Pamela R. D. Williams, Exponent, 149 Commonwealth Drive, Menlo Park, CA 94025
650-688-1748 (phone), pwilliams@exponent.com
Patrick J. Sheehan, Exponent, 1970 Broadway, Suite 250, Oakland, CA  94612
510-208-2008 (phone), psheehan@exponent.com

INTRODUCTION

Over the past decade, there has been much publicity surrounding the impact of methyl tertiary butyl ether (MTBE) on drinking water supplies in the United States.  A Blue Ribbon Panel appointed by the Administrator of the U.S. Environmental Protection Agency (EPA) and an interagency assessment by the Office of Science and Technology Policy concluded that MTBE is more likely to contaminate ground and surface water than the other components of gasoline, primarily due to gasoline releases from storage tanks, pipelines, and refueling stations (USEPA, 1999; NSTC, 1997).  In California, MTBE has been detected in many shallow aquifers located near leaking underground fuel tanks and in several drinking water wells located in prominent areas (e.g., Santa Monica and South Tahoe), resulting in extensive litigation and clean-up activity (USEPA, 2000; STPUD, 2000; Happel et al., 1998).  Due to perceived threats to the environment and water quality, the Governor of California recently issued an Executive Order to eliminate the use of MTBE statewide no later than December 31, 2002 (ARB, 1999).  Other states, such those in the Northeast, have also sought waivers to opt out of the federal reformulated gasoline program in order to reduce or eliminate the use of MTBE (NESCAUM, 1999).

Despite claims about the nature and extent of MTBE drinking water contamination, a comprehensive characterization of the available data has not been conducted.  The California Department of Health Services (CDHS), which provides summary data on MTBE drinking water detections, estimates that only 0.8% of drinking water sources and 1.9% of public water systems sampled in California had detectable levels of MTBE as of September 5, 2000 (CDHS, 2000).  Of these, less than 1% had MTBE drinking water concentrations above the State’s primary health-based standard of 13 ppb.  Although California’s MTBE monitoring databases is not representative of all drinking water supplies in the state, sampled drinking water systems are estimated to serve about 29.9 million people, or 88% of the total population (CDHS, 2000). 

In several previous analyses, we relied on the available CHDS monitoring data to characterize the detection frequency and average concentration of MTBE in California’s drinking water supplies, and to assess whether MTBE drinking water exposures are likely to pose a human health risk (Williams, 2001; Williams et al., 2000a, b).  In this paper, we provide a summary of some of the key findings from these analyses in order to provide a better perspective on the incidence and implications of MTBE in California’s drinking water.  The remainder of this paper consists of the following five sections: (1) MTBE detections in drinking water samples, sources, and systems in California; (2) detected concentrations of MTBE in drinking water; (3) the distribution of MTBE exposures for households with contaminated drinking water in California; (4) MTBE toxicity and estimated health risks for households with contaminated drinking water; and (5) discussion of key findings and recommendations for future research.

MTBE DETECTIONS IN CALIFORNIA DRINKING WATER

More than 29,000 drinking water samples in California were collected and tested for MTBE from 1995 to 2000, representing nearly 4,300 drinking water sources and about 1,700 drinking water systems.  The overall detection rate for MTBE during 1995–2000 was estimated to be 1.3% for all drinking water samples, 2.5% for drinking water sources, and 3.7% for drinking water systems in California.  The detection frequency for MTBE was the highest in 1995, ranging from 3 to 6%, likely due to the sampling of drinking water sources or systems that were suspected of having MTBE contamination.  The detection frequency for MTBE decreased 2- to 3-fold in 1996 and has remained relatively stable since then, despite increased sampling efforts in later years.  We distinguish between samples, sources, and systems because statistical analyses based on these outcomes may yield different results and various states may report their findings based on different outcomes.  In general, drinking water systems tend to encompass more than one source, and drinking water sources often contain multiple samples.

It is important to note that MTBE detections in later years may be influenced by the use of more sensitive analytical instruments and lower detection limits.  For example, although the most frequently reported detection limit for nondetect samples was 5 ppb in 1998, the most frequently reported detection limit in 1999 and 2000 was 3 ppb.  The CDHS monitoring database does not provide detection limits for all samples, however, thereby hindering a more comprehensive investigation of this issue.  It is difficult to interpret the observed decreased detection frequency for MTBE in 2000, because data have not yet been reported for the entire year.

The majority of drinking water sources and systems in California have been sampled for only one or two years.  About 36% of drinking water sources and 24% of systems were sampled for three or more years, while only 1% of sources and systems were sampled for all six years.  Reported findings of MTBE drinking water detections over time may therefore be influenced by the number of years that drinking water sources or systems are sampled, and the observed stability of MTBE drinking water detections may be an artifact of the sampling methodology.  Past detections of MTBE may also not be relevant for evaluating impacts on drinking water supplies over a longer time horizon. For example, an analysis of impacted drinking water sources that were sampled for three or more consecutive years reveals that MTBE is not detected consistently in these sources over time.  Of the 59 sources sampled for three or more consecutive years, 39% had detectable MTBE levels for only one year, 41% had detectable MTBE levels for two consecutive years, and 15% had detectable MTBE levels for three or more consecutive years.  Only five drinking water sources that were sampled for three or more consecutive years had detectable levels of MTBE for every year it was sampled. 

MTBE was detected in 31 out of 58 counties sampled in California in 1995–2000.  The detection frequency for the majority of counties is less than 5%, but can vary considerably by outcome of interest (e.g., sample, source, or system).   The greatest detection rate for MTBE (i.e., ³ 9%) appears to be concentrated in 5–9 different counties, which account for about 9–21% of the California population.  However, this does not imply that up to 21% of the population may have contaminated drinking water, since most counties have multiple sources and systems that might be used for public drinking water at any particular point in time.  The detection rate for Los Angeles, which comprises about 28% of the population in California, is fairly low.  This finding reveals that although specific communities such as Santa Monica may have been impacted by MTBE, drinking water supplies serving Los Angeles county have generally not been affected.  Most other counties in California that account for over a million persons each also have relatively low detection rates for MTBE, which suggests that MTBE contamination of drinking water supplies in California is not uniform throughout the state, but that certain geographic areas may have higher risk factors for MTBE contamination.  

DETECTED CONCENTRATIONS OF MTBE IN CALIFORNIA

For drinking water supplies with detectable levels of MTBE in 1995–2000, average detected MTBE concentrations were the highest in 1995 and 1996, ranging from 66 to 78 ppb for all drinking water samples, 37 to 58 ppb for drinking water sources, and 13 to 40 ppb for drinking water systems.  Average detected MTBE levels decreased significantly in 1997 and remained at 10 ppb or below from 1997 through 1999.  There appears to be a slight increase in MTBE detected concentrations over the last three years, with average detected MTBE levels reaching 13 ppb for all drinking water samples, 12 ppb for drinking water sources, and 15 ppb for drinking water systems in 2000.  These latter findings are difficult to interpret, however, given the lack of a complete data set for 2000.  The smaller number of drinking water samples collected in 1995 and 2000 is reflected in the greater variability in average detected MTBE concentrations for these years. 

It is important to recognize that these estimates are based only on drinking water supplies with detectable levels of MTBE, and are therefore, not representative of MTBE drinking water levels or exposures for the general population in California.  Over 95% of drinking water supplies in California had nondetectable levels of MTBE in 1995–2000, and inclusion of these samples in the analysis would significantly reduce average MTBE levels.  Indeed, in our earlier assessments we found that average MTBE levels for all drinking water sources ranged from <1 ppb to 6 ppb in 1995–1999, depending on whether nondetect samples were assumed to equal zero or the analytical detection limit, respectively (Williams et al., 2000a, b).

Approximately 73% of drinking water samples and 86% of drinking water sources and systems with detectable levels of MTBE, contain MTBE at concentrations below the State’s primary MCL (health-based standard) of 13 ppb.  In addition, about 56% of all drinking water samples and 70% of drinking water sources and systems have detectable MTBE levels below California’s secondary MCL (aesthetic-based standard) of 5 ppb.  These findings suggest that, although some drinking water supplies in California have been affected by MTBE, the majority of these sources or systems contain MTBE at concentrations that are unlikely to be of health (or aesthetic) concern.

HOUSEHOLD EXPOSURES TO MTBE FROM DRINKING WATER

To estimate the actual distribution of MTBE exposures in California for households with contaminated drinking water (from 1995-1999), we conducted a probabilistic exposure analysis based on the ingestion of MTBE in drinking water, dermal contact with MTBE during showering, and the inhalation of MTBE from volatilized water in the home.  This latter scenario includes exposure to volatilized MTBE during showering, in the bathroom, and in the household from multiple sources (e.g., washing dishes, washing clothes, etc.).  Exposure calculations are based on the CalTOX Multimedia Total Exposure Model by Cal-EPA (1994), which assumes that everyone showers rather than bathes, and tends to overpredict the actual risk for those who bathe because of conservative assumptions about inhaled MTBE vapors during showering. 

To account for the variability and uncertainty in the exposure model, all relevant input parameters were characterized by distributions rather than point estimates, and these are reported in our earlier publications (Williams et al., 2000a, b).  The probabilistic analysis was performed using Crystal Ball software and the Latin Hypercube sampling method for 10,000 iterations.  The average daily dose (ADD) and lifetime average daily dose (LADD) of MTBE were estimated by aggregating doses from each of the three exposure pathways, which in turn, were calculated using the following equations:

Ingestion of MTBE in Drinking Water

Inhalation of MTBE from Volatilized Water

Dermal Contact with MTBE during Showering

where:

ADD                Average daily dose (mg/kg-day);

LADD              Lifetime average daily dose (mg/kg-day);

C                     Concentration of MTBE in drinking water (mg/L);

EF                    Exposure frequency (days/year);

ED                   Exposure duration (years);

BW                  Body weight (kg);

AT                   Averaging time (days);

IR                    Drinking water ingestion rate (L/day);

Aingest             Oral absorption of MTBE (unitless);

Cs,Cb,Ch         MTBE concentrations in shower air,
                       bathroom air, and household air,
                       respectively (mg/m3);

ETs,ETb,ETh   Exposure time in the shower, bathroom, and house, respectively (hrs/day);

BR                   Breathing rate (m3/hr);

Ainhal               Lung absorption of MTBE (unitless);

SA                   Surface area of the skin (cm2);

PC                   Permeability coefficient (cm/hr);

F                      Fraction of skin in contact with water
                        (unitless); and

CF                   Conversion factor (0.001 L/cm3).

Volatilized concentrations of MTBE in the shower, bathroom, and household air were calculated using the following equation by Finley et al. (1993):

where:

Ci         MTBE air concentration in the ith compartment
            (shower, bathroom, or house)   (mg/m3);

Wi        Water use rate in the ith compartment (L/hr);

fI         Mass transfer efficiency from water to air for the ith
           compartment (unitless);

C         MTBE groundwater concentration (mg/L); and

VRi      Air exchange rate (m3/hr).

The estimated ADD of MTBE from all routes of exposure is about 0.1 g/kg/day at the 50th percentile and 1.4 g/kg/day at the 95th percentile.  Exposures via ingestion account for the greatest contribution to total MTBE daily dose at both the 50th and 95th percentiles.  Dermal contact accounts for less than 5% of the total daily MTBE dose.

MTBE TOXICITY AND ESTIMATED HEALTH RISKS

To date, no national or international regulatory agency has formally classified MTBE as a human carcinogen.  The available genotoxicity data also suggest that MTBE is not highly mutagenic (ECETOC, 1997).   The U.S. EPA (1997) has determined that MTBE is an animal carcinogen, however, and poses a carcinogenic potential to humans.  The California Office of Environmental Health Hazard Assessment (OEHHA) also considers MTBE to be a possible human carcinogen, and based on this presumption, has recently derived an upper-bound cancer slope factor (CSF) for MTBE in drinking water of 1.8´10-3 mg/kg/day-1 (OEHHA, 1999).  This estimate is based on the geometric mean of three potency estimates obtained from two animal studies by Chun et al. (1992) and Belpoggi et al. (1995), for which tumors were observed at multiple target sites and under inhalation and gavage MTBE dosing regimes.  A modified physiologically based pharmacokinetic (PBPK) model was also used by OEHHA to estimate the absorbed dose of MTBE in animals (OEHHA, 1999).

For illustrative purposes, we rely on the OEHHA cancer slope factor to estimate potential carcinogenic risks to Californians from exposures to MTBE in drinking water, but make no claims about the reliability of this estimate.  Assuming that MTBE is carcinogenic to humans, the estimated lifetime cancer risk from drinking water exposures to MTBE is calculated by the following equation:

Risk = LADD ´ CSF

where:

Risk                 Lifetime cancer risk from MTBE
                        exposures;

LADD              Sum of lifetime average daily dose for all
                       three exposure pathways (mg/kg/day);
                       and

CSF                 Theoretical upper-bound cancer potency
                        of MTBE (mg/kg/day)-1

The LADD was calculated for three different exposure durations.  First, a five-year exposure duration was used to estimate potential cancer risks based on MTBE drinking water exposures in California from 1995 through 1999.  Second, an 8-year exposure duration was used to estimate potential cancer risks based on past and projected exposures to MTBE from 1995 through 2002—i.e., the phase-out date for MTBE.  Because MTBE drinking water exposures may not cease after MTBE use is discontinued, due to a lag time in the fate and transport of released MTBE in groundwater, a third exposure duration of 13 years was evaluated (i.e., 1995–2007).  This latter exposure duration was selected after modeling the maximum time it would take for MTBE to be observed at a drinking water well 100 meters downgradient of a leaking underground storage tank.  Assuming an infinite source and sandy-loam soil conditions, preliminary modeling efforts suggest that MTBE would reach steady state in about 5 years. 

At the 50th percentile of exposure, lifetime cancer risks are estimated to be 1´10-8 for a 5-year exposure period, 2´10-8 for an 8-year exposure period, and 3´10-8 for a 13-year exposure period.  At the 95th percentile, estimated MTBE cancer risks over a lifetime are 2´10-7, 3´10-7, and 5´10-7 at 5, 8, and 13 years, respectively.  Although not reported here, estimated cancer risks are considerably less for the general population in California, which includes households with and without contaminated drinking water.

To assess potential non-cancer health effects, estimated ambient MTBE concentrations were compared to EPA’s chronic reference concentration (RfC) of 3 mg/m3 for inhalation exposures, and to the Agency for Toxic Substances and Disease Registry’s (ATSDR) intermediate oral Minimal Risk Level (MRL) of 0.3 mg/kg/day for ingestion and dermal exposures (IRIS, 2000; ATSDR, 1998).  Specifically, the following equation was used to establish a Hazard Index (HI) for MTBE in drinking water from all exposure routes:

HI = HQinhale + HQingest + HQdermal

where:

HQinhale          Hazard quotient for inhalation estimated as the concentration of MTBE inhaled ¸ RfC;

HQingest          Hazard quotient for ingestion estimated as the ADD from oral MTBE exposures ¸ oral MRL; and

HQdermal        Hazard quotient for dermal uptake estimated as the absorbed dermal MTBE dose ¸ oral MRL.

A hazard index less than one (HI<1) means that estimated exposures are not expected to pose an adverse health hazard.  The HI for MTBE is significantly less than one at the 50th and 95th percentiles for households with contaminated drinking water.  This finding suggests that household exposures to MTBE in drinking water are not expected to pose significant non-cancer health effects in California, even for more highly exposed groups. 

DISCUSSION

The findings of our previous analyses are in stark contrast to many media reports that suggest MTBE contamination of public drinking water supplies in California is widespread and growing.  Based on review of the available monitoring data in California, we find that the percentage of sampled drinking water sources with detectable levels of MTBE is quite low, and has remained relatively stable from 1995–2000.  We also find that many drinking water sources are not routinely sampled for MTBE, and in those sources that appear to be affected by MTBE, the compound is not consistently detected.  In addition, the majority of MTBE detections appear to be concentrated in several geographic areas, which may be at higher risk for MTBE contamination.  Reported findings do not include private drinking water wells, however, and may be influenced by limited sampling efforts and changes in analytical detection limits.   

Detected concentrations of MTBE in California’s drinking water supplies have also not changed dramatically over the last several years, with the majority being less than the State’s primary drinking water standard of 13 ppb.  Furthermore, our probabilistic exposure analysis suggests that MTBE is unlikely to pose a significant health risk for households with contaminated drinking water in California.  However, there are some uncertainties in this assessment due to incomplete data on MTBE.  For example, even though there is no clear evidence to indicate that this is likely, certain segments of the population, such as young children and the elderly, could potentially be more susceptible to the toxic effects of MTBE than the general population (ATSDR, 1998).  Estimated exposures to MTBE in the future are also based on the assumption that the incidence of MTBE detections and the levels of MTBE in drinking water will not change dramatically in California from current estimates.  In reality, household drinking water exposures to MTBE may increase or decrease over time depending on many factors, including future releases of MTBE in the environment, the transport and degradation rate of MTBE in groundwater, and ongoing efforts to upgrade or retrofit leaking underground storage tanks (Davidson and Creek, 1999).

To reach an informed decision about the impact of MTBE on drinking water supplies, decision makers will need to be presented with all the relevant information.  In particular, a careful review of the available drinking water monitoring data is needed.  This information can be used not only to evaluate trends in drinking water detections, but also to assess potential public health risks from MTBE drinking water exposures.  Our findings suggest that, although some drinking water supplies in California have been affected by MTBE, the majority of drinking water sources and systems have not been affected or they contain MTBE at concentrations that are below levels likely to be of health concern. 

Decisions about how or whether to regulate MTBE based on perceived threats to water quality will also need to be weighted against the notable air quality benefits of MTBE, including reductions in many criteria and toxic air pollutants.  In addition, the risks and benefits of reformulated gasoline containing MTBE should be compared to the potential environmental and health consequences of alternatives to MTBE.  For example, substitution of MTBE with ethanol or a non-oxygenated blend could result in backsliding on air quality or contribute to greater water contamination by other gasoline constituents, such as benzene.  We recommend that decision-makers embark on a comprehensive risk-benefit analysis of alternative gasoline formulations in order to minimize potential environmental and health risks, while maximizing public health benefits.

Top