Drinking water standards and risk assessment

REGULATORY

TOXICOLOGY

ANDPHARMACOLOGY

8,288-299

Drinking Water Standards

(1988)

and Risk Assessment’

JOSEPH A. COTRUVO O&e of Drinking Water, U.S. Environmental Protection Agency, Washington, D.C. 20460

Received April Ii, 1988

The role and use of risk assessment methods in the establishment of drinking water standards are described with emphasis on recent applications. The process essentially includes an attempt to quantify human exposure from all routes including drinking water, animal toxicology, and human epidemiology, when available, to arrive at drinking water concentrations at which exposure would result in “no known or anticipated adverse effectson health, with a margin of safety.” The process itself is straightforward; however, the application to decision making for substances that are considered to be potentially nonthreshold acting in their toxicity (e.g., carcinogenic) requires many policy choices beyond the scientific data and is subject to considerable controversy.

INTRODUCTION Risk avoidance or minimization certainly should be principal elements in the determination of drinking water standards, but they may not be the only elements. Technological and economic factors also enter into the ultimate definitive quality parameters for public drinking water supplies. Anesthetic factors of taste, odor, and appearance must be important considerations even if they do not relate to the safety of the water. Consumer acceptance and confidence in the quality and safety of drinking water are important in developed countries where consumers have options, and, through the political process, they can ensure their demands for water quality be met. The establishment of drinking water standards can have a significant impact in the broad arena of environmental policy and protection well beyond just the immediate application to public water supplies. The case could be made that drinking water quality is the pivotal raison d%tre for all programs aimed at protecting the water environment, since drinking water could be the principal human exposure vector for many substances being controlled. Those environmental programs could include surface water industrial discharge controls, domestic sewage treatment requirements ’ Presented at the Institution of Water and Environmental Management Symposium, March 1988, and the International Society of Regulatory Toxicology and Pharmacology Annual Meeting, December lo1 I, 1987, Arlington, VA. The opinions expressed in this paper are those ofthe author. They do not necessarily reflect the position of the USEPA. 288

0273-2300188 $3.00

DRINKING

WATER

STANDARDS

AND RISK ASSESSMENT

289

(location and extent), siting and controls on subsurface disposal of hazardous wastes, standards for cleanup of ground waters or surface waters contaminated by hazardous chemical wastes, disposal of radioactive wastes, and even agricultural practices (fertilizer and pesticide use) where ground and surface waters could be affected. Thus, drinking water standards in their broadest application could effectively predetermine many of the political, economic, and technological decisions in those seemingly remote environmental programs, and they could, to a considerable degree, be the logical philosophical basis for all of them which are intended to protect health of humans who might be exposed to those drinking water contaminants which had originated from those sources. Risk assessment is fundamentally an attempt to quantify the possible health consequences of human exposure in particular circumstances. In the case of drinking water the conclusion would be expressed in terms of the probability (within specified levels of uncertainty) of cases of adverse effects (e.g., fatalities) in the reference population group: for example, an incremental upper bound risk of bladder cancer of one per million ( 1OP6) in a population typically consuming 2 liters of drinking water per day for 70 years. The lower bound risk might be from one per billion ( 10e9) to zero. All of these computations and “conclusions” are limited in their reliability and credibility by the quality of the exposure and toxicological data, the mathematical expressions used, and the lack of scientific understanding of the mechanisms of carcinogenesis operative at low environmental doses in genetically diverse humans, as opposed to the high doses to which test animals are exposed. In addition, the question of lowdose interaction is a complete unknown. RISK

ASSESSMENT

FOR DRINKING

WATER

CONTAMINANTS

In its lowest terms a risk assessment could be represented as follows: RA = (dose distribution)

X (number of persons exposed at each dose) X (risk per dose) X (time).

The basic information required to perform a qualitative and quantitative risk assessment includes quantitative information on (1) the occurrence, (2) human exposure, and (3) toxicology of the substance. Although methodologies are available to quantify each of these factors, in practice, data limitations and analytic complexities often lead to many simplifying assumptions. Comprehensive quantitative data on the frequency and concentration range of the contaminant in public drinking water supplies are a basic necessity in order to determine the potential for human exposure under a variety of conditions and to predict the exposure consequences of any control options being considered. The same analyses should be done for food, air, occupational, and other contributions to total human exposure. Typically, drinking water data are the most feasible to obtain, given the finite nature of drinking water sources and the availability of analytical methods of high sensitivity and reliability. Statistically based surveys can be designed to generate highquality data on national distributions of drinking water contaminants by source type, population, season, or other variables. Food contributions are more difficult for many substances due to analytical difficulties and dietary variability. Ambient and indoor

290

JOSEPH A. COTRUVO

air data are limited for most substances, and data on contributions from occupation, smoking, or other sources are also limited. Computing human exposure from occurrence data requires detailed information on water and food consumption patterns and other life-style factors that often are very difficult to model. These would be age-, size-, season-, and location-dependent. Water consumption has been studied in several countries and reasonable distributional data are available. For example, the average drinking water consumption estimated from eight studies was 1.63 liters per day. A recent dietary study (Ershow and Cantor, 1988) concluded that the median daily water consumption in the United States was 1.2 to 1.4 liters per day and that 80 to 85% consumed less than 2 liters, and about 1% consumed more than 4 liters per day. This included all tap water including coffee, tea, and reconstituted juices, soups, and food water (e.g., from rice). Dietary patterns are, however, much more complex and data bases amenable to extrapolation to populations are not very extensive. Localized ambient air inhalation data are available for a few substances. Indoor air quality data are potentially of greatest interest but also limited. Water can also contribute to indoor air exposure to volatile substances such as trihalomethanes or radon, or even Legionelh organisms. This indirect exposure should also be considered when projecting total exposure and the drinking water contribution. Toxicology assessments require highly qualified health scientists to examine and weigh the complex data describing the toxicology of a substance, select the appropriate valid studies from the often conflicting or incomplete data, and arrive at a judgment on the relevance of the information to human health risks. The analyses should include all relevant toxicology, including whole animal acute, subchronic (90-day), and chronic (lifetime) studies, reproductive and developmental studies, neurotoxicology and other relevant animal data, in vitro studies of mutagenesis and cytogenetics, and human epidemiology if available. The objective is to have at hand a comprehensive dossier of the toxicology, then identify the endpoints that are the most sensitive and significant for the sensitive groups and then calculate the appropriate “safe” exposure levels, commonly called the acceptable daily intake (ADI) (or the risk), and also called the reference dose (RfD). This, in conjunction with the occurrence and human exposure, constitutes the basis for the risk assessment. SAFETY

AND

RISK

DETERMINATION

FOR CHEMICAL

AGENTS

Toxicity has been defined as the intrinsic quality of a chemical to produce an adverse effect. The toxicology of chemical substances found in drinking water is commonly divided into two broad classes: (1) acute or chronic toxicity and (2) carcinogenicity. The same substance may be capable of causing classic toxic effects and imparting risks of carcinogenicity. The distinguishing characteristic between these categories of effects lies (1) in the probably unverifiable assumption that dose thresholds exist for chronic toxicity effects and (2) in the also unverifiable assumption that dose thresholds do not exist (or have not been demonstrated) for carcinogenic effects. In the case that dose thresholds exist for chronic toxicity effects, the nominal basis for standard setting is to achieve a total daily dose of the substance that is with practical certainty below the level at which any injury would result to any individual in the population. For toxicants assumed to be acting by nonthreshold mechanisms, it fol-

DRINKING

WATER

STANDARDS

291

AND RISK ASSESSMENT

lows that some finite risk may exist at any nonzero dose level. Thus, standard setting objectives range from zero, which is not quantifiable and often not practically achievable, to a daily dose level that contributes only a negligible theoretical incremental increase in the lifetime risk of the effect to individuals and/or the population exposed. NONCARCINOGENIC

EFFECTS:

SAFETY

FACTORS

Numerous substances detected in drinking waters are known to induce toxicity but usually at dose levels much higher than those found in water. When appropriate data are available from human epidemiology or animal studies, the use of the AD1 concept is a well-accepted procedure for determining concentration levels for standard setting. The AD1 of a chemical is defined as the dose that is anticipated to be without lifetime risk to humans when taken daily. The AD1 does not guarantee absolute safety, however, and it is not an estimate of risk. The assumption of one threshold for all individuals in a large population is simplistic; the population is genetically heterogeneous with a varied history of exposure, prior disease states, nutritional status, and stresses. Thus, it is likely that each individual has a unique threshold, and certain individuals in the population will be at inordinately high risk, whereas others may be at very low risk. The AD1 is usually derived from a detailed analysis of the toxicology of the chemical being examined. The no observed adverse effect level (NOAEL) is determined for the most sensitive adverse effect in the test system (usually animals but occasionally humans), and a safety or uncertainty factor is applied to the NOAEL dose to derive the safe level for the general human population. The AD1 is computed by multiplying the experimental NOAEL (in milligrams per kilogram per day) by the reference weight of a typical adult (70 kg) and dividing by the safety (uncertainty) factor: AD1 (mg/person/day)

= NOAEL

(mg/kg/day) X 70 (kg/person)/safety

(uncertainty)

factor.

Because an AD1 is intended to account for total daily intake of the toxicant, inhalation and food intake as well as water should be accounted for when attempting to arrive at the maximum drinking water level or the adjusted AD1 for drinking water at the maximum drinking water level considering only health factors. Thus, in the optimum case when such information is available, the daily uptake from air and the daily intake from food (if 100% uptake is assumed) should be subtracted from the ADI. Finally, for the determination of the acceptable drinking water concentration value, the assumption in the United States is that adults consume 2 liters of water per person per day; thus, the final value should be divided by a factor of 2: drinking

water target (mg/liter)

= AD1 (mg/day) - inhalation

(mg/day)

- food (mg/day)/(2

liters/day)

Figure 1 is a general illustration of a process for the calculation of an AD1 for a particular substance. The solid line to point A is the dose-response curve determined by the multiple-dosing experiment. Point A is the highest no observed effect level in milligrams per kilogram per day for the most sensitive adverse endpoint that was

292

JOSEPH A. COTRUVO

FIG. 1. General illustration of a process for the calculation of an ADI for a particular substance.

determined from the animal multiple-dose chronic study. Points B, D, and E are the presumed thresholds for the adverse effect in the human population if the extrapolated dose-response curves AB, AD, or AE are correct. Point C would be the AD1 concentration determined by application of the selected safety (uncertainty) factor to the dose at point A. Because lines AB, AD, and AE are extrapolations, the true doseresponse curve in the range of concern is unknown; thus, any of the curves could be correct in a given case. The intention of the standard setter is that AB would be the true curve because the no (actual) response dose would be greater than the calculated AD1 value C; thus, the safety (uncertainty) factor was chosen appropriately. However, if AD or AE were the true dose-response curve, then the calculated AD1 was too large; thus, the safety (uncertainty) factor was too small, and some members in the human population might suffer the adverse effect. AE indicates a nonthreshold doseresponse. The size of the gap between C and B is also of interest because if it were excessively large, overregulation could result in excessive control expenditures without any benefit. The value of an AD1 is entirely dependent on the quality of the experimental data and the judicious selection of the safety (uncertainty) factor, which is entirely judgmental. Among the factors influencing the quality of the experimental data, beyond the mechanics, are the selection of the appropriate animal model as the human surrogate, the number of animals at each dose and the number and range of the doses for acceptable statistical significance of results and shape of the experimental curve, the actual detection of the most sensitive adverse effect (which could be only biochemical change or frank organ damage), the length of the study (lifetime studies versus shorter-term studies), and the appropriate route of exposure (inhalation, gavage, ingestion in food or water, etc.). The quality of the experimental evidence determines the magnitude of the safety (uncertainty) factor to be applied. SAFETY

(UNCERTAINTY)

FACTORS

The safety factor is a number that reflects the degree or amount of uncertainty that must be considered when experimental data are extrapolated to the human population. When the quality and quantity of dose-response data are high, the uncertainty

DRINKING

WATER

STANDARDS

AND

RISK

293

ASSESSMENT

factor is low; when the data are inadequate or equivocal, the uncertainty be larger. The following general guidelines have been adopted by the National Sciences (NAS) Safe Drinking Water Committee, and they are also USEPA in the development of drinking water standards and guidelines advisories.

factor must Academy of used by the and health

( 1) 10 Factor: Valid experimental results from studies on prolonged human ingestion with no indication of carcinogenicity. (2) 100 Factor: Experimental results of studies of human ingestion not available or scanty. Valid results from long-term feeding studies on experimental animals or, in the absence of human studies, on one or more species. No indication of carcinogenicity. (3) 1000 Factor: No long-term or acute human data. Scanty results on experimental animals. No indication of carcinogenicity. The NAS also examined the application of quantitative models such as log probit and log logistic for human risk assessments for noncarcinogenic substances but found these to be of limited value for contaminants in drinking water. These models could be used to estimate the risk of a toxic effect, but they require data from lifetime feeding studies with sufficient numbers of animals and with a demonstrated dose-response. Data of this type are seldom available; thus, the NAS concluded that the ADI approach is most useful at this time. However, in those cases in which such data can be obtained, a risk estimate approach can be employed. RISKS

FROM

POTENTIAL

NONTHRESHOLD

TOXICANTS

In 1977, the NAS Safe Drinking Water Committee outlined four principles that it said should be useful in dealing with the assessment of hazards that involve chronic irreversible toxicity or the effects of long-term exposure. These principles (paraphrased as follows) were intended to apply primarily to cancer risks from substances whose mechanisms involve somatic mutations and may also be applicable to mutagenesis and teratogenesis: (1) Effects in animals, properly qualified, are applicable to man. Large bodies of data indicate that exposures that are carcinogenic to animals are likely to be carcinogenic to humans, and vice versa. (2) Methods do not now exist to establish a threshold for long-term effects of toxic agents. Thresholds in carcinogenesis that would be applicable to a total population cannot be established experimentally. (3) The exposure of experimental animals to toxic agents in high doses is a necessary and valid method of discovering possible carcinogenic hazards in humans. Only dosages that are high in relation to expected human exposures must be given to animals under the experimental conditions that are used because there is no choice but to use numbers of animals that are small relative to exposed human populations, and then to use biologically reasonable models in extrapolating the results to estimate risk at low doses.

294

68

JOSEPH A. COTRUVO

10-l

r

TRICHLOROETHYLENE uwestlon)

CONCENTRATION

IN DRINKING

FIG. 2. Sample risk extrapolations for trichloroethylene Cothem).

WATER(micrograms/liter)

using four different models (calculations by R.

(4) Material should be assessed in terms of human risk rather than as safe or unsafe. Extrapolation techniques may permit the estimation of upper limits of risk to human populations. To do so, data are needed to estimate population exposure; valid, accurate, precise, and reproducible animal assay procedures are required, and appropriate statistical methods are necessary. Decisions cannot involve merely risk; benefit evaluations should include the nature, extent, and recipient of the benefits. It is often necessary to accept risks when the benefits warrant the risk, but risks imposed on persons who gain no benefits are generally not acceptable. RISK

EXTRAPOLATION

Numerous mathematical models have been developed in attempts to estimate potential risks to humans from low-dose exposures to carcinogens. Each model incorpo rates numerous unverifiable assumptions. Lowdose calculations are highly model dependent, widely differing results are commonly obtained, and none of the models can be firmly justified on either statistical or biological grounds. Thus, the decision to use this approach and the choice of how to do the calculations are matters of judgment. Among the choices that the decision makers must consider are which model(s) to employ, which assumptions to incorporate, and which acceptable risk to allow. See Fig. 2 for an example of the model dependency of risk calculations.

DRINKING

WATER

STANDARDS

295

AND RISK ASSESSMENT

IDENTIFICATION OF COMPOUNDS LIKELY CARCINOGENIC TO HUMANS

TO BE

The fundamental question of risk assessment for potential human carcinogens requires definition of substances that exceed an evidentiary threshold. Once the scientific evidence establishes a substantial basis for conclusion of known or potential human cancer, it is then in order to determine a procedure for risk quantification. Quantitative risk assessments must always be read with the qualitative evidence of the likelihood of carcinogenicity. The International Agency for Research on Cancer (IARC) has provided guidelines for assessing the epidemiological and animal toxicological data base leading to a conclusion of the strength of the evidence of carcinogenicity of numerous substances. USFPA has recently proposed a similar approach with some added refinement. USEPA

QUALITATIVE

ASSESSMENT

OF CARCINOGENS

The USEPA applies a qualitative weight of evidence scheme in assessing the potential for a contaminant to increase the risk of cancer in humans. This scheme consists of five groups: Group A-Human

Carcinogen: Sufficient evidence in humans.

Group B-Probable Human dence in humans but sufficient Group C-Possible Human in the absence of human data. Group D-Not Classifiable available. Group E-No Evidence of least two species. USEPA’S

Carcinogen: Limited evidence in humans or no evievidence in animals. Carcinogen: Limited or equivocal evidence in animals as to Human Carcinogenicity

THREE-CATEGORY

Carcinogenicity: for Humans:

APPROACH

Inadequate

or no data

Negative evidence in at

FOR SETTING

MCLGs

Maximum contaminant level goals (MCLGs) are nonenforceable health goals that serve as the target for setting the enforceable drinking water standard, the maximum contaminant level (MCL). The Office of Drinking Water (ODW) employs a threecategory approach for setting MCLGs, based on the qualitative carcinogenicity classification scheme listed above. Category I: Group A and B substances: Goal equals zero as an aspirational goal. Category II: Group C substances: Goal equals lop5 to 10e6 excess cancer risk, or goal equals reference dose (ADI) value with an additional safety factor applied to allow for an adequate margin of safety due to uncertainties in the substances’ carcinogenic potential to humans. Category III: Group D and E substances: Goal calculated by the AD1 approach with a portion allocated to drinking water.

296

JOSEPH A. COTRUVO TABLE 1 VOCs:

Trichloroethylene Carbon tetmchloride Vinyl chloride 1,2-Dichloroethane Benzene para-Dichlorobenzene 1,l Dichloroethylene 1, 1,l -Trichloroethane

U.S. NATIONAL

FINAL

MCLGs

AND

MCLs

(IN

maliter)

EPA ranking

EPA category

Final MCLG

Final MCL

B2 B2 A B2 A C C D

I I I 1 I II II III

Zero Zero Zero Zero Zero 0.075 0.007 0.2

0.005 0.005 0.002 0.005 0.005 0.075 0.007 0.2

DRINKING

WATER

REGULATIONS

The U.S. Safe Drinking Water Act requires that EPA set drinking water standards in two steps after it makes a determination that a substance “may” have any adverse effect on health. First, it must determine the MCLG, the level in drinking water that would result in “no known or anticipated adverse effect on health” with a margin of safety. Second, the MCL is the enforceable standard to be set as close as feasible to the MCL taking costs into consideration. The statute does not formally describe that a risk assessment is part of the process and the final standards are technology based, not risk/benefit or cost/benefit based.

VOLATILE

SYNTHETIC

ORGANIC

CHEMICALS

In July 1987, EPA established drinking water standards for eight volatile synthetic organic chemicals (VOCs) (see Table 1). Five chemicals, benzene, vinyl chloride, carbon tetrachloride, 1,2-dichloroethane, and trichloroethylene were regulated as known or probable human carcinogens. Two, para-dichlorobenzene and 1,l dichloroethylene, were regulated as having equivocal evidence of carcinogenicity, and one, 1, 1,l -trichlorethane, was regulated as a noncarcinogen. MCLGs for threshold toxicants were set on the classical basis of determining a NOAEL and allocation of a certain portion of daily intake to drinking water, assuming consumption of 2 liters per day of drinking water. Thus, 1, 1,l -trichloroethane, which was in Category III, was given an MCLG of 0.2 mg/liter based upon a safety factor of 1000 applied to the NOAEL, and the MCL was also 0.2 mg/liter since the MCLG is the floor value even though it is technically feasible to achieve much lower levels. The evidence for a compound to be considered in the “equivocal” evidence of carcinogenicity Category II, Group C, would be limited as follows: (a) the studies involve a single species, strain, or experiment;

or

DRINKING

WATER

STANDARDS

AND

RISK

ASSESSMENT

297

(b) the experiments were restricted by inadequate dosage levels, inadequate duration of exposure or inadequate follow-up, poor survival, too few animals, or inadequate reporting; or (c) an increase in benign tumors only. Category II, Group C, substances includes those substances where some limited but insufficient evidence of carcinogenicity exists from animal data. MCLGs should reflect the fact that some evidence of carcinogenicity has been reported. Two approaches were used to set the MCLGs-either (a) setting the goal based upon noncarcinogenic endpoints (the ADI) then applying an additional uncertainty (safety) factor of up to 10, or (b) setting the goal based upon a nominal lifetime risk calculation in the range of 10e5 to 10e6 using a conservative calculation model. The first approach is preferred; however, the second is used when valid noncarcinogenicity data are not available and adequate experimental data are available to perform the risk calculation. The most controversial decision related to the establishment of nonregulatory goals for substances to be regulated based upon the risk of human carcinogenicity (i.e., Category I, Groups A and B). Three options were considered as interpretations of the statutory directive (“no known or anticipated adverse effect”) and a brief statement in the legislative history that MCLGs for non-threshold activity substances should be zero. Those three options were ( 1) MCLG = zero, (2) MCLG based upon a target calculated risk, and (3) limit of analytical detection. Little support for the analytical detection limit approach was received. The choice between the zero option and the risk option was ultimately based upon the statutory direction and the legislative history. The basic question was whether it was appropriate as an “ideal” to set a target that theoretically permitted some number of cancer deaths. The “ideal,” of course, would be zero cancers. MCLGs are “ideal” or “aspirational” goals, not standards. EPA also pointed out that setting an MCLG at zero did not imply that actual harm would occur at levels somewhat above zero. ENFORCEABLE

NATIONAL

DRINKING

WATER

STANDARDS

(MCLs)

MCLs are set as close to the MCLGs as is feasible. This is not difficult for Category II and Category III substances which have finite goals. The dilemma occurs for Category I substances with zero MCLGs. Theoretically, as what is technically feasible changes over time, the MCLs would also have to be changed. MCLs were determined based upon an assessment of a variety of factors including the availability and performance of granular activated carbon and packed tower aeration and their costs in a variety of water system conditions, the number of supplies affected, total national costs, and the reliability of analytical methods. Then EPA also examined the nominal residual lifetime risks that were theoretically associated with exposures at the technologically determined MCLs. BEST AVAILABLE

TECHNOLOGIES

Granular carbon (GAC) and packed tower aeration (PTA) were determined to be the best available technologies based upon their demonstrated ability to remove 90 to 99% of the volatile synthetic organic chemicals in field tests.

298

JOSEPH

UNIT

A. COTRUVO

COSTS

Cost estimates ranged from $0.10 to $0.85 per 1000 gallons of production and $0.05 to $0.30 per 1000 gallons for PTA. TOTAL EPA estimated that 1300 have to install treatment at liter. MCLs at 0.00 1 mg/liter It concluded that the latter national costs.

for GAC

COSTS

community water suppliers in the United States would a total cost of $280 million for the MCLs at 0.005 mg/ would involve 3800 systems at a cost of $1300 million. costs were not warranted considering the incremental

ANALYTICAL

CONSIDERATIONS

The practical quantitation levels (PQLs) were defined be reliably achieved within specified limits of precision laboratory operations. These were +40% of true value 0.0 10 mg/liter and +20% of true value for concentrations RISK

as the lowest levels that can and accuracy during routine for concentrations less than greater than 0.0 10 mg/liter.

CONSIDERATIONS

Alter the above conclusions were considered, the MCLs for the probable carcinogens (except vinyl chloride) were set at 0.005 mg/liter and established for most of the VOCs in this group. EPA then examined the putative risks at the MCL levels to determine whether they would be acceptable from a safety standpoint. A target reference risk range of 1Oe4to 1Oe6was considered to be safe and protective of public health, when calculated by a typically conservative linear multistage model. This is consistent with the concept expressed in the World Health Organization 1984 Drinking Water Guidelines, which selected 10v5 as a general guideline value for “carcinogens” and then explained that the application could vary within by a factor of 10 on either side, i.e., 10m4to lo-?

Trichloroethylene Carbon tetrachloride 1,2-Dichloroethane Benzene Vinyl chloride

MCL (mg/liter)

1o-4 to 1o-6 range (mg/liter)

0.005 0.005 0.005 0.005 0.002

0.3 to 0.003 0.03 to 0.0003 0.04 to 0.0004 0.1 to 0.001 0.2 to 0.002 (0.002 to 0.00002)

One of the reasons for the downward adjustment of the MCL for vinyl chloride was the higher unit risk obtained from a more conservative assessment of the animal toxicology.

DRINKING

WATER

STANDARDS

AND RISK ASSESSMENT

299

CONCLUSION In the United States, risk assessment is an integral part of the regulatory decision process particularly in the qualitative determination of the strength of evidence relating to carcinogenicity and the classification within the EPA ranking system. That then leads to an “aspirational” MCLG of zero for “probable” carcinogens and nonzero values based upon classical toxicology for “noncarcinogens” and a related system for “equivocal” evidence substances involving either additional safety factors or a nonthreshold risk model calculated target. Legally enforceable drinking water standards (MCLs) are required to be set as near as technically and economically feasible to the MCLGs. For “noncarcinogen” and “equivocal” evidence substances the MCL in usually the same as the MCLG. For “probable carcinogens” the MCL is set based on a variety of technological performance/cost factors, but also a “reference risk” rank is targeted between 1Oe4and 1Oe6 (incremental lifetime risk using a conservative model unlikely to have underestimated the risk). Standards falling in that range are concluded to be safe and protective of public health. REFERENCES COTRUVO, J. A. (1985). Organic micropollutants in drinking water: An overview. In Organic Micropollutants in Drinking Water (H.A.M. de Kruijf and H. J. Kool, Eds.). Elsevier, Amsterdam.

COTRUVO, J. A. (1987). Risk assessment and control decisions for protecting drinking water quality. In Organic Pollutants in Water, Advances in Chemistry Series, No. 214 (I.H. Suffet and M. Malaiyandi, Eds.). Amer. Chem. Sot., Washington, DC. Guidelines for carcinogen risk assessment. Fed. Regist. 51,33,992 (September 24, 1986). National Academy of Sciences (NAS) (1975). Principles for Evaluating Chemicals in the Environment. Washington, DC. National primary drinking water regulations; Volatile synthetic organic chemicals. Fed. Regist. 50,(2 19), 46,880-46,900 (November 13, 1985). National primary drinking water regulations; Volatile synthetic organic chemicals, Final Rule. Fed. Regist. 52( 130), 25,690-25,717 (July 8, 1987). The Safe Drinking Water Act, Public Law 99-339, June 19, 1986.