CONCLUSIONS FOR SAFE REGULATION OF DRINKING WATER

What can we conclude about the adequacy of the U.S. EPA’s current drinking water standard for arsenic?

The present EPA drinking water standard, as an enforceable Maximum Contaminant Level (MCL), is 50 micrograms of arsenic per liter water (50 µg/L, equivalent to 50 parts per billion, or ppb). This value has not changed since 1942, and was promulgated with few scientific underpinnings. There is therefore little scientific support for its regulatory adequacy. This MCL was issued before the accumulation of the large body of scientific and human health data produced over the last 30 to 40 years, a period that included the Taiwanese studies and numerous authoritative treatises on arsenic, including some from the NAS and EPA. As long ago as 1962, the U.S. Public Health Service recommended that water containing more than 10 µg/L (or ppb) of arsenic (one-fifth of the still-current standard) should not be used for domestic supplies.

Congress has directed EPA to update the 1942 arsenic standard three times -- in 1974, 1986, and 1996. A court ordered EPA to complete this task in the early 1990’s, but several extensions were granted. EPA still has not updated the standard. In a legislative mandate in the Safe Drinking Water Act Amendments of 1996, Congress again directed EPA to publicly propose an updated arsenic standard based on current evidence by January 1, 2000, a deadline that EPA has now, again, missed. EPA is then required to promulgate the final arsenic standard by January 1, 2001.

The current scientific and health risk assessment status of arsenic within that mandate makes it clear that EPA’s current MCL of 50 µg/L is grossly inadequate for protecting public health. The extent of that inadequacy is effectively captured in the NAS report, Arsenic in Drinking Water (NAS, 1999). The report focused heavily on risk assessment estimates for human cancer frequencies as a function of drinking water and food arsenic and derived cancer risks for arsenic in environmental media, particularly drinking water. Our analysis concurs strongly with the academy’s findings and recommendations as well as the following conclusion:

On the basis of its review of epidemiological findings, experimental data on the mode of action of arsenic, and available information on the variations in human susceptibility, it is the subcommittee’s consensus that the current EPA MCL for arsenic in drinking water of 50 µg/L does not achieve EPA’s goal for public-health protection and, therefore, requires downward revision as promptly as possible (NAS, 1999, pp. 8-9).

The NAS report did not recommend a specific MCL below 50 that would be fully health protective. It did, however, provide a series of cancer risk assessments for cancers of the skin and internal organs. This approach for bladder and lung cancers employed the traditional straight-line extrapolation from rates at elevated arsenic exposures. Put differently, the NAS assumed -- as is usually assumed by scientists based on traditional principles of toxicology, unless there is strong evidence to the contrary -- that there is a direct, linear relationship between cancer risk and arsenic exposure. The academy committee members, correctly and conservatively (with respect to the best health protection), noted that low-dose extrapolation models based on available data may or may not be "sublinear" compared to linear extrapolation. That is, arsenic at extremely low doses may, or may not, cause relatively less cancer risk per microgram than it does at high doses. However, the NAS experts concluded, the evidence for such "non-linear" models of arsenic-associated cancer risk is not compelling enough to rule out the traditional linear approach, so the health-protective linear approach should be used. The NAS scientists then used studies of people who had been exposed to arsenic in their tap water at elevated levels (for example in Taiwan) to model, or estimate, the risks of people exposed to lower levels.

The 1999 NAS report calculated that arsenic consumption in drinking water at the current EPA MCL would produce a male fatal bladder cancer lifetime risk of 1 per 1,000 to 1.5 per 1,000, using a linear extrapolation approach. Factoring in lung cancer risk and its relative robustness compared to bladder cancer (lung cancer risk is about 2.5 times greater than bladder cancer risk), an overall internal cancer risk rate "could easily result in a combined lung cancer risk" of 1 percent, or 1 in 100, according to the NAS’s 1999 report (p. 8). The high level of cancer risk from arsenic ingestion in water at the present MCL does not account for concurrent intakes of carcinogenic arsenic from food or idiosyncratic sources (for example, certain prepared ethnic remedies that contain arsenic). In the past, EPA estimated a lower cancer risk from arsenic in tap water than did NAS in 1999. For example, EPA’s Integrated Risk Information System (EPA, 1998) estimated about a 10-fold lower cancer risk for arsenic than the more recent NAS study (NAS, 1999), apparently in part because EPA evaluated only bladder cancer risks, whereas NAS considered the higher risk of lung cancer as well, based on recent studies. We believe the NAS risk estimates are more reliable and should be adopted by EPA.

The lifetime risks of dying from internal cancers due to drinking water arsenic estimated in this paper based on linear extrapolations in this paper from the NAS 1999 arsenic report are generally supported by studies of people drinking relatively low levels of arsenic in their tap water. For example, a recent study from Finland (Kurttio et al., 1999), found that Finns who drank water containing low levels of arsenic (less than 0.1 ppb) had about a 50 percent lower risk of getting bladder cancer than their countrymen who drank water containing somewhat more arsenic (0.1 ppb to 0.5 ppb). Significantly, people who drank more than 0.5 ppb arsenic had more than a 140% increase in bladder cancer rates compared to those who consumed levels less than 0.1 ppb.

The pros and cons of models that characterize cancer risk bring up the role and judgment of risk assessors. The NAS’s 1983 seminal document on risk assessment in regulatory agencies and elsewhere in the federal government (NAS, 1983) suggested a four-part paradigm for quantifying health risk that is now widely used in various incarnations by governmental agencies and others. The 1983 report also repeatedly made note of the role of judgment in the risk assessment process, a fact too often ignored by interested parties viewing regulatory risk assessment models. Without a totally clear scientific consensus on the guaranteed best scientific approach, or in the face of equally acceptable approaches, we must opt for the scientific approach that provides the maximum protection for human populations. The linear extrapolation approach adopted by the NAS subcommittee is in full accord with this principle, which should apply to assessment of cancer risks for environmental contaminants.

What can we conclude about the adequacy of other regulatory guidelines or standards for arsenic, for example the EPA reference dose (RfD) for ingested arsenic?

EPA issues guidelines for the intake levels of environmental contaminants that the agency generally considers to be free of toxic risk during long-term, that is, lifetime, exposures. In the case of oral intakes these values are called reference doses, RfDs. They are expressed in milligrams (mg) of contaminant daily intake per unit body weight in kilograms (kg-day). RfDs, being derived for oral intakes, do not usually take account of other routes of intake. Inhalation of contaminants might be a significant exposure route, in which case a reference concentration, RfC, expressed as milligrams per cubic meter of ambient air, may also be used. It is important to note that if more than one exposure route is significant, we must recognize that the RfD is less protective than we would otherwise conclude if we thought that arsenic in drinking water was the sole route of exposure. EPA, in its general description of the RfD approach, notes the need to take account of other intake routes (EPA, 1993).

EPA has set the RfD for ingested inorganic arsenic, the amount viewed as not being linked to any health risk, at 0.0003 mg/kg-day (0.3 µg/kg-day). This value is derived for skin hyperpigmentation and keratosis and potential vascular effects. Analyses in the preparation of this paper, including a review of health effects data for the United States, found no currently valid and convincing reasons to say this value is too low. Thus, no higher RfD is warranted.

EPA’s failure to fully consider risks to children in the RfD derivation is of concern. It is true that early childhood is only a fraction of the total lifetime interval considered when deriving an RfD for lifetime effects of arsenic. However, the relatively inefficient detoxification of a potent carcinogen and toxin by children, and the increased sensitivity (and higher exposure per unit of body mass) of children to arsenic-associated central nervous system effects, are serious issues. EPA should revise the current RfD downwards to account for the apparent elevated vulnerability of children; the data certainly do not support any upward revision of the current value.

In addition, EPA has not reconciled the health risks represented by the current RfD value based on noncancer toxic effects with the internal cancer risk estimates calculated for drinking water arsenic in the 1999 NAS report. The current RfD permits a "safe" daily intake by a 70 kg adult male of 21 µg arsenic per day. Risk-characterization estimates in the NAS report for the MCL value permit calculation of a cancer risk for this "safe" 21 µg daily intake that markedly exceeds any acceptable regulatory risk management guideline for cancer. Put differently, the amounts of arsenic intake that may be safe for noncancer risks are unsafe for cancer risks.

To protect children and infants, an RfD at least three-fold lower, 0.1 µg/kg-day, is certainly more defensible and more protective of identifiable at-risk populations in the United States. This adjustment is based upon standard EPA use of "uncertainty" factors for the RfD. The current uncertainty factor of three should be increased 10, the next generally permitted level for such a factor, based on concerns about the special susceptibility of children. Even such a lower RfD, it should be noted, would still present a cancer risk higher than EPA would generally consider acceptable. We recommend that the RfD be reduced to at most this level.

What can we conclude about what a health-protective level of arsenic in U.S. drinking water supplies should be to prevent cancer and noncancer effects in the U.S. population?

According to the data, we need a much lower and more protective EPA standard for drinking water arsenic and a much lower and more protective reference dose guidance level for arsenic.

Given the risk estimates for all internal cancers provided in the NAS’s 1999 report, the current EPA MCL for arsenic must be revised downward to no higher than a value at the Practical Quantitation Level (PQL) of 3 ppb. EPA completed a thorough review of laboratory capabilities in 1999, and concluded that the PQL is 3 ppb (Miller, 1999). Thus, a new MCL of 3 ppb is reasonable, based on the newest analytical methodology assessment from EPA (which is more current than the 4 ppb figure cited by NAS, 1999, a level based on earlier studies, see, Eaton et al., 1994; Mushak and Crocetti, 1995).

• Our conclusion that the MCL should be 3 ppb is driven by practicality, that is, one cannot regulate below what one can measure for compliance. This does not say that values lower than the PQL of about 3 ppb pose no cancer risk; it only recognizes that quantification of these lower levels in drinking water is problematic at this time. While many laboratories can reliably detect arsenic at levels below one ppb, reviews of a variety of laboratories to date have found that many others are unable to reliably detect and quantify the concentration of arsenic at these levels. As the NAS recommended in its 1999 report on arsenic in drinking water, EPA should immediately seek to reduce the PQL for arsenic by developing and standardizing improved analytical techniques for arsenic. The only alternative to setting an MCL at the PQL would be for EPA to establish a "treatment technique" for arsenic, an approach that seems difficult to justify here since arsenic is reliably detectable down to the low ppb range.
  
  
• There is no scientifically sound reason for increasing the noncancer RfD value from 0.3 µg/kg-day to a higher value. To the contrary, as noted above, there is good reason to adjust the value lower. Adults ingesting the "safe" arsenic dose for noncancer effects will simultaneously be at too high a risk for internal organ cancers. While EPA’s risk management guideline for permissible skin cancer risk was changed to 1 in 10,000 in 1988, the guideline for the more dangerous, more often fatal internal cancers should remain at 1 in 1,000,000. One cannot get to anything near this cancer rate guideline with the present RfD value if one assumes significant contribution of carcinogenic inorganic arsenic from food.
  
  
• For these reasons, an RfD at least three-fold lower, 0.1 µg/kg-day, is certainly more defensible and more protective of identifiable at-risk populations in the United States.

How can we prevent arsenic from getting into drinking water, or remove it from drinking water once it’s there?

Arsenic gets into drinking water from a variety of sources. Sources from human activities include:

• Leaking of arsenic from old industrial waste dumps. Arsenic is one of the most common contaminants found at Superfund sites, for example.
  
  
• Leaching of arsenic from mines and mine tailings. Some hard-rock and other mines expose arsenic-bearing rock to the elements, "liberating" the arsenic into the environment, and in some cases causing serious arsenic contamination of ground and surface water.
  
  
• Runoff or leaching of old arsenic-containing pesticides from sites where they were heavily used. In some cases, the old arsenic-based pesticides remain in the areas where they were applied, manufactured, or disposed of years ago, and can get into water supplies.
  
  
• Heavy groundwater pumping. Recent studies in Wisconsin and elsewhere have shown that heavy pumping of groundwater has increased arsenic levels in some wells. In some cases heavy pumping appears to have pulled water out of heavily arsenic-contaminated layers of rock that were not the primary aquifer being tapped but had not been sealed off from the well. In other cases, possibly because overpumping appears to have caused groundwater levels to drop, increasing arsenic-bearing rock contact with air and thereby increasing arsenic leaching).

Cleaning up old dumpsites under Superfund and related programs may reduce arsenic contamination in some systems affected by arsenic from industrial sites. Additionally, arsenical pesticide hot spots, and certain mine waste sites, are sometimes covered by Superfund or other cleanup laws and should be addressed in order to reduce water contamination.

Efforts to reduce leaching and drainage from mines and mine tailings by improving reclamation and mining practices should also be undertaken to reduce arsenic loading into many water sources. Furthermore, it is worth investigating whether reworking contaminated wells (for example, using a casing and cement to seal off arsenic-bearing rock layers that may be leaking water into the well) and/or reducing pumping rates may in some cases reduce arsenic levels in systems. Government officials and water systems should work with citizens to remedy these problems so water supplies are not contaminated by arsenic and do not need to be treated for arsenic removal.

The best way to avoid arsenic contamination from reaching our taps is to prevent it from getting into the environment in the first place. Where prevention is not possible, as when the arsenic occurs naturally, and when no alternative water source is available and the system cannot consolidate with another, cleaner water system, water treatment is readily available. Treatment already in use by some progressive water utilities has been demonstrated to reduce or essentially eliminate arsenic contamination of tap water. Among the effective arsenic treatment options EPA has identified (EPA, 1999; EPA 1994) are:

• Modifying Existing Coagulation and Filtration. Large water systems that already have coagulation and filtration technology (as most surface water systems do) can take simple steps to modify these processes to substantially reduce arsenic levels. Changing their use of iron or manganese oxidation, use of ferric chloride or ferric sulfate, and alum coagulation and filtration can reduce arsenic by 80 to 95 percent. These steps are relatively inexpensive.
  
  
• Water Softening with Lime. Many water systems already use lime to "soften" their water (that is, to reduce water "hardness" by removing the minerals calcium and magnesium). We now know that softening, if optimized, can reduce arsenic levels by 60 to 90 percent. It is about as inexpensive as coagulation and filtration modifications.
  
  
• Activated Alumina. Activated alumina can be packed into beds through which water is run in a treatment plant to remove arsenic. While this method works well for most waters, if the source water has high levels of selenium, fluoride, or sulfate, it is not as effective at arsenic removal.
  
  
• Ion Exchange. This technology, already used by many water systems, can remove arsenic effectively in most water. Again, however, if levels of certain other chemicals (such as sulfate, selenium, fluoride, or other dissolved solids) are too high, pretreatment using other technologies is needed to assure that adequate levels of arsenic are removed.
  
  
• Electrodialysis Reversal. Essentially the same process as used to clean blood at dialysis centers, electrodialysis takes advantage of the charge of particles (like arsenic) and a special membrane under the influence of an electric current, and can remove about 80 percent of arsenic from water.
  
  
• Reverse Osmosis and Nanofiltration Membranes. RO and NF membranes can remove 90 percent to more than 95 percent of arsenic. These membranes can reject substantial amounts of water, and therefore waste-stream recovery or other actions may be necessary in the arid West. Also, particularly if arsenic levels in the raw water are high, treatment or disposal of the concentrated brine created by removing the arsenic from the water can increase costs.
  
  
• Point of Use and Point of Entry Treatment. Under the 1996 Safe Drinking Water Act Amendments, water suppliers are authorized, under strict conditions, to use point-of-use filters (for example, RO units installed under kitchen sinks) or point of entry filters (for example, treatment devices in the basement at the point water goes into the home) to comply with drinking water standards. EPA studies have shown that these devices can be affordable and effective to treat for arsenic, and may be cheaper for small systems than installing centralized treatment. For this to work in a national rule, EPA would have to clarify utilities’ utility responsibility in assuring the continued operation and maintenance of such devices.

For several years, EPA has been evaluating the cost of installing treatment to meet various Maximum Contaminant Levels (MCL) for arsenic. EPA's most recent public analysis (Taft, 1998) found that if the standard were lowered from the current 50 ppb down to 5 ppb, it would cost most households (those served by city systems serving 100,000 people or more) about $2 a month, and would cost up to $14 a month for people living in smaller towns (with 10,000 to 100,000 people). Even a standard as low as 2 ppb would cost city dwellers with arsenic problems about $5 a month, and those living in affected towns as small as 10,000 people would pay about $14 a month.

Systems serving over 10,000 people serve the vast majority of people affected by arsenic contamination. Our analysis of EPA’s 25-state arsenic database shows that about 9 out of 10 people (87 percent) who consume arsenic at a significant level in their tap water (over 1 ppb) are served by these systems serving more than 10,000 customers.

For the 13 percent of consumers who get their water from smaller systems, however, treatment costs can be significantly higher than they are for consumers in cities, because of the lack of economies of scale. Thus, EPA estimates that people drinking water from a system serving 3,300 to 10,000 people may have to pay as much as $20 a month, and the smallest systems (assuming the worst case and that no point-of-use or other devices were allowed) could reach $100 a month (Taft, 1998).

Using these figures, EPA has estimated that a 5 ppb arsenic rule would cost about $686 million per year, and a 2 ppb standard would cost $2.1 billion. However, EPA recently admitted (Taft 1998) that both these national cost estimates and the individual household cost estimates are probably overstatements of the true costs of treatment for several reasons:

• Most important, EPA assumed that all systems that exceeded the MCL would install full treatment of all of their water to get it well below the MCL. More recent analysis shows, however, that most water systems would actually treat only some of their water and then would blend it with untreated water, in order to produce water just under the MCL, to keep the costs down.
  
  
• EPA assumed that if a water system with multiple wells has just one or a few wells exceeding the arsenic MCL, the system will treat all of its wells, including those below the MCL; EPA now understands that this is extremely unlikely.
  
  
• EPA's estimates did not account for recent advances in treatment technologies, such as the newly understood ability of the relatively inexpensive ion-exchange treatment to effectively treat all but the highest sulfate waters.
  
  
• EPA’s estimates failed to account for improvements in water quality that are expected to be required by other EPA rules, such as the groundwater rule, the Stage 2 Microbial and Disinfection Byproducts rule, and the uranium rule, all of which are expected to drive many water systems to use treatment that will also reduce arsenic.
  
  
• The older EPA estimates do not consider the availability of point-of-use and point-of-entry devices now authorized by the 1996 SDWA Amendments, technologies that are substantially less expensive than centralized treatment for many small systems.
  
  
• EPA's cost estimates do not account for expected reductions in treatment costs as more treatment technology is installed.

---

Figure 4: Percent of Population Drinking Arsenic at Significant Levels* Served by Large vs. Small Systems

---

Even with these reasons to believe EPA is overestimating costs, it is clear that at least some small systems will have to pay relatively high costs per household to have arsenic-safe water. For these smaller systems, federal and state assistance to improve treatment is available, and arsenic contamination should be a high priority for these drinking water funds. Additional federal and state funding through State Revolving Funds (SRF), USDA's Rural Utility Service, and other programs may also be needed. The SRF established by the Safe Drinking Water Act Amendments of 1996, which has not been fully funded since the act's passage, should be funded at least to the full authorized amount ($1 billion per year) to help smaller systems with arsenic problems.

Therefore, even using EPA’s high cost estimates,^[4] a strict arsenic standard for tap water would be both sound public health policy and affordable for consumers. It is EPA’s obligation to protect the American public from arsenic contaminated tap water, by issuing a strict MCL of 3 ppb arsenic.

CONCLUSIONS

Americans should be able to turn on their taps and be sure that their drinking water is safe. Arsenic is perhaps the worst example of EPA’s failure to address a serious health risk from a chemical contaminant in drinking water. The agency has had over a quarter century, since the Safe Drinking Water Act passed in 1974, to adopt a modern tap water standard for arsenic, but has failed to do so. The time has come for the agency to act. Specifically, we recommend that:

• EPA Must Immediately Propose and Finalize by January 1, 2001 a Health-Protective Standard for Arsenic in Tap Water. The National Academy of Sciences (NAS) has made it clear, and we agree, that EPA should expeditiously issue a stricter Maximum Contaminant Level standard for arsenic. Based on available scientific literature and NAS risk estimates, this standard should be set no higher than 3 ppb -- the lowest level reliably quantifiable, according to EPA. Even an arsenic standard of 3 ppb could pose a fatal cancer risk several times higher than EPA has traditionally accepted in drinking water.
  
  
• EPA Must Revise Downward its Reference Dose for Arsenic. EPA's current reference dose likely does not protect such vulnerable populations as infants and children. Furthermore, "safe" arsenic intakes in the RfD present unacceptably high cancer risks. To protect children, EPA should reduce this reference dose from 0.3 micrograms per kilogram per day (µg-kg/day) to at most 0.1 µg-kg/day. For concordance with cancer risk numbers, EPA should reevaluate the RfD in more depth as expeditiously as feasible.
  
  
• EPA Should Assure that Improved Analytical Methods Are Widely Available to Lower Detection Limits for Arsenic. EPA must act to reduce the level at which arsenic can be reliably detected in drinking water, so that it can be reliably quantified by most labs at below 1 ppb, the level at which it may pose a health risk.
  
  
• Water Systems Should be Honest With Consumers about Arsenic Levels and Risks. It is in public water systems’ best long-term interest to tell their customers about arsenic levels in their tap water and the health implications of this contamination. Only when it is armed with such knowledge can the public be expected to support funding and efforts to remedy the problem.
  
  
• Water Systems Should Seek Government and Citizen Help to Protect Source Water. Water systems should work with government officials and citizens to prevent their source water from being contaminated with arsenic.
  
  
• Water Systems Should Treat to Remove Arsenic, and Government Funds Should be Increased to Help Smaller Systems Pay for Improvements. Readily available treatment technology can remove arsenic from tap water, at a cost that is reasonable ($5 to $14 per month per household) for the vast majority of people (87 percent) served by systems with arsenic problems. Very small systems serving a small fraction of the population drinking arsenic-contaminated water, however, will often be more expensive to clean up per household. Assistance to such systems should be a high priority for drinking water funds such as the SRF and USDA's Rural Utility Service programs. The SRF should be funded at at least $1 billion per year to help systems with arsenic problems.
  
  
• EPA Should Improve its Arsenic, Geographic Information, and Drinking Water Databases. EPA should upgrade its Safe Drinking Water Information System to include and make publicly accessible all of the arsenic and unregulated contaminant data, as required by the Safe Drinking Water Act. EPA also should require water systems to provide accurate lat-long data using GPS systems, which will have widespread use in GIS systems by federal, state, and local officials, and the public, for source water protection, developing targeted and well-documented rules, and for other purposes.

Note

4. The Association of California Water Agencies and the American Water Works Association have charged the EPA has underestimated national arsenic treatment costs. However, EPA has responded in detail to these allegations and thoroughly rebutted these arguments.