Toxicology Letters 144 (2003) 383 /395 www.elsevier.

com/locate/toxlet

Probabilistic hazard assessment of environmentally occurring pharmaceuticals toxicity to sh, daphnids and algae by ECOSAR screening
Hans Sanderson *, David J. Johnson, Christian J. Wilson, Richard A. Brain, Keith R. Solomon
Center for Toxicology, University of Guelph, Bovey Building, Gordon Street, Guelph, Ont., Canada N1G 2W1 Received 8 April 2003; received in revised form 30 May 2003; accepted 30 May 2003

Abstract The risks associated with occurrence of pharmaceuticals in water resources are mostly unknown. In the absence of extensive toxicological data, we scanned all the compounds observed in the environment for toxicological properties by (Quantitative) Structure Activity Relationship ((Q)SAR). The results of the probabilistic distribution of environmental and effect concentrations and hazard quotients (HQs) do not indicate significant acute risks prior to application of assessment factors. Compared with measured effect concentrations SAR predictions were more sensitive 80% of the time. The long-term effects of subtle and chronic changes, additive or synergistic effects and effects on other endpoints e.g. reproduction, behavior, metabolism, bacterial resistance etc. are still uncertain. (Q)SARs can be important prioritization tools for subsequent experimental risk assessment of pharmaceuticals in surface waters, due to the prevalent lack of ecotoxicological data. # 2003 Elsevier Ireland Ltd. All rights reserved.
Keywords: Pharmaceuticals; ECOSAR; EC50; Probability; Hazard quotients

1. Introduction Recently medical and personal care products have received increasing attention from environmental and health agencies across the European Union and in North America. Surveys and reports
* Corresponding author. Tel.: '/1-519-824-4120x54794; fax: '/1-519-837-3861. E-mail address: hsander@uoguelph.ca (H. Sanderson).

on the occurrence of pharmaceuticals in the environment (primarily surface waters) show that medical compounds are ubiquitous (Daughton and Ternes, 1999). Pharmaceuticals are created with the intent of causing a biological effect, they often have similar types of physio-chemical behavior that are characteristic of harmful xenobiotics e.g. they are able to pass membranes, and they are relatively persistent in order to avoid being inactivated before having their therapeutic effect.

0378-4274/03/$ - see front matter # 2003 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/S0378-4274(03)00257-1

384

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

These compounds are excreted through feces and urine as a mixture of metabolites and unchanged substances. They, therefore, predominately enter the environment via wastewater effluent, aggravated by the fact that, in practice the majority of people flush unused drugs down the drain or dispose of it with household garbage (Jones et al., 2001). Other sources include, direct application in aqua farming, manure run-off, as run-off from the application of sewage sludge and manure on farmland as fertilizers (Halling-Srensen et al., 1998), via hospital effluent (Ku mmerer, 2001) or, finally, via landfill leaching (Richardson and Bowron, 1985). Since the sophistication of analytical methods has increased, so has the range of detection of xenobiotics in the environment. Hence, pharmaceuticals have been proven to occur in surface water (Kolpin et al., 2002). Even if the environmental half-life of the parent pharmaceutical compounds may not be relatively great compared, this is compensated, however, by continuous replacement of the compounds in the environment, which serves to sustain perpetual life-cycle exposure for aquatic organisms (Daughton and Ternes, 1999). The quantities of several of pharmaceuticals used throughout the world are comparable to agrochemicals (Jones et al., 2001). Directives by the US Food and Drug Administration (FDA) since 1995 (CDER, 1995) and in the EU since 1993 for human and veterinary compounds (Straub, 2002) stipulating that an environmental risk assessment should be part of the approval procedure of new medical substances. Few new medical substances have been subjected to a complete risk assessment (Halling-Srensen et al., 1998) primarily due to the fact that in most instances the calculated environmental concentrations lie below the proposed cut-off values, making further ecotoxicological studies unnecessary. However, the importance of identifying emergent risks such as pharmaceuticals in the environment is reflected in the fact that pharmaceuticals are one of the top five goals of the Strategic Plan 2000 for the US Environmental Protection Agencys Office of Research and Development (Daughton and Ternes, 1999).

The current US regulatory guidance requires new pharmaceuticals to undergo standard acute toxicity tests (algae, Daphnia magna and fish) if the predicted or measured environmental concentration (PEC/MEC) of the active ingredient is !/1 mg l(1. In the EU the cut-off PEC value is 0.01 mg l(1, and no environmental concerns are apparent no further testing is deemed necessary. In the second tier a crude predicted no-effect concentration (PNEC) for the aquatic compartment is to be extrapolated by dividing the lowest E(L)C50 from standard tests by an assessment factor of up to 1000 in the EU. If the PEC/PNEC is B/1 no further assessment is necessary. The third tier is a case-by-case study. Regulations may result in labeling or restricted use (e.g. in hospitals, insurgery, etc.) (Straub, 2002). Due to the scarcity of ecotoxicological data and the presence of pharmaceuticals in water, the primary question is whether medical substances at low environmentally realistic concentrations (parts per billions or trillions) will have any effect at all on different trophic levels and/or on ecosystems. This analysis combines the findings of pharmaceuticals in surface waters in the US (Kolpin et al., 2002) with those found in the EU reported in surveys and reviews by Richardson and Bowron (1985), Halling-Srensen et al. (1998), Daughton and Ternes (1999), Ayscough et al. (2000), Jones et al. (2001), Ku mmerer (2001), Halling-Srensen et al. (2002) and Sturer-Lauridsen et al. (2002). We performed an ecotoxicological (Quantitative) Structure Activity Relationship ((Q)SAR) screening (ECOSAR) of all the compounds reported in the aquatic environment in an attempt to frame the above question. We include; MECs and the effect concentration where 50% of the organisms either die or in other ways are adversely impaired (EC50). Covered are values for fish (96 h and 14 days), daphnids (48 h and 21 days) and algae (48 h), chronic effects values are included when available ( /75%) from the ECOSAR for all model species. Effect measures in the ECOSAR are based on data reported by the industry to the OECD or USEPA, and are all according to USEPA toxicity test guidelines for algae (typically growth inhibition of Selenastrum capricornutum ; lethality and reproduction of D. magna, and fish

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

385

Fathead minnows (Pimephales promelas ) personal communication, Nabholz, 2003). The exact toxic mode of action of the pharmaceuticals to non-target test organisms is not known nor accounted for in the SARs. The specificity of the pharmacodynamic activity and the ecotoxicological mode of action of pharmaceuticals does not easily translate into an ecotoxicological mode of action of pharmaceuticals. The concentrations of pharmaceuticals needed to elicit intended pharmacodynamic responses will exceed environmental concentrations by factors in the range of 104 /106 (Seiler, 2002). Furthermore, if the effect is driven by receptors that may be lacking in non-target organisms, concentrations needed to evoke any effect may then be even higher: as, e.g. serotonin reuptake inhibitors, beta-blockers etc. are not present in most plants and insects. However, more basic mechanisms of cellular functions like those connected with signal transduction or cell division that are generally well conserved in evolution and can thus be identified throughout the living world from unicellular to mammal organisms are targeted by more recently developed pharmaceuticals (Seiler, 2002). Due to the data scarcity and unknown risks associated with pharmaceuticals in the environment the European Commission Scientific Committee on Toxicity, Ecotoxicity and the Environment (CSTEE) recognizes that a prioritization procedure needs to be developed for pharmaceuticals and their environmental risk assessment. To ensure harmonization, this should follow the general scheme for chemicals as described in the White Paper for future chemicals strategy (EU, 2001a). The main tool for prioritization stressed therein is the use of QSARs (EU, 2001b). The most extensively validated and used QSAR is the USEPA EPIWIN suit with ECOSAR. ECOSAR predictions does not replace the need for experimental assessment of the environmental risks posed by pharmaceuticals, but can serve as an initial prioritization tool to estimate potential hazards of pharmaceuticals in the environment. ECOSAR has previously been successfully (low false negative rates) applied to screening pharmaceuticals (Jones et al., 2002) and other complex

compounds such as fragrance materials (Salvito et al., 2002).

2. Methods 2.1. ECOSAR The SARs in the ECOSAR are used to predict the aquatic toxicity of chemicals based on the similarity of structure to chemicals for which the aquatic toxicity has been previously measured. Since 1981, the US Environmental Protection Agency has used SARs to predict the aquatic toxicity of new industrial chemicals in the absence of test data. The acute toxicity of a chemical to fish (both fresh and saltwater), water fleas (daphnids), and green algae has been the focus of the development of SARs. These organisms are group model-organisms and thus not specific species. SARs are developed for chemical classes based on measured test data that have been submitted by industry or they are developed and structural similarities. Using the measured aquatic toxicity values and Kow values, regression equations (currently more than 150 for more than 50 chemical classes) can be developed for a class of chemicals. Inserting the Kow into the regression equation and correcting the resultant value for the molecular weight of the compound may then calculate toxicity values for new and similar yet nonassessed chemicals (Nabholz, 2001). The ECOSAR class program is a computerized version of the ECOSAR analysis procedure as currently practiced by the Office of Pollution Prevention and Toxics (OPPT). It has been developed within the regulatory constrain of the Toxic Substances Control Act (TSCA) and is a pragmatic approach to SAR as opposed to a theoretical approach (Meyland and Howard, 1998). The ECOSAR program can freely be downloaded via the USEPA (http://www.epa.gov/oppt/exposure/docs/episuitedl.htm). A validation assessment of ECOSAR predictions has been performed and they indicate an 87 /90% agreement between predictions and measured data for more than 2000 different chemicals and with B/3% false negatives (Nabholz, 2001).

386

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

2.2. Probabilistic risk assessment Probabilistic risk assessment (PRA) methods are being assessed and considered for incorporation into risk assessment procedures in a number of regulatory jurisdictions. The method used in this analysis has been implemented by the USEPA (Hendly and Giddings, 1999). The use of PRA recognizes that there are no absolutes in risk assessment. Instead there are continuums of potential exposure and effect situations and a range of certainty, which can be reported (Solomon et al., 2000). This is the same in any risk assessment ranging from economic, ecological, or nuclear power plant safety. Implicit in the term risk (or in positive terms chance) dwells a distribution of probabilities, which can be more or less thoroughly elucidated (Bernstein, 1996). The use of distribution curves for exposure (MECs) and toxicity (in this case ECOSAR estimated EC50 values for fish, daphnids and algae) allows the application of joint probability method to describe the nature of risks posed by the MECs of pharmaceuticals and the estimated effect concentrations. The straight-line transformations are converted by probit transformation for the probability (ranked percentages) (1. axis) versus logtransformation of the concentrations (2. axis). The analysis was preformed in an EXCEL spreadsheet designed to perform the double probability ecotoxicology risk assessment procedures outlined in Solomon et al. (2000). Overlap between the measured concentrations and the estimated effect concentrations are then indicative of the existence of risk. 2.3. Hazard quotients We also calculated the hazard quotient (HQ) (MEC/EC50) for the compounds: Values B/1 indicate an insignificant risk and no need for further risk assessments in a tiered procedure (Maund et al., 2001), depending on the assessment factors that are being applied, whose values vary between the US and the EU. A conservative assessment factor of 106 has been proposed to apply for ECOSAR predictions of fragrances by Salvito et al. (2002), whereas the USEPA typically

applies an assessment factor of 2 /10 to ECOSAR predictions. The application of assessment factors is a risk management decision outside the scope of this paper, which focuses on the characterization of environmental hazards associated with pharmaceuticals reported in surface waters; assessment factors is thus not considered. 2.4. Conservative approach The highest environmental concentration of pharmaceuticals found in water, and the lowest effect concentration from the ECOSAR estimations, respectively, were used in the analysis, to secure homogeneity and initially optimal conservatism throughout the test. A few chemicals found in the aquatic environment could not be estimated in ECOSAR due to lack of SMILES notation for the compounds (Meyland and Howard, 1998). The precision of the ECOSAR predictions increases for compounds where ECOSAR identifies an SAR that allows assessment of excess toxicity beyond the narcotic toxicity towards aquatic organisms (personal communication, Nabholz, 2002). This was the case for !/90% of the compounds scanned. Most modern pharmaceuticals are optimized for a specific pharmacodynamic modes of action, which the ECOSAR will not identify, pharmaceutical targets, e.g. membranes, enzymes, or bacterial components, are not restricted to mammalian physiology, as many of these are ubiquitously present on many levels of biological organization (Seiler, 2002). Furthermore, the available experimental ecotoxicological data for human and veterinary pharmaceuticals were compared with the ECOSAR estimates. The chemicals White Paper (EU, 2001b), stresses the foundation and application of the precautionary principle, therefore, we used only the lowest EC50 predictions and the highest MECs including inside sewage treatment plant concentrations in this analysis.

3. Results Table 1 illustrates the available concentration data of pharmaceuticals in environmental water samples from the EU and the US ( /2002) in the

Table 1 ECOSAR scan of environmentally occurring pharmaceuticals in Europe and North America Compound CAS# Water concentration Fish Daphnid Algae log Kow HQ Fish EU Acetemineophen Acetylsalicylic acid Betaxolo Bezafibrate Bisprolol Carazolol Carbamazepine Clenbuterol Clofibrate Clofibric acid Cyclophosphamide Dextropropoxyphene Diatrizoate Diazepam Diclofenac Dimethylaminophenazone 17a-Etinylestradiol Fenofibrate Fenofibric acid Fenoterol Gemfibrozil Ibuprofen Ifosfamide Indomethacine Iopamidol Iopromide Ketoprofen Methaqualone Methotrexate Morphine Metoprolol Naproxen Paracetamol Phenazone Propranolo Propyphenazone Albuterol Salicylic acid Terbutaline Theophylline 3,4,5,6-Tetrabromo-o -cresol Timolol 103-90-2 50-78-2 63659-18-7 41859-67-0 66722-44-9 57775-29-8 298-46-4 37148-27-9 637-07-0 882-09-7 50-18-0 1639-60 737-31-5 439-14-5 15307-79-6 58-15-1 57-63-6 49562-28-9 42017-89-0 13392-18-2 25812-30-0 15687-27-1 3778-73-2 53-86-1 60166-93-0 73334-07-3 22071-15-4 72-44-6 59-05-2 57-27-2 37350-58-6 22204-53-1 103-90-2 60-80-0 525-66-6 479-92-5 18559-94-9 69-72-7 23031-25-6 58-55-9 576-55-6 26839-75-8 0:006 0:0015 0.00019 /0:0046 /0:0029 0.00012 /0:0063 0.00005 0.00004 /0:0016 0.00002 /0:001 0.00023 0.00004 /0:0012 0.00034 0.000043 0.00003 0.00028 0.00006 /0:0015 0.00053 /0:00191 0.0002 /0:015 /0:011 0.00012 /0:001 0.00001 0.0009 /0:022 0.00039 /0:014 0.00095 0.00059 /0:01 0.000035 /0:041 0.00012 /0:001 /0:0001 0.00001
/ /

Daphnid 1.46 )/10 ( 4 1.69 )/10 ( 7 1.27 )/10 ( 4 1.84 )/10 ( 4 3.63 )/10 ( 4 4.80 )/10 ( 5 5.68 )/10 ( 5 2.50 )/10 ( 5 6.15 )/10 ( 6 5.46 )/10 ( 6 1.11 )/10 ( 8 4.17 )/10 ( 5 4.71 )/10 ( 10 2.00 )/10 ( 5 2.37 )/10 ( 7 4.10 )/10 ( 5 2.05 )/10 ( 5 8.57 )/10 ( 5 7.37 )/10 ( 6 3.43 )/10 ( 6 2.50 )/10 ( 4 1.39 )/10 ( 5 1.06 )/10 ( 6 7.69 )/10 ( 6 2.04 )/10 ( 8 1.44 )/10 ( 9 4.84 )/10 ( 7 6.67 )/10 ( 4 1.54 )/10 ( 8 2.81 )/10 ( 5 2.75 )/10 ( 3 2.60 )/10 ( 5 3.41 )/10 ( 4 1.42 )/10 ( 4 2.57 )/10 ( 4 2.86 )/10 ( 3 1.17 )/10 ( 6 6.95 )/10 ( 4 4.44 )/10 ( 6 1.10 )/10 ( 5 1.43 )/10 ( 4 1.11 )/10 ( 6

Algae 2.40 )/10 ( 6 2.46 )/10 ( 5 4.32 )/10 ( 5 2.56 )/10 ( 4 2.07 )/10 ( 4 2.00 )/10 ( 5 9.00 )/10 ( 5 5.00 )/10 ( 6 8.00 )/10 ( 5 8.33 )/10 ( 6 1.82 )/10 ( 6 1.00 )/10 ( 3 9.13 )/10 ( 10 7.27 )/10 ( 6 4.12 )/10 ( 7 2.62 )/10 ( 4 2.15 )/10 ( 5 3.00 )/10 ( 4 1.08 )/10 ( 5 2.40 )/10 ( 6 3.75 )/10 ( 4 2.04 )/10 ( 5 1.74 )/10 ( 4 1.11 )/10 ( 5 3.97 )/10 ( 8 2.97 )/10 ( 9 7.32 )/10 ( 7 1.00 )/10 ( 3 5.21 )/10 ( 8 2.31 )/10 ( 5 1.57 )/10 ( 3 1.77 )/10 ( 5 5.49 )/10 ( 6 8.64 )/10 ( 4 1.07 )/10 ( 4 1.00 )/10 ( 2 9.72 )/10 ( 7 8.54 )/10 ( 4 3.75 )/10 ( 6 1.33 )/10 ( 5 7.69 )/10 ( 4 6.45 )/10 ( 7

258.00 796.00 20.00 5.30 113.00 31.00 101.00 30.00 5.00 53.00 70.00 13.00 6.14 )/105 28.00 532.00 3.70 2.10 0.80 7.70 20.00 0.90 5.00 140.00 3.90 8.66 )/105 8.65 )/106 32.00 1.10 3.83 )/105 257.00 116.00 34.00 40.00 3.00 29.50 0.80 38.00 1.28 1.05 1679.00 0.01 126.00

41.00 8858.00 1.50 25.00 8.00 2.50 111.00 2.00 6.50 293.00 1795.00 24.00 4.88 )/105 2.00 5057.00 8.30 2.10 0.35 38.00 17.50 6.00 38.00 1795.00 26.00 7.35 )/105 7.66 )/106 248.00 1.50 651.00 32.00 8.00 15.00 41.00 6.70 2.30 3.50 30.00 59.00 27.00 91.00 0.70 9.00

2549.00 61.00 4.40 18.00 14.00 6.00 70.00 10.00 0.50 192.00 11.00 1.00 2.52 )/105 5.50 2911.00 1.30 2.00 0.10 26.00 25.00 4.00 26.00 11.00 18.00 3.78 )/105 3.70 )/106 164.00 1.00 192.00 39.00 14.00 22.00 2549.00 1.10 5.50 1.00 36.00 48.00 32.00 75.00 0.13 15.50

0.27 1.13 2.98 /4:25 1.84 2.66 2.25 2.00 /3:62 2.84 0.97 /3:20 (/1.28 2.70 0.57 0.60 /4:12 /5:19 /4:00 1.22 /4:77 /3:79 0.97 /4:23 (/1.38 (/2.49 /3:00 /4:33 (/1.28 0.72 1.69 /3:10 0.27 0.59 2.60 2.05 0.64 2.24 0.67 (/0.39 /5:62 1.75

2.33 )/10 ( 5 1.88 )/10 ( 6 9.50 )/10 ( 6 8.68 )/10 ( 4 2.57 )/10 ( 5 3.87 )/10 ( 6 6.24 )/10 ( 5 1.67 )/10 ( 6 8.00 )/10 ( 6 3.02 )/10 ( 5 2.86 )/10 ( 7 7.69 )/10 ( 5 3.75 )/10 ( 10 1.43 )/10 ( 6 2.26 )/10 ( 6 9.19 )/10 ( 5 2.05 )/10 ( 5 3.75 )/10 ( 5 3.64 )/10 ( 5 3.00 )/10 ( 6 1.67 )/10 ( 3 1.06 )/10 ( 4 1.36 )/10 ( 5 5.13 )/10 ( 5 1.73 )/10 ( 8 1.27 )/10 ( 9 3.75 )/10 ( 6 9.09 )/10 ( 4 2.61 )/10 ( 11 3.50 )/10 ( 6 1.90 )/10 ( 4 1.15 )/10 ( 5 3.50 )/10 ( 4 3.17 )/10 ( 4 2.00 )/10 ( 5 1.25 )/10 ( 2 9.21 )/10 ( 7 3.20 )/10 ( 2 1.14 )/10 ( 4 5.96 )/10 ( 7 8.33 )/10 ( 4 7.94 )/10 ( 8

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395 387

388

Table 1 (Continued ) Compound CAS# Water concentration Fish Daphnid Algae log Kow HQ Fish Tolfenamic acid USA Chlorotetracycline Ciprofloxacin Erythomycin Norfloxacin Lincomycin Oxytetracycline Roxithromycin Sulfadimethoxine Sulfamethazine Sulfamethizole Sulfamethaxazole Tetracycline Trimethoprim Tylosin Cimetidine Codeine Diltiazem Enalaprilat Fluoxetine Gemfibrozil Metformin Ranitidine Acetaminophen Caffeine Cotinine 1,7-Dimethylxanthine Ibuprofen Acetophenone 17a-Etynyl-estradiol 13710-19-5 57-62-5 85721-33-1 114-07-8 70458-96-7 154-21-2 79-57-2 80214-83-1 122-11-2 57-68-1 144-82-1 723-46-6 60-54-8 738-70-5 1401-69-0 51481-61-9 76-57-3 42399-41-7 76420-72-9 54910-89-3 25812-30-0 657-24-9 66357-35-5 103-90-2 58-08-2 486-56-6 611-59-6 15687-27-1 98-86-2 57-63-6
/

Daphnid
(3

Algae
(4

0:0016

0.40 1.39 )/105 2.46 )/105 61.00 1.40 )/104 1391.00 1.66 )/105 50.00 226.00 517.00 1113.00 890.00 16.00 795.00 27.40 571.00 238.00 23.00 7.30 )/104 1.70 0.90 3.32 )/104 1076.00 1.00 805.00 4747.00 1679.00 5.00 181.00 2.10

1.70 357.00 991.00 7.80 1449.00 82.00 2432.00 6.00 3.50 4.00 5.00 4.50 550.00 4.80 66.00 35.00 16.00 2.90 3690.00 0.17 6.00 1345.00 63.00 42.00 46.00 4535.00 91.00 38.00 190.00 2.10

1.30 267.00 938.00 4.30 1232.00 86.00 2294.00 4.00 24.00 38.00 60.00 51.00 475.00 2.60 16.00 40.00 23.00 1.90 2523.00 0.80 4.00 511.00 66.00 2549.00 46.00 2577.00 75.00 26.00 116.00 2.00

5:38

4.00 )/10

9.41 )/10

1.23 )/10 ( 3 2.58 )/10 ( 6 3.20 )/10 ( 8 3.95 )/10 ( 3 9.74 )/10 ( 8 8.49 )/10 ( 6 5.23 )/10 ( 8 4.50 )/10 ( 5 2.50 )/10 ( 6 3.16 )/10 ( 6 2.17 )/10 ( 6 3.73 )/10 ( 4 2.32 )/10 ( 7 2.73 )/10 ( 4 1.75 )/10 ( 5 1.45 )/10 ( 5 4.35 )/10 ( 5 2.58 )/10 ( ( 5 1.82 )/10 ( 8 1.50 )/10 ( 5 1.98 )/10 ( 4 2.94 )/10 ( 7 1.52 )/10 ( 7 3.92 )/10 ( 6 1.30 )/10 ( 4 3.49 )/10 ( 7 1.47 )/10 ( 6 3.85 )/10 ( 5 1.29 )/10 ( 6 4.16 )/10 ( 4

0.00069 0.00003 /0:017 0.00012 0.00073 0.00012 0.00018 0.00006 0.00012 0.00013 /0:019 0.00011 0.00071 0.00028 0.00058 /0:001 0.000049 0.000046 0.000012 0.00079 0.00015 0.00001 /0:01 /0:006 0.0009 0.00011 /0:001 0.00015 0.000831

(/0.68 0.01 2.48 (/0.31 0.29 (/2.87 2.72 1.17 0.76 0.41 0.48 (/1.33 0.73 1.05 0.57 1.28 2.79 (/0.94 /4:65 /4:77 (/2.64 0.29 0.27 0.16 0.34 (/0.39 /3:79 1.67 /4:12

4.96 )/10 ( 9 1.22 )/10 ( 10 2.79 )/10 ( 4 8.55 )/10 ( 9 5.25 )/10 ( 7 7.23 )/10 ( 10 3.60 )/10 ( 6 2.65 )/10 ( 7 2.32 )/10 ( 7 1.17 )/10 ( 7 2.13 )/10 ( 5 6.88 )/10 ( 6 8.93 )/10 ( 7 1.02 )/10 ( 5 1.02 )/10 ( 6 4.20 )/10 ( 6 2.13 )/10 ( 6 6.30 )/10 ( 10 7.06 )/10 ( 6 8.78 )/10 ( 4 4.52 )/10 ( 9 9.29 )/10 ( 9 1.00 )/10 ( 2 7.45 )/10 ( 6 1.90 )/10 ( 7 6.55 )/10 ( 8 2.00 )/10 ( 4 8.29 )/10 ( 7 3.96 )/10 ( 4

1.93 )/10 ( 6 3.03 )/10 ( 8 2.18 )/10 ( 3 8.28 )/10 ( 8 8.90 )/10 ( 6 4.93 )/10 ( 8 3.00 )/10 ( 5 1.71 )/10 ( 5 3.00 )/10 ( 5 2.60 )/10 ( 5 4.22 )/10 ( 3 2.00 )/10 ( 7 1.48 )/10 ( 4 4.24 )/10 ( 6 1.66 )/10 ( 5 6.25 )/10 ( 5 1.69 )/10 ( 5 1.25 )/10 ( 8 7.06 )/10 ( 5 1.32 )/10 ( 4 1.12 )/10 ( 7 1.59 )/10 ( 7 2.38 )/10 ( 4 1.30 )/10 ( 4 1.98 )/10 ( 7 1.21 )/10 ( 6 2.63 )/10 ( 5 7.89 )/10 ( 7 3.96 )/10 ( 4

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

C(/1 (underlined indicate should be risk assessed by FDA ( !/1 mg l ( 1 0/ !/0.001 mg l ( 1) all exceeded the EU cut off of 0.000001 mg l ( 1) and the estimated log Kow (critical value for significant bioaccumulation potential !/3, underlined) the compound name is underlined if both criteria are fulfilled at the same time. The final row is the HQ (MEC/EC50) for the three endpoints, ranked in Fig. 2.

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

389

Fig. 1. Displays the percent rank distribution of the environmental concentrations from the literature and the effect concentrations in mg l ( 1 derived from the ECOSAR screening.

open literature (Richardson and Bowron, 1985; Halling-Srensen et al., 1998; Daughton and Ternes, 1999; Ayscough et al., 2000; Jones et al., 2001; Ku mmerer, 2001; Halling-Srensen et al., 2002; Sturer-Lauridsen et al., 2002; Kolpin et al., 2002; Schulman et al., 2002). The list depicts the differences found in the EU and the US for compounds of potential need for environmental risk assessment and for significant bioaccumulation. The differences detected for certain compounds between the US and the EU surveys, could either indicate significant differences in use patterns, wastewater treatment, manure and sludge management, environmental conditions or simply, and most likely, differences in analytical focus. Twenty-three of the substances were in worstcase scenarios detected at levels !/1 mg l (1, thus fulfilling the FDA the requirements for an environmental risk assessment. All of them fulfilled the EU criteria of 0.01 mg l(1, which is virtually the detection limit. Thirteen had potential for significant bioaccumulation (log Kow !/3). Bezafibrate, Dextropropoxyphene, Gemfibrozil, Ibuprofen,

Methaqualone and Tolfenamic acid (or 8%) fulfilled both risk characteristics. Fig. 1 illustrates that acute risks are probably not significant, as there is no overlap between the distribution of the MECs and the estimated EC50 values. There are 1 /2 orders of magnitude in difference between the environmental concentrations and the effect concentrations at the level where 10% (10th centile) of the compounds effect concentrations would be exceeded 5% (95th centile) of the time. The probability of the highest MEC exceeding the 10th centile of EC50s for fish, daphnids and algae are all 0.3%. In a theoretical worst-case scenario there might be an overlap of probabilities for the lowest fish EC50 and the least frequent and highest environmental concentration. 3,4,5,6-Tetrabromo-o -cresol has the lowest EC50 for fish, 0.01 mg l (1 while its environmental concentration is 0.0001 mg l(1 (see Table 1 and Fig. 2), so the theoretical overlap is not a consistent risk, before application of assessment factor.

390

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

Fig. 2. Graphical illustration of ranked HQs (MEC/EC50) for pharmaceuticals reported in the environment for sh, daphnids and algae. More than 99.9% of the HQs were B/1 with an approximately median of 10 ( 5, before application of an arbitrary assessment factor.

On average, the order of susceptibility among the three endpoints was: algae ]/daphnids !/fish. Note that the cytostatic pharmaceuticals (Cyclophosphamide and Ifosfamide) represent a special high risk for mammals and potentially other trophic levels in the environment, as these are known to be carcinogenic, mutagenic and embryotoxic (Ku mmerer, 2001). More than 99.9% of the ranked HQs (Maund et al., 2001) in Fig. 2 are less than 1, before application of an assessment factor. On average, the MECs were five orders of magnitude smaller than the related effect concentrations. This indicates no significant environmental risks based on the ECOSAR estimates and the available MECs. For the exact HQs for each individual compound and endpoint see Table 1. Because most (Q)SAR models (including ECOSAR) uses lipophilicity, plus additional excess toxicity due to structure to develop models to predict toxicity these models should, when possible be authenticated by comparing modeled versus

measured data (Nabholz, 2001). In this case it maybe extra important as 82% of the compounds were hydrophilic, thus we compared the modeled predictions with data from the open literature. Figs. 3 /5 is a graphically representation of modeled versus experimental EC50 data for fish, daphnids and algae (Halling-Srensen et al., 2002; Sturer-Lauridsen et al., 2002; Wilson et al., 2002; Johnson et al., 2002) for 20 different pharmaceuticals reported to occur in surface waters that have experimental data for either fish, daphnids and/or algae. In 80% of the cases where both measured and modeled data were available, the ECOSAR EC50 estimations were the lower (or over-protective) than the measured effect concentration. Cleuvers (2003) found that for all endpoints and compounds he tested the QSAR derived EC50 predictions were lower than the measured EC50 values, even though only Clofibrinic acid had a log Kow !/3. He concluded that the compounds he worked with all acted unspecifically by non-polar narcosis and that

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

391

Fig. 3. Measured sh effect concentration from the literature vs. ECOSAR estimated sh effect concentration for 20 different pharmaceuticals.

toxicity thus may be associated with log Kow rather than any specific toxic action in the nontarget organism (Cleuvers, 2003).

4. Discussion The HQs were derived by comparing the highest MECs from the literature with the lowest ECOSAR prediction, indicating low acute risk to aquatic organisms (median HQ :/10 (5). However, if an assessment factor of 1000, as advised in the EU, is applied to the (Q)SAR predictions 14% of the compounds HQ would exceed 1 and require further testing, which is consistent with findings of 13% for high volume pharmaceuticals exceeding 1 (EU, 2001a). Caution, due to uncer-

tainty connected to the regressions in the ECOSAR, has been raised by Kaiser et al. (1999). In 80% of the cases where both an experimental and modeled effect concentration were available, the estimated values were lower than the corresponding lowest measured effect concentration (see Figs. 3 /5). Intra- and inter-laboratory variability of standard single species toxicity tests needs to be taken into account when assessing the sensitivity and quality of SAR estimates versus experimental values. Personne and Janssen (1994) have found that the average coefficient of variation (CV 0/ S.D./mean )/100) in single species laboratory bioassays exceeds 25% and can be as high as 50% in some cases. Furthermore, changes in laboratory environmental factors such as temperature, light or pH can modulate the toxicity of

392

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

Fig. 4. Measured daphnids effect concentration from the literature vs. ECOSAR estimated daphnids effect concentration for 20 different pharmaceuticals.

compounds by up to two orders of magnitude (Personne and Janssen, 1994). The EU-Directive 93/67/EEC classifies compounds according to their EC50 values: B/0.1 mg l(1 0/extremely toxic to aquatic organisms; 0.1 /1 mg l (1 0/very toxic to aquatic organisms; 1 /10 mg l (1 0/toxic to aquatic organisms; 10 /100 mg l(1 0/harmful to aquatic organisms; B/100 mg l (1 0/non-toxic to aquatic organisms. According to EU-Directive 93/67/EEC, the anti-fungal 3,4,5,6-tetrabromo-o -cresol was extremely toxic; 12% of all the compounds found in surface waters were toxic, 41% were harmful and 47% were nontoxic. Thus despite low risks, more than half of these pharmaceuticals may due to their intrinsic toxicity cause unwanted harm in aquatic environ-

ments and are liable to be labeled N;R50/53 (Dangerous for the environment; very toxic to aquatic organisms, may cause long-term adverse effects in the aquatic environment) or R52/53 (Harmful to aquatic organisms, may cause longterm adverse effects in the aquatic environment) according to EU-Directive 93/67/EEC). The current battery of ecotoxicological testing of chemicals is not tailored for a risk assessment of pharmaceuticals in terms of mechanistic knowledge and statistical analysis in terms of replication and statistical power (Weiss, 1998). As most pharmaceuticals are designed to affect mammalian physiology, it is not known what effects they could have on other forms of life, e.g. aquatic fauna or plants (Seiler, 2002). Knowledge of the availability

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395

393

Fig. 5. Measured algae effect concentration from the literature vs. ECOSAR estimated algae effect concentration for 20 different pharmaceuticals.

of pharmaceuticals to cellular targets is required for an effective risk assessment. Due to the continued low exposure the effects of most interest will be chronic and subtle effects on the organisms function, reproduction, behavior, metabolism, genotoxicity etc. (Jones et al., 2001). The current protocols are primarily designed to suit other chemicals such as pesticides, where acute effects on algae, aquatic insects or fish are expected because the ecotoxicological modes of action of the compounds are better elucidated. Moreover, the risk management and risk communication process is also more complex for pharmaceuticals than other chemicals as pharmaceuticals intuitively are perceived as good, therapeutic compounds, and the environmental risks are easily

outweighed, as pharmaceuticals benefits to humans are a greater priority (Henschel et al., 1997).

5. Conclusions Due to the low MECs acute risks are not likely, simple extrapolation of effects from higher concentrations does not necessarily have relevance at lower concentrations. !/50% of the reported pharmaceuticals were intrinsically toxic potentially leading to the necessity for labeling in the EU. The complicated issue of mixtures and additive, synergistic or antagonistic effects need to be addressed (Cleuvers, 2003) along with assessment of chronic, population and ecosystem effects. Without these

394

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395 EEC Directive 93/67. Commission of the European Communities, 1996. Technical guidance document in support of commission directive 93/67/EEC on risk assessment for new notied substances and commission regulation (EC) No. 1488/94 on risk assessment for existing substances. Part II: Environmental Risk Assessment. Luxembourg. EU, 2001a. CSTEE. Discussion paper on environmental risk assessment of medical products for human use (nongenetically modied organisms (non-GMO) containing). CPMPpaperRAssessHumPharm12062001/D(01). Brussels, Belgium. EU, 2001b. White paper. Strategy for a future chemicals policy, COM (2001) 88 Final. http://europa.eu.int/comm/environment/chemicals/0188_en.pdf (accessed 28 March 2003). Hendly, P., Giddings, J.M., 1999. ECOFRAM. Aquatic guidance report. http://www.Aqex_ecofram_peer01_may499.doc. http://www.epa.gov/oppefed1/ecorisk/index.htm. Halling-Srensen, B., Nors-Nielsen, S., Lansky, P.F., Ingerslev, F., Holten-Lu tzhft, H.C., Jrgensen, S.E., 1998. Occurrence, fate and effects of pharmaceutical substances in the environment */a review. Chemosphere 36, 357 /393. Halling-Srensen, B., Nors-Nielsen, S., Jensen J., 2002. Environmental assessment of veterinary medicinal products in Denmark. Environmental Project No. 659. Danish Environmental Protection Agency. Henschel, K.P., Wenzel, A., Diedrich, M., Fliedner, A., 1997. Environmental hazard assessment of pharmaceuticals. Regul. Toxicol. Pharmacol. 25, 220 /225. Johnson, DJ., Brain, R.A., Richards, S.M., Wilson, C.J., Mabury, S.A., Lam, M., Sibley, P.K., Solomon, K.R., 2002. Ecotoxicology of pharmaceutical mixtures in aquatic microcosms part IV: responses in freshwater sh. SETAC platform presentation 064, SETAC NA 23rd annual meeting 16 /20 November 2002. Jones, O.A.H., Voulvoulis, N., Lester, J.N., 2001. Human pharmaceuticals in the aquatic environment */a review. Environ. Technol. 22, 1383 /1394. Jones, O.A.H., Voulvoulis, N., Lester, J.N., 2002. Aquatic environmental assessment of the top 25 English prescription pharmaceuticals. Water Res. 36, 5013 /5022. Kaiser, K.L.E., C.J. Dearden, W. Klein, T.W. Schultz, 1999. A note of caution to users of ECOSAR. Water Qual. Res. Jour. of Canada 34, 179 /182. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S.D., Barber, L.R., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999 /2000: a national reconnaissance. Environ. Sci. Technol. 36, 1202 /1211. Ku mmerer, K., 2001. Drugs in the environment: emission of drugs, diagnostic aids and disinfectants into wastewater by hospitals in relation to other sources */a review. Chemosphere 45, 957 /969. Maund, S.J., Travis, K.Z., Hendley, P., Giddings, J.M., Solomon, K.R., 2001. Probabilistic risk assessment of cotton pyrethroids: V. Combining landscape-level exposures

analyses and careful consideration of the statistical power and detectability of the test (Sanderson and Petersen, 2002), it would be unwise as well as statistically and scientifically false to conclude that pharmaceuticals are not causing effects in the environment at all (Jones et al., 2001). The present analyses indicate that the regulatory and risk management context concerning pharmaceuticals in the aquatic environment is more complicated than risk management of other chemicals. The uncertainty concerning pharmaceutical mode of action in environmentally relevant non-target organisms, mixture interactions, degradation products, bioavaliability, low acute risks, but intrinsic toxicity and bacterial resistance contributes to the risk management challenges. Ultimately, the environmental risk assessment and management framework for pharmaceuticals must balance these uncertainties on a case-by-case basis against the human health benefits of pharmaceuticals. QSARs can be used as a prioritization tool for the risk assessment and management of pharmaceuticals.

Acknowledgements The authors greatly acknowledge The Canadian Network of Toxicology Centers for supporting this work.

References
Ayscough, N.J., Fawell, J., Franklin, G., Young, W., 2000. Review of Human Pharmaceuticals in the Environment. The Environmental Agency R&D dissemination centre, Wilts, UK. Bernstein, P.L., 1996. Against the Gods the Remarkable Story of Risk. Wiley, New York, USA. FDA, 1995. Centre for drug evaluation and research (CDER). Guidance for industry for the submission of an environmental assessment in human drug applications and supplements. FDA, Rockville, USA. Cleuvers, M., 2003. Aquatic ecotoxicity of pharmaceuticals including the assessment of combination effects. Toxicol. Lett. 142 (3), 185 /194. Daughton, C.G., Ternes, T.A., 1999. Special report. Pharmaceuticals and personal care products in the environment: agents of subtle change. Environ. Health Pers. 107, 907 / 938.

H. Sanderson et al. / Toxicology Letters 144 (2003) 383 /395 and ecotoxicological effect data to characterize risks. Environ. Toxicol. Chem. 20, 687 /692. Meyland, W.M., Howard, P.H., 1998. Users Guide for the ECOSAR Class Program. USEPA, North Syracuse, New York, USA. Nabholz, V.J., 2001. SAR validation study. Poster presentation at Society of Toxicology annual meeting, March 2001 in San Francisco, USA. Personne, G., Janssen, C.R., 1994. Field validation of predictions based on laboratory toxicity tests. In: Graney, R.L., Kennedy, J.H., Rodgers, J.H. (Eds.), Aquatic Mesocosm Studies in Ecological Risk Assessment. CRC Press, USA (Chapter 25). Richardson, M.L., Bowron, J.M., 1985. The fate of pharmaceutical chemicals in the aquatic environment */a review. J. Pharm. Pharmacol. 37, 1 /12. Salvito, D.T., Senna, R.J., Federle, T.W., 2002. A framework for prioritizing fragrance materials for aquatic risk assessment. Environ. Toxicol. Chem. 21, 1301 /1308. Sanderson, H., Petersen, S., 2002. Power analysis as a reexive scientic tool for interpretation and implementation of the precautionary principle in the European Union. Environ. Sci. Pollut. Res. 9, 221 /226. Schulman, L.J., Sargent, E.V., Naumann, B.D., Faria, E.C., Dolan, D.G., Wargo, J.P., 2002. A human health risk

395

assessment of pharmaceuticals in the aquatic environment. Hum. Ecol. Risk Assess. 8, 657 /680. Seiler, J.P., 2002. Pharmacodynamic activity of drugs and ecotoxicology */can the two be connected. Toxicol. Lett. 131, 105 /115. Solomon, K.R., Giesy, J.P., Jones, P., 2000. Probabilistic risk assessment of agrochemicals in the environment. Crop Protec. 19, 649 /655. Straub, J.O., 2002. Environmental risk assessment for new human pharmaceuticals in the European Union according to the draft guideline/discussion paper of January 2001. Toxicol. Lett. 131, 137 /143. Sturer-Lauridsen, F., Hansen, L., Birkved, M., Kjlholt, J., Mikkelsen, S., 2002. Litteraturudredning vedrrende medicin i miljet. Environmental Project No. 661. Danish Environmental Protection Agency. Weiss, B., 1998. A risk assessment perspective on the neurobehavioral toxicity of endocrine disruptors. Toxicol. Ind. Health 14, 1341 /1359. Wilson, C.J., Brain, R.A., Richards, S.M., Johnson, D.J., Mabury, S.A., Lam, M., Sibley, P.K., Solomon, K.R., 2002. Ecotoxicology of pharmaceutical mixtures in aquatic microcosms part II: responses of freshwater plankton communities. SETAC platform presentation 062, SETAC NA 23rd annual meeting 16 /20 November 2002.