0 . ,:u o. . ..,, i- 0 o 0 00 0 ' 0 : ,, '_7-. U 0^ .. 0; 0 80 , 1323'0Vp{.;.4 "-'12 64 ' ... , .......... .................. Calculated as the difference between the number of systems expected to have i0.5 ug/L TCE and the total number of systems in that l category This group includes those having no contamination and those with levels <0.5 pg/L , .,, -?V-&&&*$$&? 112 Environ. Sei. Techno!., Vol. 20, No. 2,1986 SL 034471 to have levels > 5 /*g/L: none is esti mated to have TCE >40 jxg/L. Using the combined data from sur face water and groundwater supplies, it was estimated that nearly 190 million persons (88.3% of the population served by public drinking-water sys tems) receive uncontaminated water or water contaminated with TCE at levels <0.5/ig/L. Of the 25 million persons (11.7%) receiving water that contains TCE at levels >0.5 jzg/L, an estimated 1.8 million (0.9%) are exposed to lev els >5/ig/L. About 212.000 persons (0.1%) are probably exposed to levels >50/tg/L, and 42,000 (<0.1 %) are estimated to be exposed to levels >100Mg/L. Of the approximately 23 million per sons exposed to levels ranging from 0.5 fig/L to 5 ng/L, 76% obtain water from surface water supplies. However, of the 1.8 million persons exposed to levels >5 jtg/L, 62% use groundwater sources. All exposure to TCE in drink ing water in levels >40 /ig/L is pro jected to be from groundwater sources. Several assumptions have been made during the assessment of TCE contami nation of drinking water, among them that sites selected in surveys are repre sentative of drinking-water supplies across the nation and that grab samples are representative of water quality throughout the water supply system. Using scientific judgment, the uncer tainties in these estimates were ana lyzed. The uncertainties include those in the measurements themselves and those involved in projecting the existing data to estimate natural occurrence. In general, the overall uncertainty is less than an order of magnitude. Human exposure to contaminants The average adult in the United States uses 48-55 gal of water each day (3, 4). This means that a household of two children and two adults uses ap proximately 255 gal of water each day (J). Water pollutants pass through the home in water that is used for launder ing clothing, washing dishes, removing wastes, bathing, cooking, and drinking. The level of exposure to any pollu tant present in drinking water is the result of many personal daily choices and several factors over which we have very little direct control. Where we live, what we eat and drink, and how old we are all have a profound influ ence on the magnitude of exposure. Also, the physical and chemical charac teristics of the pollutant govern the amount that stays in the water and that which is transported into indoor air. Table 2 provides an example of the potential differences in exposure of TABLE 2 Comparative model of absorbed a volatile pollutant in drinking water* - r . ; i Absorbed doaapig/d) " ',~'v----R'r '* Exposure rouU pA." Formula-fad ^ Infant <4 kg) Fluid ingestion 80 Inhalation of ' enriched indoor air Ato Inhalation of W_v. enclosed shower air Dermal absorption ; \ Bathing ' Swimming i' <*? ' Ibrai absorbed dose $; },20itQ/kg/d Prefean (32 kg) Adult famale Adult mala* *- (80 kg) (70 kg) ,<* 200 j ' 100 10-300 50 * 300 200 ' '^7 Vir * ' - *-- 10-20 uglkqld 10 ^g/kg/d 7<<g/kg/d *ioohq/l '<' *Showsring adult male , Vi'". ' - sag ' >-4 family members to a VOC resulting from personal choices. This table has been prepared for a theoretical VOC, similar in chemical characteristics to TCE. We assumed that 100% of the compound is absorbed for all oral in take, 50% for respiratory intake, and a very small amount is absorbed in direct proportion to water flux across the skin. Cancer risks projected for TCE are based on assumed lifetime ingestion of 0.03 L of drinking water for each kilo gram of body weight and on the as sumption that 100% of the ingested chemical is absorbed. The factor for in gestion was developed using an intake standard of 2 1770 kg adult weight. In dividuals may experience many times this intake rate. Formula-fed infants and young children, for example, have average intake rates that are as much as eight times greater than those of aver age adults. Adults in tropical areas may consume twice as much liquid as the average, as may athletically inclined adults when engaged in strenuous physical activity. Persons who are ill also may consume much more water than the average. So cial behavior ritualized around the drinking of tea or coffee may lead to increased water consumption, although boiled water may contain smaller amounts of contaminants. A recent Ca nadian study shows that children under 5 years old, 10% of the children 6-17 years old, and 2% of the adult populaton consume more than 0.03 L/d of drinking water for every kilogram of SL 034472 Environ. Sci. Technol., Vol. 20, No. 2,1980 113 FIGUBE2 Bioassay data and model extrapolations for exposure to TCE through ingestion* Buutay 0*ta ar ttw starred point* tn the upper nght-hand cornet. The Wettme risk par parson exposed was calculated assuming that tha nsk due to inhalation and dermal exposure is equal to that due to ingestion exposure. The error introduced by this simplifying assumption i* less than tha widths of tha lines shown Through ingestion, mha&ion. ana dermal exposure, assuming that ingestion equals inhalation plus dermal body weight (<S). Each of the VOCs considered, in cluding TCE, has been shown to trans fer from water into air if the water is heated or aerated. Monitoring data in dicate that this process is continuous within the home and leads to an imme diate enrichment of respirable air at the point of water use and to a diffusion throughout the home. Showers taken within an enclosed bathroom, using 20-30 gal of water, result in the libera tion of all or a portion of the VOCs into the air. A person who showers inhales air containing a highly enriched level of the chemical while inside the bath room. VOCs from other sources as well mix with general home air, resulting in higher levels of the pollutants being relased into the indoor air. Very few data are available on the enrichment of general indoor air with VOCs. However, research has been conducted on the radon enrichment of indoor air. and it is assumed that the fraction of radon released is about the same as that for VOCs. These data indi cate that an enrichment of indoor air (the amount of radon in a liter of air) is about 10* to the concentration found in a liter of drinking water (7-10). This model assumes an indoor air change rate of one per hour. For our analysis it has been assumed that respiratory exposure to VOCs in drinking water is roughly equivalent to that from ingestion. To compare the rel ative exposures from inhalation and in gestion, assume that for both pathways 100% of the VOCs that enter the body reach the bloodstream. Then if water with a concentration of X/ig/L is in gested, the daily intake to the blood stream is 2X/*g/d. For example, X /xg/L x 2 L/d = 2X fig/d (1) If water with a concentration of X ng/L transfers all of the VOC to indoor air and it all goes to the bloodstream, then the intake to the bloodstream is again 2X Mg/d. For example, Xfxg 10~4 L (water) L L (air) 20 m3 (air) 106 cm3 (air) d m-1 (air) x L (air) = 2XMg/d 1000 cm3 (2) Under these conditions exposure from ingestion and inhalation is equal. Individual risk rates Estimates of individual risk rates are based on bioassay data derived from animal experiments. These data are converted to a continuous human equivalent exposure for use in risk esti mates. The bioassay data selected to es timate carcinogenic potency of VOCs in drinking water are from experiments TABLE 3 !' ' ' '' Population risk estimates for current TCE levels In drinking water i > -'VStT r- ,v .Stef; " ,,____ ... _____ . `-'I''.V' Total lifetime individual risk _________ for the mean concentration* concentration Number of persona kH'D_____________ served Law (probit) ' High (Wetbull) -- -:'3^^Ufetlnie"'' .v" v/r population 0.25 '' ' 1.9 x io* - <10-' 2.75 2.3 x 107 <10-, 7.5 4.3 x 10s ' <io-,B 3tV 15 .- ' ' 2.1 x 10s <10', 35 .^Tvv.'" 7.4 x 10s 7 x 10-* 45 - ; 2,6 x 10* 3 x 10-7 55 . / . 4.2 x 10* 4 x IO"7 . 75 1.3 x 10* 6 x 10"7 .> 100 , 4.2 x 10* > 1.2 x i(H ? Total* ' - ;J<vV 'r"-- . 2.4 x 10-* 7.3 x 10-* .1.3 X 10-* 1.7 x itr 2.3 x 10-* 2.6 x 10-* 2.8 x 10-3 3.2 x 10-3 3.7 x 10-J The total individual risk was determined by assuming that tha risk due to inhalation is equal to that due to ingestion `Bounded to one significant figure ' >.^<1-16.790 1-13.76072 1-416 1-155 . .iSl-IWWI ",V ,'.c .. f, 114 Environ. Set. Technol., Vol. 20, No. 2, 1986 SL 034473 reviewed and selected by the National Academy of Sciences and the EPA Car cinogen Assessment Group (II, 12). When the carcinogenic data were from inhalation experiments, they were con verted to ingestion values. The bioassay data arc fit to four ana lytical models: logit, multistage, pro bit, and Weibull. These four models are chosen somewhat arbitrarily and are representative of models currently in use. It is well-known that other models could be used to fit the same data (1316). There are no biologically based criteria for choosing one model over another. The models are in the form of analytical expressions that can be used to fit the high-dose data mathemati cally. The dose-response curves for the models are fit to equivalent human dose-response data converted using the relative surface area of animals and hu mans. The curves are extrapolated to the lower environmental levels. A loglog plot of the entire region shows the two nonzero data points for TCE (Fig ure 2). The solid lines are the point esti mates. and the error bars show the up per 95% confidence limits. The GLOBAL computer program was used for the multistage model, and the Krcwski computer program was used for the logit, probit, and Weibull models. The curves shown in Figure 2 are not the only models that could be fit to the data. Additionally, there is no assurance that the actual curve is in the range of those shown. It is possible, although highly unlikely, that the actual curve lies outside those shown. For example, the probit model generally decreases extremely quickly relative to dose, whereas the Weibull model decreases very slowly. Although the dose-response curves projected by each model in Figure 2 start at the same points, they diverge significantly at lower dose levels. At a concentration of 50 fig/L TCE in drink ing water, the Weibull model provides a risk estimate approximating 1 x 10'2, whereas the probit model provides an estimate of 10'10. These estimates pro vide a range of uncertainty equivalent to not knowing whether one has enough money to buy a cup of coffee or pay off the national debt. Population risk calculations The population at risk of health dam age due to TCE in drinking water was estimated. The population risks are de termined by a product of the number of people affected and the individual risk rate (Table 3). In each case the concen tration is in the average of the range. The lowest concentration (<0.5 ng/L) represents the minimum reported con centrations for the national surveys and does not necessarily represent the de tection limit of the chemical in drinking water. Although we have used the esti mated average of 0.25 jtg/L for this category, it is possible that all of the persons in this category could be ex posed to 0/ig/L or that all of them could be exposed to the maximum of 0.5 /xg/L. For each chemical the indi vidual lifetime risks were determined from the model fits. In each case the curves in Figure 2 giving the lowest and the highest indi vidual risks were used to determine the range of population risks. The uncer tainty in choice of model in Figure 2 is by far the largest uncertainty of any in volved in the estimates of risk. Thus it alone was used to calculate the range of risk. When calculations yielded less than one person, the number was listed as a whole number rather than as the fractional number of persons. The risk reduction analysis (Table 4) is a determination of the number of cancer cases averted if the standard is set corresponding to individual lifetime risk rates of 1CL4, 10"5, or 10`6. For these individual risk rates, the corres ponding range of concentration and the cases averted are listed. The concentra tion range was determined from the graph shown earlier in Figure 2. The population risk estimates are listed as ranges; the range is indicative of the uncertainty in this estimate. For TCE, the lower estimate for each TABLE 4 Risk reduction analysis* Maximum allowable drinking-water concentration (*g/i> .. Approximate Individual risk rata lor maximum -1 = c---o--n---c--e- --nt--rat*i--on - =: :/ ' .:i; Cumulativeiccaa*a*e*a fvfeidSS* "3^.. , , -a--v--e---r-ted" 100 i t x 10*7 to 1 x 10** jO-200 45 1 x 10-* to 1 x 10"a W^i-1.000 7.5 <i x i0',to6 x io- f .;<i-4.ooo ' 2,75 -:1-; <1 x IQ-'0 to 4 x IQ-3 ' Trlfeff< 1-20.000 ' For limiting TCE concentration In drinking water 'ssIIfestsiS 1 "Number of caaea averted lor concentration shown In first column f L'.'' -V,; y-.; concentration range is less than one. The sum of the upper population risk estimates is 70.000. when rounded off to one significant figure. The largest portion of the population risk is con tributed by the lowest concentration category. This category represents cal culations below the detection limit and may be an artifact of the mathematical analysis technique used to estimate the population risk. The range is largely a reflection of the uncertainty in choice of the mathematical model selected to estimate individual risk. The contribu tion to uncertainty from other factors is negligible by comparison. If a drinking-water standard--a max imum contaminant level--for TCE were established at 7.5 jtg/L, nation wide compliance would result in the avoidance of anywhere from < 1 to 4000 cases. Explanation The development of regulations cov ering contaminants in drinking water requires the assembly of data from many different sources. One important piece of information needed is the esti mate of the expected number of health effects and how many of these can be prevented by regulation. Risk assess ment usually involves using skimpy data concerning the levels of a contami nant in drinking water and tiding health effects data from animal experiments conducted at exposure levels much higher than those usually found in the environment. The risk can be estimated using TCE in drinking water as the example, al though the range of uncertainty is quite large. If there were no data, the range of estimated health effects would be be tween none and all the people. Using the available data for TCE in drinking water, the estimated range of health ef fects lies between < 1 and 70,000. If all exposure to levels >7.5 jxg/L were eliminated, as few as one or less of these health effects could be prevented, or perhaps as many as 4000. These esti mates are the best representation of the existing data for TCE in drinking water. Their value to the regulatory decision making process must be judged when placed alongside the other information used in that process. Acknowledgment We thank Joseph A. Cotruvo for his en couragement and helpful discussions on the work described here. The thoughts and ideas expressed in this article are those of the authors and are not necessarily those of the EPA. References (1) Cothern, C. R.; Coniglio. W. A.; Marcus. W. L. "Techniques for the Assessment of Carcinogenic Risk to the U.$. Population Due to Exposure from Selected Volatile Or SL 034474 Environ. Sci. Technol., Uol. 20. No. 2,1986 IIS ganic Compounds from Drinking Water via the Ingestion. Inhalation and Dermal Routes." EPA 570/9-85*001; EPA: Wash ington, D.C., 1985. (2) Vogel. T. M.; McCarty. R L. Appl, Envi ron. Microbiol. 1985,49. 1080-83. (3) Watson. K. S.; Farrell. R. P.; Anderson, J. S. J. Water Poilut. Control Fed. 1967, 30(12), 2039-54. (4) Lenaweaver, F. E. Jr. In "A Study of Flow Reduction and Treatment of Wastewater from Households"; Bailey. J.R.. Ed.; Gen eral Dynamics report to the Federal Water Quality Administration. Washington, D.C., 1969. (5) Bailey, J. R. et al. "A Study of Flow Re duction and Treatment of Waste Water from Households." Water Pollution Control Re search Senes II050FKE; General Dynamics report for the Federal Water Quality Admin istration; Department of the Interior: Cin cinnati, Ohio, 1969. (6) "Tap Water Consumption in Canada." 82EHD80; Canadian Environmental Health Directorate, Ministry of National Health and Welfare: Ottawa. Ont., 1981. (7) "Sources and Effects of Ionizing Radia tion," Report to the General Assembly of the United Nations by the United Nations Scientific Committee on the Effects of Atomic Radiation. 1977. (8) Hess, C. T.; Weiffcnbach, C. V; Norton, S. A. Environ. Int. 1982, 8, 59. (9) Kahlos, H.; Asikainen, M. Health Phys. 1980 39 108 (10) Nazaroff, W. W. et al. "Potable Water as a Source of Airborne Radon-222 in U.S. Dwellings: A Review and Assessment," LBL 18154: Lawrence Berkeley Labora tory: Berkeley, Calif., 1985. (11) Drinking Water and Health; National Academy of Sciences: Washington, D.C., 1982. (12) Anderson, E. L. et al. Risk Anal. 1983,3. 277-95. (13) Van Ryzin, J. J. Occup. Med. 1980, 22, 321-26. (14) Van Ryzin, J.; Rai, K. In The Scientific Basis of Toxicity Assessment; Witschi, H., Ed.; North-Holiand Biomedical Press: New York, N.Y., 1980; p. 273. (15) Krewski. D.; Van Ryzin, J. In Statistics and Related Topics; Csorgo, M. et al., Eds.; North-Holland: New York, N.Y., 1981; p. 145. (16) Munro, T. C.: Krewski, D. R. Food Cosmet. Toxicol. 1981, 19, 549-60. C. Richard Coihern is a physicist with EPA. He received his Ph.D. in physics from the University of Manitoba. He is in volved in setting standards for radionu clides in drinking wafer and in assessing risks that result from drinking-water con taminants. William A. Coniglio (I.) is a biologist with EPA. He has an M.S. in economic biology from Rutgers University and received pub lic health training at the University ofAla bama. He develops background informa tion on the occurrence of and exposure to drinking-water contaminants. William L. Marcus (r.) is a toxicologist with EPA. He received his Ph.D. in toxi cology and pharmacology from Howard University. He reviews and analytes toxi cological reports and data and advises on toxicological matters. CIVIL ENGINEERS ENVIRONMENTAL ENGINEERS CHEMICAL ENGINEERS HYDROGEOLOGISTS Dames & Moore, one of the world's largest firms in engineering and applied earth sciences, is continuing to expand its practice in waste management and water resources engineering. Our Subur ban Washington, D. C. office has several immediate vacancies: Civil Engineers--1 to 5 years experience in all aspects of civil engineering with a concentration in water resources and hazardous waste management; Chemical and Environmental Engineers--2 to 15 years ex perience in hazardous waste remedial investigations and al ternative technologies studies. RCRA Part B permitting, and advanced wastewater treatment systems development; Hydrogeologists--2 to 10 years experience in design and review of groundwater sampling and monitoring programs, groundwater modeling and analysis of contaminant trans port. All positions require excellent communication skills. Hydrogeo logy positions require a minimum ol a BS in geology or hydrogeo logy. Engineering positions require a minimum of BS in Engineer ing and P.E. or eligibility. Interested candidates should send their resumes to: Dames& Moore 7101 Wisconsin Avenue Bethesda, MD 20814 Attention: Personnel Dept. CEM EOE/AA TANK LEAK? LINE LEAK? NO LEAK? GET THE ANSWER FAST, WITHOUT EXCAVATION! accuracy, speed, traln^pef 1 sonnet, economy; natfona/M program capability, state-- * tvxjdf-art technology?^ Meets NFPA329standards^ leakjokator environmental SERVICES. INC.J, 115 DEWALTAW.Vt:i^t^OtCrOHl6"i^# SL 034475