Environ. Sci. Technol. 2010, 44, 5619–5626
Toxicological Relevance of Pharmaceuticals in Drinking Water G R E T C H E N M . B R U C E , * ,† RICHARD C. PLEUS,† AND SHANE A. SNYDER‡ Intertox, Inc., 600 Stewart Street, Suite 1101, Seattle, Washington 98101, and Southern Nevada Water Authority, Applied R&D Center, 1001 S. Valley View Boulevard, Las Vegas, Nevada 89153
Received February 12, 2010. Revised manuscript received May 27, 2010. Accepted May 31, 2010.
Interest in the public health significance of trace levels of pharmaceuticals in potable water is increasing, particularly with regard to the effects of long-term, low-dose exposures. To assess health risks and establish target concentrations for water treatment, human health risk-based screening levels for 15 pharmaceutically active ingredients and four metabolites were compared to concentrations detected at 19 drinking water treatment plants across the United States. Compounds were selected based on rate of use, likelihood of occurrence, and potential for toxicity. Screening levels were established based on animal toxicity data and adverse effects at therapeutic doses, focusing largely on reproductive and developmental toxicity and carcinogenicity. Calculated drinking water equivalent levels (DWELs) ranged from 0.49 µg/L (risperidone) to 20,000 µg/L (naproxen). None of the 10 detected compounds exceeded their DWEL. Ratios of DWELs to maximum detected concentrations ranged from 110 (phenytoin) to 6,000,000 (sulfamethoxazole). Based on this evaluation, adverse health effects from targeted pharmaceuticals occurring in U.S. drinking water are not expected.
Introduction Pharmaceuticals and personal care product (PPCP) ingredients have doubtlessly been released into the environment for as long as they have been in use. However, increasing sensitivity of analytical methods and growing scientific interest only recently resulted in the detection of a significant number of these compounds in water supplies (1-3). Although most pharmaceuticals are tested for human safety prior to marketing, the potential for adverse effects in nontarget populations exposed to low environmental levels has not been established. A review by Daughton and Ternes (1) suggested possible health effects of long-term exposure to pharmaceuticals via drinking water could include endocrine disruption, antibiotic resistance, genotoxicity, carcinogenicity, or allergic reactions, as well as effects on reproduction or fetal/child development, noting, “Although drugs are usually designed with a specific * Corresponding author phone: 206-443-2115; fax: 206-443-2117; e-mail:
[email protected]. † Intertox, Inc. ‡ Southern Nevada Water Authority. Current address: Professor of Chemical and Environmental Engineering, University of Arizona, 1133 E. James E. Rogers Way, Harshbarger 108, Tucson, AZ 857210011. 10.1021/es1004895
2010 American Chemical Society
Published on Web 06/24/2010
mode of action in mind..., they can also have numerous effects on nontarget, or as yet unknown, receptors and possibly cause side effects in the target organism” (1). Although no regulatory limits have been established for pharmaceuticals in drinking water, agencies are beginning to recommend monitoring (4, 5). The California Department of Health Services specified in its Draft Groundwater Recharge Reuse Regulations that PPCPs, endocrine disrupting compounds (EDCs), hormones, and other indicator compounds should be monitored in recycled water used to recharge groundwater basins designated as domestic water supplies (4). In its third Contaminant Candidate List (CCL3), the U.S. EPA has listed several pharmaceutical ingredients, including estrogenic hormones and erythromycin (5). Several investigators (6-10) assessed the potential for adverse effects from exposure to pharmaceuticals in water by comparing exposures to therapeutic doses divided by uncertainty factors (UFs) to extrapolate safe levels for populations including sensitive individuals. Others (10-13) examined health risks of pharmaceuticals in drinking water based on no-effect levels from animal toxicity studies or human exposures. Provisional “no-effect levels” for several pharmaceuticals detected in Dutch drinking water were calculated based on the maximum residue levels for veterinary pharmaceuticals in milk (14). In addition, the Australian Environment Protection and Heritage Council (EPHC) (15) has developed a decision tree for establishing drinking water guidelines for a range of compounds, including pharmaceuticals, based on toxicity data and therapeutic doses. In general, these authors concluded that human health risks from exposures to pharmaceuticals in water are not expected; however, an examination of the significance of pharmaceuticals detected in U.S. drinking waters has not been conducted. To further address the possible impact of pharmaceutically active ingredients in water systems on human health in the U.S., a study was undertaken to identify a target list of pharmaceuticals likely to be present or represent a potential health risk, obtain occurrence data from drinking water sources, develop health risk-based screening levels, and compare screening levels to detected concentrations. A total of 222 samples, including blanks and QA/QC samples, were collected from 19 drinking water treatment plants (DWTPs) (3). Samples were also collected at 15 distribution sites representing 13 of the DWTPs.
Experimental Section Selection of Target Analytes. More than 3,000 pharmaceuticals are currently approved for prescription in the U.S. and thousands more are approved for over-the-counter use. In the current study, target pharmaceuticals for analysis were selected based on likelihood of presence and potential toxicity, as well as interest among the public or water utilities. To emphasize compounds with the highest rates of use, the majority of target analytes were selected from the 200 most prescribed drugs in 2003 and the 300 most prescribed drugs in 2004 (16), representing a range of drug categories (Table 1); these compounds were administered in 2.2 million or more prescriptions per year in the U.S. From this list, compounds were screened to exclude those that did not meet at least one minimum toxicity requirement indicative of potential to cause adverse effects at low, chronic exposure levels, based on information provided in toxicity summaries (e.g., drug labels and monographs): (1) In FDA Pregnancy Category “C”, “D”, or “X” (indicating a potential for fetal risk, including fetal abnormalities and death; Table 1); (2) evidence VOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5619
TABLE 1. Target Pharmaceuticals Selected for Risk Assessmenta drug
trade name
class description
2003 Rx rank
atenolol
Tenormin beta-blocker
atorvastatin
Lipitor
antilipidemic
carbamazepine
Tegretol
anticonvulsant
diazepam
Valium
diclofenac
generic
benzodiazepine tranquilizer NSAID
enalapril fluoxetine gemfibrozil
Vasotec Prozac Lopid
ACE inhibitor antidepressant antilipidemic
122 34 126
meprobamate naproxen
generic Aleve
antianxiety agent NSAID
NA 73
phenytoin
Dilantin
anticonvulsant
147
risperidone
Risperdal antipsychotic
79
simvastatin
Zocor
15
toxicity summary and FDA pregnancy category
4
reproductive, developmental, evidence of carcinogenicity in rodents; D 2 reproductive, developmental, evidence of carcinogenicity in rodents; X 252 (2004) reproductive, developmental, evidence of carcinogenicity in rodents; D 117
reproductive, developmental; D
181 (2004) developmental; C reproductive, developmental; C, D some developmental; C developmental, evidence of carcinogenicity in rodents though relevance to humans not clear; C reproductive, developmental; D developmental; B
sulfamethoxazole generic
antibiotic
148
developmental, evidence of carcinogenicity in rodents; D reproductive, developmental, evidence of carcinogenicity in rodents; C reproductive, developmental, evidence of carcinogenicity in rodents; X developmental; C
trimethoprim
antibiotic
148
developmental; C
generic
antilipidemic
past detections SW (17), WW (18, 19) NA DW (20, 21), GW & SW (20, 22), WW (23) DW (17, 20), SW (20, 24, 25) DW (26), GW & SW (20, 25), WW (27) SW (2, 28) SW (2, 25), WW (23) DW (20), SW (2, 20, 25), WW (29) DW (20), SW (25) DW (20), SW(20, 25, 29), WW (27) SW (24, 25), WW (30) NA NA DW (20), GW & SW(20, 25, 31), WW (23) DW (20), GW & SW(20, 25, 31), WW (23)
a ACEsAngiotensin-converting enzyme; DWsDrinking water; GWsGround water; NAsNot available; NSAIDsNonsteroidal anti-inflammatory drug; SWssurface water; WWswastewater. FDA Pregnancy Category (21 CFR § 201.57): BsEither (a) Animal studies have not shown risk to fetus, but no adequate studies in pregnant women, or (b) Animal studies have shown risk to fetus but human studies show no risk; CsEither (a) Animal studies have shown an adverse effect on fetus, but no adequate studies in humans; or (b) No animal studies and human studies are inadequate to judge risk; DsData from humans provide some evidence of fetal risk; XsStudies in animals or humans have shown fetal abnormalities.
of carcinogenicity in animals or humans; or (3) evidence of reproductive or developmental toxicity in animal testing or in humans at therapeutic doses. Reported detections in wastewater, drinking water, and surface water were considered to select among drugs of apparently similar toxicity. Finally, prescription ranking was considered in selecting among similar drugs. Derivation of Health Risk-Based Screening Levels. Screening levels were derived using methodologies consistent with those applied by U.S. EPA and others to establish exposure levels not likely to be associated with adverse health effects following chronic exposure (32-34). For noncarcinogenic end points, threshold doses for toxic effects were identified from studies in animals or in humans (e.g., clinical trials) that identified no observed adverse effect levels (NOAELs) and lowest observed adverse effect levels (LOAELs) for reproductive, developmental, systemic, and other toxicity end points. Study types of most relevance for evaluating longterm low-level exposures to compounds in water were assumed to be subchronic, chronic, reproduction, and teratology studies with exposure primarily via the oral route. Sources of information included the National Library of Medicine PubMed database of toxicological literature citations, National Toxicology Program (NTP) reports, information submitted to the FDA as part of the drug approval process (e.g., Drugs@FDA) (35), drug labels or monographs (e.g., the Physician’s Desk Reference), and documents prepared by 5620
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010
the U.S. EPA, World Health Organization (WHO), International Agency for Research on Cancer (IARC), and other agencies. NOAELs and LOAELs were divided by UFs to account for potentially sensitive populations and uncertainties in the data set. The lowest resulting dose was selected as the point of departure (POD) for estimating noncarcinogenic risks. Selected UFs addressed five types of uncertainty/variability, with values consistent with U.S. EPA recommendations (36): extrapolation from animals to humans (10); intraspecies differences (10, unless the critical study is based on a sensitive population group, e.g., a developmental end point, in which 3 was used); extrapolation from a LOAEL to a NOAEL (10); study duration (10, to extrapolate from a subchronic to chronic study, or 3 if the critical study is a reproductive/ developmental study and the fetus/child is the most sensitive group); and database uncertainties (3, if either a prenatal study or a 2-generation reproduction study is missing from the database, or 10 if both are missing). For compounds showing evidence of carcinogenicity in animal studies, a linear extrapolation model was used to predict the tumorigenic response at low doses if data on tumor incidence were available (37). The resulting slope factor (SF) is an upper-bound estimate of risk per increment of dose (1 mg/kg-d of exposure), and was used to estimate the probability of developing cancer at a given dose. For drugs with cited evidence (e.g., in drug monographs) of carcinogenicity in animals but for which tumor incidence
data were not located, a method proposed by Gaylor and Swirsky Gold (38) was used to predict the cancer potential. They reviewed the results of 2-year cancer bioassays for 139 chemicals tested by the NTP and determined that a “virtually safe dose” corresponding to a cancer risk of 1 in a million can be estimated by dividing a chemical’s maximum tolerated dose (MTD) from 90-day studies in rodents by 740,000. The maximum tolerated dose is usually the high dose selected for a carcinogenicity study (39). Drinking Water Equivalent Levels. For comparison to concentrations in water, screening levels were converted to drinking water equivalent levels (DWELs) by multiplying by an average adult body weight of 70 kg and dividing by a daily drinking water ingestion rate of 2 L/d (40). DWELs for carcinogens were estimated assuming a de minimis lifetime excess cancer risk of one additional cancer per million lifetime exposures (10-6), and continuous exposure for 30 years over a 70-year lifetime (40): DWEL (µg/L) )
10-6 × 70 kg × 25,550 d × 1,000 µg/mg SF (mg/kg-d)-1 × 2 L/d × 30 yr × 365 d/yr (1)
Risk Assessment of Pharmaceuticals Detected in Water. DWELs were compared to the highest concentration detected in drinking water collected from the 19 DWTPs or 15 distribution sites (3, 41). For three compounds (atorvastatin, fluoxetine, and simvastatin), primary active metabolites (ohydroxy atorvastatin, p-hydroxy atorvastatin, norfluoxetine, and simvastatin hydroxyl acid) were also analyzed; these contribute significantly to the parent compound’s overall toxicity (42-44) and the screening levels developed for the parent compound were assumed to apply.
Results Selected Compounds. Target analytes (Table 1) comprised a range of drug classes, including three of the top 25 compounds prescribed during 2003 or 2004. Meprobamate, an antianxiety agent introduced in 1955 but not on recent lists of most prescribed drugs, was included because of evidence of frequent detection (29). Carisoprodol, a frequently prescribed skeletal muscle relaxant that is metabolized to meprobamate, likely accounts for the detections of meprobamate (45). The list includes seven pharmaceuticals that California DHS suggests be included in monitoring programs for groundwater recharge reuse projects: atorvastatin, carbamazepine, gemfibrozil, meprobamate, phenytoin, sulfamethoxazole, and trimethoprim (4, 46). The 19 target analytes were subsequently subject to analysis in water collected from the DWTPs and distribution sites (3). Screening Levels. DWELs ranged from 0.49 µg/L (risperidone) to 20,000 µg/L (naproxen). For eight of the 15 parent compounds, screening levels were based on noncarcinogenic effects (Table 2). Of these, seven were based on reproductive or developmental end points, as these yielded the lowest NOAELs or LOAELs from available studies; these compounds had composite UFs ranging from 300 to 1,000. For meprobamate, the screening level was based on effects of carisoprodol on the liver (55) with a UF of 10,000 applied because data were limited: only two studies of sufficient quality were identified and neither considered reproductive or developmental end points. Animal studies of diclofenac or naproxen identified no toxicological effect at any dose; for these compounds, screening levels were based on the highest no-effect dose from developmental toxicity studies. Screening levels for the remaining seven compounds were based on evidence of carcinogenicity. Sufficient tumor incidence data to develop SFs were identified for gemfibrozil and phenytoin; for the other five compounds, screening levels were calculated from the MTD. Screening levels calculated
from the MTD for gemfibrozil and phenytoin (0.41 and 0.049 µg/kg-d) were similar to values calculated using tumor data (0.39 and 0.058 µg/kg-d), demonstrating that this method produces acceptable estimates of carcinogenicity potential for compounds without available tumor data. For compounds with evidence of carcinogenicity, the screening level based on carcinogenicity was always lower than the screening level based on noncancer end points. Risk Assessment of Pharmaceuticals in Water. Maximum-detected concentrations in finished or distribution water were compared to DWELs (Table 3). Ten compounds were detected in one or both sources (3). The remaining nine compounds were not detected at detection limits of 0.5 ng/L or lower in either source. Of the detected compounds, estimated risks associated with atenolol, diazepam, fluoxetine, meprobamate, norfluoxetine, and sulfamethoxazole were low, with a margin of exposure (MOE, the ratio of the DWEL to the maximum detected concentration) of 3,800 or greater. Screening levels for these compounds were based on noncarcinogenic effects, although data of sufficient quality to develop a screening level were limited to fewer than three studies for meprobamate and sulfamethoxazole. Screening levels for the four remaining detected compounds (carbamazepine, gemfibrozil, phenytoin, and risperidone) are based on carcinogenicity. Of these, risperidone appears to be the most potent, with reported increases in pituitary gland and endocrine pancreatic adenomas, mammary gland adenocarcinomas, and renal tubular adenomas and adenocarcinomas in rodents (62). However, risperidone was detected above the reporting limit (2.5 ng/L) in only one of 15 distribution water samples and in no finished drinking water or source water samples (3), yielding a MOE of 170. Regarding the detection of risperidone (and that of norfluoxetine, also detected in only one distribution water sample), Benotti et al. (3) indicate, “The reason for these detections is not known, though it is important to note that both concentrations were just above their respective MRLs [method reporting limits].” The remaining detected carcinogens were detected in onequarter or more of the finished and distribution water samples, with MOEs ranging from 110 (phenytoin) to 6,500 (gemfibrozil). Although a SF was calculated for gemfibrozil using tumor incidence data, the estimated cancer potency is uncertain because the mechanism assumed to lead to cancer (proliferation of peroxisomes in the liver) occurs in rat but not human liver, and epidemiologic studies of gemfibrozil found no excess cancer mortality in humans (63). IARC determined that gemfibrozil is a Group 3 carcinogen (not classifiable as to its carcinogenicity in humans), based on inadequate evidence in humans and limited evidence in animals (63). By contrast, NTP and IARC consider phenytoin a possible human carcinogen based on evidence of the carcinogenicity in rodents but equivocal evidence in humans (57, 64). Based on the IARC assessment, California lists phenytoin as “Known to Cause Cancer” under Proposition 65 (65).
Discussion Overall, detected concentrations of pharmaceuticals at the DWTPs and distribution sites in the U.S. were well below DWELs. These conclusions are consistent with the findings of other investigations. Cunningham et al. (10) found no “appreciable risk to human health” associated with estimated environmental concentrations of 44 active pharmaceutical ingredients marketed by GlaxoSmithKline. Schriks et al. (66) compared tolerable daily intakes for 50 emerging contaminants potentially present in drinking water, including five pharmaceuticals (acetylsalicylate, carbamazepine, clofibric acid, metoprolol, and sulVOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5621
TABLE 2. Toxicological Data Used to Develop Screening Levels and Comparison to Minimum Therapeutic Doses (Minimum Therapeutic Doses from Approved Drug Labels)a (36) drug atenolol atorvastatin carbamazepine diazepam diclofenac enalapril fluoxetine
gemfibrozil meprobamate naproxen phenytoin risperidone simvastatin sulfamethoxazole trimethoprim
screening level (µg/kg-d)
screening level basis evidence of carcinogenicity; no tumor data. MTD ) 1,500 mg/kg-d (47). evidence of carcinogenicity; no tumor data. MTD ) 100 mg/kg-d (42). evidence of carcinogenicity; no tumor data. MTD ) 250 mg/kg-d (48). developmental effects (behavior, immune) in rats exposed gestationally, LOAEL ) 1.0 mg/kg-d (49, 50). UF ) 1,000 (10, 3, 10, 3, 1). no effects in mice exposed gestationally, NOAEL ) 20 mg/ kg-d (51). UF ) 300 (10, 3, 1, 3, 3). developmental effects (malformations) in offspring of women exposed during pregnancy, LOAEL ) 0.070 mg/kg-d (52). UF ) 300 (1, 3, 10, 3, 3). developmental effects (short gestation, reduced birthweight, poor adaptation) in offspring of women exposed during pregnancy, LOAEL ) 0.29 mg/kg-d (53). UF ) 300 (1, 3, 10, 3, 3). testicular tumors in rats (0 mg/kg-d: 1/50; 9.1 mg/kg-d: 8/50; 91 mg/kg-d: 17/50)(54). SF ) 6.0 × 10-3 (mg/kg-d)-1 systemic effects (increased liver weights) in mice exposed to carisoprodol for 13 wks, NOAEL ) 75 mg/kg-d (55). UF ) 10,000 (10, 10, 1, 10, 10). no effects in mice exposed gestationally, NOAEL ) 170 mg/kg-d (56). UF ) 300 (10, 3, 1, 3, 3). liver adenomas and carcinomas in mice (0 mg/kg-d: 5/48; 7.5 mg/kg-d: 14/49; 24 mg/kg-d: 30/50) (57). SF ) 4.0 × 10-2 (mg/kg-d)-1 evidence of carcinogenicity; no tumor data. MTD ) 10 mg/ kg-d (58). evidence of carcinogenicity; no tumor data. MTD ) 100 mg/kg-d (44). developmental effects (cleft palate) in rats exposed gestationally, NOAEL ) 512 mg/kg-d (59). UF ) 1,000 (10, 3, 1, 3, 10). developmental effects (cleft palate) in rats exposed gestationally, NOAEL ) 192 mg/kg-d (59). UF ) 1,000 (10, 3, 1, 3, 10).
minimum therapeutic dose (µg/kg-d)
2.0
360
0.14
300
0.34
1,000
1.0 67
29 1,400
0.23
36
0.97
330
0.39
17,000
7.5
7,000
570
4,000
0.058
4,300
0.014
26
0.14
200
510
13,000
190
8,000
a LOAELsLowest observed adverse effect level; MTDs Maximum tolerated dose; NOAELsNo observed adverse effect level; SLs Screening level; UF-Uncertainty factor. UFs in parentheses indicate value assigned for: extrapolation from animal data to human, intra-species differences, LOAEL to NOAEL, subchronic to chronic, database. Tumor incidence data reported for human equivalent doses (39).
famethoxazole) to maximum concentrations measured in the Rhine and Meuse river basins and in Dutch drinking water, and estimated margins of exposure for drinking water ranging from 0.00007 (sulfamethoxazole) to 0.04 (metoprolol). Limited data on target pharmaceuticals in drinking water from other locations are available; however, those that are available indicate no significant risk to human health at detected drinking water concentrations compared to DWELs (Table 3). DWELs were calculated assuming the rate of exposure to drinking water was higher than likely on average (a water consumption rate of 2 L/day corresponds to the 84th percentile for adultssthe mean is about 1.4 L/day (67)); however, California EPA has calculated public health goals for compounds in drinking water assuming a body weightto-tap water consumption ratio for the 95th percentile of pregnant women, or 25.2 kg-day/L (2.8 L/day assuming a weight of 70 kg) (68). Using this assumption would decrease DWELs and corresponding MOEs by about 28%. This assessment assumes exposure only through drinking water; however, people could also be exposed through other environmental sources, including agricultural products and fish. Carbamazepine was detected in soils irrigated with wastewater (69) and atenolol was detected at ppt levels in 5622
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010
cow’s milk (41). However, Munoz et al. (70) estimated potential exposures to a number of compounds, including 13 pharmaceuticals, from ingestion of crops irrigated with wastewater, and concluded that potential human health risks were minimal. Kumar and Xagoraraki (9) estimated exposures to three pharmaceuticalssmeprobamate, carbamazepine, and phenytoinsfrom fish consumption and incidental ingestion of surface water using U.S. surface water concentrations and fish concentrations computed from bioconcentration factors (BCFs), and concluded that combined average daily exposures through these routes were no more than 21% of exposures from finished drinking water consumption. Cunningham et al. (10) reached similar conclusions, estimating that for 39 of 44 pharmaceuticals, average daily exposures from fish consumption were 4% or less of exposures associated with drinking water ingestion. Higher uptake from fish was estimated for five pharmaceuticals, all of which had high BCF values of 27 L/kg or higher. When setting action levels for contaminants in surface water systems, regulatory agencies often incorporate a relative source contribution (RSC) to account for exposure through other sources, typically ranging from 20-80% for drinking water, with a default value of 20% (33, 40, 71). If an RSC is
TABLE 3. Comparison of DWELs to Maximum Drinking Water Concentrationsa concentrations in U.S. DWTPs (3)
drug
detection frequency: finished drinking water systems (n ) 18), screening level distribution water maximum water minimum margin (µg/kg-d) DWEL (µg/L) systems (n ) 15) concn. (µg/L) of exposure
maximum literature drinking water concentrations (µg/L)
atenolol carbamazepine diazepam fluoxetine norfluoxetine gemfibrozil meprobamate phenytoin risperidone sulfamethoxazole
2.0 0.34 1.0 0.97 0.97 0.39 7.5 0.058 0.014 510
70 12 35 34 34 14 260 2.0 0.49 18,000
44%, 53% 44%, 40% 6%, 0% 11%, 7% 0%, 7% 39%, 27% 78%, 73% 56%, 67% 0%, 7% 22%, 7%
0.018 0.018 0.00033 0.00082 0.00077 0.0021 0.042 0.019 0.0029 0.0030
3,800 660 100,000 41,000 44,000 6,500 6,200 110 170 6,000,000
