Pharmaceuticals, hormones and bisphenol A in untreated source and finished drinking water in Ontario, Canada — Occurrence and treatment efficiency

Science of the Total Environment 409 (2011) 1481–1488

Contents lists available at ScienceDirect

Science of the Total Environment j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s c i t o t e n v

Pharmaceuticals, hormones and bisphenol A in untreated source and finished drinking water in Ontario, Canada — Occurrence and treatment efficiency Sonya Kleywegt a, Vince Pileggi a, Paul Yang b,⁎, Chunyan Hao b, Xiaoming Zhao b, Carline Rocks c, Serei Thach b, Patrick Cheung c, Brian Whitehead c a b c

Standards Development Branch, Ontario Ministry of the Environment, 40 St. Clair Avenue West, Toronto, Ontario, Canada M4V 1M2 Laboratory Services Branch, Ontario Ministry of the Environment, 125 Resources Road, Etobicoke, Ontario, Canada M9P 3V6 Environmental Monitoring and Reporting Branch, Ontario Ministry of the Environment, 125 Resources Road, Etobicoke, Ontario, Canada M9P 3V6

a r t i c l e

i n f o

Article history: Received 7 October 2010 Received in revised form 28 December 2010 Accepted 6 January 2011 Keywords: Hormones Antibiotics Bisphenol A (BPA) Granulated activated carbon Ultraviolet irradiation Removal efficiciency

a b s t r a c t The Ontario Ministry of the Environment (MOE) conducted a survey in 2006 on emerging organic contaminants (EOCs) which included pharmaceuticals, hormones and bisphenol A (BPA). The survey collected 258 samples over a 16 month period from selected source waters and 17 drinking water systems (DWSs), and analyzed them for 48 EOCs using liquid chromatography–tandem mass spectrometry (LC-MS/ MS) and isotope dilution mass spectrometry (IDMS) for the highest precision and accuracy of analytical data possible. 27 of the 48 target EOCs were detected in source water, finished drinking water, or both. DWSs using river and lake source water accounted for N 90% detections. Of the 27 EOCs found, we also reported the first detection of two antibiotics roxithromycin and enrofloxacin in environmental samples. The most frequently detected compounds (≥ 10%) in finished drinking water were carbamazepine (CBZ), gemfibrozil (GFB), ibuprofen (IBU), and BPA; with their concentrations accurately determined by using IDMS and calculated to be 4 to 10 times lower than those measured in the source water. Comparison of plant specific data allowed us to determine removal efficiency (RE) of these four most frequently detected compounds in Ontario DWSs. The RE of CBZ was determined to be from 71 to 93% for DWSs using granulated activated carbon (GAC); and was 75% for DWSs using GAC followed by ultraviolet irradiation (UV). The observed RE of GFB was between 44 and 55% in DWSs using GAC and increased to 82% when GAC was followed by UV. The use of GAC or GAC followed by UV provided an RE improvement of BPA from 80 to 99%. These detected concentration levels are well below the predicted no effect concentration or total allowable concentration reported in the literature. Additional targeted, site specific comparative research is required to fully assess the effectiveness of Ontario DWSs to remove particular compounds of concern. © 2011 Elsevier B.V. All rights reserved.

1. Introduction Trace amounts of emerging organic contaminants (EOCs) such as pharmaceuticals, hormones, and bisphenol A (BPA) can enter the environment from various anthropogenic activities (Daughton and Ternes, 1999; Richardson and Ternes, 2005; Kolpin et al., 2002a). It wasn't until the mid 1990s however, that researchers began to identify and quantify these EOCs in sewage treatment plant (STP) influent and effluent, surface water and drinking water in Europe, the United States, and Canada (Anderson et al., 2004; Ternes, 1998a; Rowney et al., 2009). EOCs that are heavily used and readily available (e.g. those found in consumer products or over the counter prescription medications) have a greater potential to migrate into the natural environment and be detected in environmental media. Some EOCs enter the natural

⁎ Corresponding author. Fax: +1 416 2355900. E-mail address: [email protected] (P. Yang). 0048-9697/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2011.01.010

environment in runoff from non-point sources such as agricultural areas (from manure or biosolid applications) or from point sources such as discharges from municipal sewage treatment plants to surface waters. Once in the natural environment, EOCs may undergo photodegradation or other degradation/transformation processes; thus the proximity of the sources (domestic or agricultural) to drinking water intakes is also a contributing factor in what might be detected in drinking water. Finally, the effectiveness of a drinking water system's treatment process to reduce the levels of EOCs will affect the concentrations found in drinking water. Recent reviews indicate that to date more than 30 different pharmaceuticals or other EOCs of concern have been detected in finished drinking waters world-wide (Daughton and Ternes, 1999; Richardson and Ternes, 2005; Kolpin et al., 2002a; Anderson et al., 2004; Ternes, 1998a; Rowney et al., 2009). The detection of these compounds in drinking water has been attributed to their presence in the source water and the inability of the treatment process at the drinking water system to reduce these EOCs below the current

1482

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

method detection limits (DLs). In order to determine the occurrence and concentrations of selected pharmaceuticals, hormones and BPA in Ontario source water and finished drinking water, the Ontario Ministry of the Environment (MOE) conducted a survey with the goal to establish the occurrence and reliable point estimates of the concentrations of selected EOCs in Ontario during a 16-month monitoring period. A total of 258 field grab samples were collected, 130 from source water and 128 from finished drinking water. Using a liquid chromatography/tandem mass spectrometry (LC/MS-MS) method (Hao et al., 2008; Ministry of the Environment, 2010), 48 selected analytes were investigated with detection limits (DLs) in the ng/L (parts per trillion, ppt) range. We report in this paper results of this survey including concentrations and percent detections for selected pharmaceuticals, hormones and BPA in both source water and finished drinking water. Also, by comparing the distributions of the concentrations of the most frequently detected compounds in source and finished waters, and individual treatment processes for 5 DWSs, we were able to quantify removal efficiency (RE) from selected DWSs.

membrane filtration, carbon adsorption, and disinfection (chlorine, ozone, and ultraviolet irradiation); including the use of granulated activated carbon (GAC), ultraviolet irradiation (UV) and membrane filtration (Table 1). Samples collected ranged from one to 15 samples for each site depending on the site characteristics including source water used and treatment technologies. Field grab samples were collected in pre-cleaned 1-L amber glass bottles, using a Teflon lid cap, and preserved with 250 mg/L of sodium thiosulfate to quench residual chlorine in the finished water. The samples were packed in an ice shuttle, couriered back to the laboratory overnight, and stored in the dark between 2 and 6 °C until ready for extraction (usually within seven days). The source and finished water samples were collected on the same day with the assumption that the source water characteristics would not vary significantly throughout the day and the effect of retention time would be minimal.

2. Experimental

Sample preparation and analysis were done according to the MOE method E3454 (Ministry of the Environment, 2010). Accredited by the Canadian Association for Laboratory Accreditation, method E3454 uses Waters (Millford, MA, USA) hydrophilic–lipophilic balanced solid phase extraction (SPE) cartridges (6 mL, 200 mg) to extract analytes in one single neutral extraction. Sample extracts were separated by an Agilent 1100 LC (Mississauga, Ontario, Canada), and analyzed by an Applied Biosystems API 4000 Q-trap mass spectrometer (Foster City, CA, USA). Samples were prepared in batches of 24 that included four QC samples (laboratory blank and three method spikes) and 20 field samples to maximize operational efficiency. Duplicate pure water method spikes and tap water method spiked samples were used to monitor within-run method precision and matrix effects. Typically, 800 mL of drinking/surface water samples were used in the sample preparation. Prior to extraction, 4 g of EDTANa2, 100 μL of IDQS solution and 20 mL of 0.25 M ammonium acetate solution were added into each sample, homogenized on a roller (Wheaton Science, NJ, USA) for 10 min, and the pH value adjusted to 6.95 ± 0.05 using 10% (w/v) NaOH and 10% (v/v) H2SO4 solution. Wastewater samples were

2.1. Chemicals, isotopically labeled compounds and analytical standards Analytical standards for the 48 EOCs, methanol, acetonitrile, sulphuric acid, sodium hydroxide, heptafluorobutyric acid (HFBA), ammonium acetate (N99%), HPLC grade water and ethylenediaminetetraacetic acid disodium salt (EDTANa2, ACS reagent grade), were purchased from Sigma Aldrich (Oakville, ON, Canada). Fifteen 13C- and 2H-isotope labeled compounds (ILC) were purchased from Cambridge Isotope Laboratories (Andover, MA, USA) and CDN-Isotopes (Montreal, QC, Canada) and were used as surrogates to carry out isotope dilution mass spectrometric (IDMS) analysis. The names of the EOC standards and ILCs are listed in Table S1. High purity water was produced by passing reverse osmosis water through a Barnstead NANOpureTM water purification system (pure water) and used to prepare method blank and method spike samples. Separate stock solutions of analytical standards and ILCs were prepared by weighing approximately 10 mg each compound and dissolving in 10 mL of methanol or methanol:water (50:50/v:v) in calibrated plastic tubes (Simport, QC, Canada). Intermediate standards were prepared in 100 mL volumetric flasks by mixing the stock solutions and were used to prepare five calibration standard and one spiking solutions at conventration levels from 2 ng/mL to 3.5 μg/mL. Isotope dilution quantitation solutions (IDQS) were prepared by mixing stock solutions of ILCs (Table S1, total of 13), diluted in methanol and added into samples before extraction. Recoveries of ILCs in the IDQS were calculated and used to monitor method performance and matrix effects, and carry out IDMS analysis. In addition, injection internal standards (IIS) were also prepared by mixing ILC stock solutions (Table S1, total of three) diluted in HPLC grade water, added into the final sample extract before LC injection, and used to correct variations of sample volume occurring during the final sample concentration and sample injection processes. 2.2. Site selection and sampling Seventeen sampling sites were selected from a cross section of drinking water systems (DWSs) that participate in the drinking water surveillance program (DWSP, http://www.ene.gov.on.ca/en/water/ dwsp/) administered by the provincial Government of Ontario. Sites were chosen to reflect a range of source water types, treatment processes, and proximity to municipal sewage treatment plants and agricultural activities. A total of 17 Ontario DWSs volunteered to participate in the survey which included: 8 surface water sources from rivers; 7 from lake sources; and 2 from groundwater sources. The range of treatment processes used by these DWSs included media and

2.3. Sample preparation and analysis

Table 1 Characteristics of drinking water systems (DWSs) in the study. DWS Source

Filtration

Disinfection

Sampling events

1 2 3 4 5 6 7 8

River River River River River River River River

Sand, Sand, Sand, Sand, Sand, GAC Sand, Sand,

9 15 10 9 10 12 3 5

9 10

Lake Lake

Chlorine, Fluoridation Chlorine Chlorine, Fluoridation Chlorine, Fluoridation Chlorine, Fluoridation Chlorine, UV Chlorine, Pre UV Chlorine, Pre UV and Fluoridation Chlorine Chlorine, Fluoridation Chlorine Chlorine, Fluoridation

5 5

Chlorine, Fluoridation Chlorine, Fluoridation Chlorine

5 10 15

11 12 13 14 15 16 17

AnthrCoal GAC GAC AnthrCoal AnthrCoal AnthrCoal GAC

Sand, AnthrCoal Sand, Gravel, and AnthrCoal Lake Membrane Filtration Lake Sand, Gravel, and AnthrCoal Lake Sand, AnthrCoal Lake Sand, AnthrCoal Mixed Sand, Gravel, and AnthrCoal Ground AnthrCoal, MnSand Ground –

Chlorine Chlorine

5 5

1 4

AnthrCoal, Anthracite Coal; GAC, Granulated Activated Carbon; and MnSand, Manganese Greensand.

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

treated in the same manner except that 500 μL of IDQS solution were used to ensure that recoveries of the IDQS can be determined. Each of the SPE cartridge was conditioned by 5 mL of water, methanol, and finally 5% (v/v) ammonium hydroxide in methanol before the SPE extraction. Upon the completion of the extraction, the SPE cartridges were rinsed with 5 mL of 10% (v/v) methanol/water, dried by air, and analytes eluted from the SPE cartridges with 5 mL methanol. 5 mL of the eluate was evaporated to dryness with N2 at an ambient temperature, and reconstituted by using 0.1 mL IIS solution. Calibration standards were prepared by using a mixture of 0.3 mL IDQS solution and 0.3 mL each of different levels of target compound solutions, evaporating to dryness under N2, and reconstituted to 0.3 mL using the IIS solution before instrumental analysis. Sample extracts were injected into the LC and separated by a Hypersil Gold 3 μm C-18, 100×2.1 mm LC column (Thermo Electron, Bellefonte, PA, USA) in two separate chromatographic runs that correspond to positive and negative ionization LC-MS/MS analysis using, respectively, acidic and neutral LC mobile phases. Column temperature used was 30 °C and the injection volume was 20 μL. Mobile phases used in acidic chromatography analysis were 0.03% heptafluorobutanoic acid in HPLC water (A) and acetonitrile (B). The flow rate used was 0.2 mL/min. The gradient started at 15% for mobile phase B, increased to 100% at 13 min, stayed at 100% for another 2 min, and returned to 15% of B at 17 min. The LC column was then conditioned for another 11 min resulting in a total 28 min run time for the positive mode LC-MS/MS analysis. The negative mode LC separation used 5 mM ammonium acetate in water (A) with pH adjust to 7.0±0.1 with 0.5 M NH4OH and acetonitrile (B) and a flow rate of 0.2 mL/min. The gradient started at 10% for mobile phase B, ramped to 85% B at 15 min, 100% B at 16 min, stayed at 100% for another 2 min, and returned to 10% of B at 18 min. The LC column was then conditioned for another 10 min resulting in a total of 30 min run time for the LC-MS/MS analysis. Multiple reaction monitoring (MRM) data were acquired and processed for all compounds in either positive or negative ion mode. Optimized MRM parameters were obtained by direct infusion with the most intense ion pair of each analyte used for the analysis (Hao et al., 2008; Ministry of the Environment, 2008). Table S1 listed MRM transitions and collision energies used during the analysis. Confirmation of target compounds in field samples was done using retention times obtained from the MRM reconstructed chromatogram and their specific MRM transition which include the precursor and fragment ions. Curtain, collision, nebulizer, and auxiliary gases of the MS-MS were set at 15, 6, 35 and 45 psi, respectively. Source temperature and entrance potential were kept at 450 °C and 10 V for both positive and negative modes. Ion spray voltage, declustering potential, and collision cell exit potential used were 5200, 60 and 10 V for the positive and −4500, −90 and −5 V for the negative ESI, respectively. 2.4. Statistical analysis A significant portion of the analytical results were reported as not detected (ND) or lying in the range from zero to less than the detection limit (b DL). In order to obtain a reliable point estimate of summary statistics, without bias, all b DL sample results were considered using left-censored data analysis which incorporated all the non-detect and detected results. Through the use of regression on order statistics (ROS), point estimates of summary statistics were calculated (Minitab® statistical software, Version 15, http://www. practicalstats.com/nada, 9). This approach extended the calculated point estimates (e.g., median or 95th percentile values) to values below the DL especially when there was a large proportion of nondetect results (Helsel, 2006). 3. Results and discussion Chemical analyses were conducted for a total of 48 EOCs (Table S1). These EOCs were selected for their general occurrence in the

1483

environment (Hao et al., 2007a; Hao et al., 2007b) or due to their general usage in Ontario (Lissemore et al., 2006). The analyses were limited to only parent compounds and did not include metabolites or degradation products. Concentrations of EOCs that had a corresponding ILC standard were corrected according to the recovery of the ILC. For EOCs that had no ILC standard, concentrations were reported as is and, depending on the quality assurance data, might be less than accurate. Instrument detection limits (IDL) and detection limit (DL) listed in Table S1 were determined by the standard MOE practice for reporting drinking water results (Ministry of the Environment (MOE), 2008) and were used as descriptors of quality assurance and detections. Values of IDL was calculated from the standard deviation (SD) of quantitative results obtained from 8 consecutive analysis of level 3 calibration standard at concentrations ranging from 10 to 50 ppb, and calculated as 2.998 × SD (critical t0.010 = 2.998 @ degree of freedom (df) of seven). Values of DL were calculated in a similar manner using SD derived from quantitative results of eight method spike samples spiked at concentration level 3 of calibration standard. The approach used to determine IDLs and DLs also reflects the precision of the LC-MS/MS system and the method used in the monitoring project. 3.1. Quality assurance and quality control (QA/QC) data From Table S1 average method recoveries (Avg. %R) of the 15 EOCs measured with a corresponding ILC ranged from 100±5% (naproxen and sulfamethoxazole) to 113±13% (estrone) and 87±13% (progesterone). Target EOCs measured that had anomalous recoveries included monensin sodium (190%±49%), erythromycin (137±46%), and tylosin (177± 46%). This range of recovery is a well documented phenomenon associated with the analysis of these compounds (13, 14). For compounds without corresponding ILCs the accuracy of the measurement could not be corrected. Thus, measurements for two antibiotic compounds, penicillin G (83±70%) and virginiamycin (47±49%), were excluded from further discussions. Table S1 also lists the relative difference (%RRD) calculated from the %R of the two method spikes QC samples (%Rspike1 and %Rspike2) using Eq. (1).

%RRD =

2× j%Rspike1 −%Rspike2j %Rspike1 + %Rspike2

ð1Þ

The %RRD data was used to evaluate within-run method precision for each analyte and to demonstrate that the method is capable of measuring EOCs consistently. From Table S1 the average %RRD of the 15 EOCs with a corresponding ILC to carry out IDMS analysis ranged from 5% to 11% except for indomethacin (31%) and progesterone (19%). Due to matrix effects (Hao et al., 2007a), the three EOCs deemed to have unrealistic %R (i.e. enrofloxacin, norfloxacin and meclocycline with average %R at 407, 644 and 266% respectively) but with good method precision as showed by their respective average % RRD calculated at 11, 15 and 11% respectively; a good demonstration that analytical results obtained for these three compounds can be used for trend analysis. Values of the 8 EOCs designated by a # in Table 3 were measured by IDMS using their corresponding ILCs, are considered to have the best precision and accuracy possible (Hao et al., 2008), and are used as is. The %R QA data of lincomycin, benzafibrate and trimethoprim were summarized, respectively, at 113 ± 21%, 89 ± 21%, and 100 ± 25% and can be used to support the precision and accuracy of their monitoring data. The remaining six compounds (designated by a *) that did not have a corresponding ILC to carry out IDMS analysis, did not have good %R QA data (ranged from 137 ± 46% for erythromycin to 407 ± 45% for enrofloxacin), had a lower number of detections, and were used for qualitative statement only.

1484

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

Table 2 List of compounds analyzed in source (river and lake, N = 125 samples) and finished drinking water (N = 123 samples) samples. aChemical analysis with anomalous QA data (See Table S1). Compound

Group

Compounds detected in source water and finished drinking water Carbamazepine Pharmaceutical Gemfibrozil

Pharmaceutical

Bisphenol A

Plasticizer

Ibuprofen

Pharmaceutical

Lincomycin

Antibiotic

Sulfamethoxazole

Antibiotic

Acetaminophen

Pharmaceutical

Benzafibrate

Pharmaceutical

Trimethoprim

Antibiotic

Erythromycina

Antibiotic

Ketoprofen

Pharmaceutical

Tylosin Monensin Sodium

Antibiotic a

Antibiotic

Enrofloxacin

Antibiotic

Roxithromycin

Antibiotic

Tetracycline

Antibiotic

Norfloxacin

Antibiotic

Meclocyclin

Antibiotic

Compounds only detected in source water Naproxen Sulfamethazine Norethsterone Oxytetracycline Sulfathiazole

Pharmaceutical Antibiotic Antibiotic Antibiotic Antibiotic

Sample type

Number of sites (17)

N

%

Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished Source Finished

63 31 41 18 27 15 26 19 24 3 23 1 14 1 13 2 13 1 12 4 11 1 5 8 14 9 3 4 3 3 3 5 2 1 1 1

50 25 33 15 22 12 21 15 19 2 18 1 11 1 10 2 10 1 10 3 9 1 4 6 11 7 2 3 2 2 2 4 2 1 1 1

10 8 7 6 11 11 9 9 6 3 8 1 8 1 2 1 3 1 4 4 3 1 5 4 7 4 3 4 3 3 3 5 2 1 1 1

Source Source Source Source Source

26 12 1 1 1

21 10 1 1 1

5 4 1 1 1

2 1 1 1

2 1 1 1

2 1 1 1

Compounds only detected in finished drinking water Sulfachloropyridazine Antibiotic Clofibric acid Pharmaceutical Diclofenac Pharmaceutical Equilin Hormone

Finished Finished Finished Finished

Compounds not detected in the survey 17-α-estradiol 17- α a-ethynyl estradiol 17-β-estradiol Carbadox Chloramphenicol Chlorotetracycline Ciprofloxacin Diethylstilbesterol Doxycycline Estriol Estrone Indomethacin Lasaloid a Progesterone Sulfadiazine Sulfadimethoxine Sulfamerazine Sulfamethizole Warfarin

– – – – – – – – – – – – – – – – – – –

Hormone Hormone Hormone Antibiotics Pharmaceutical Antibiotics Antibiotics Hormone Antibiotics Hormone Hormone Antibiotics Antibiotics Hormone Antibiotics Antibiotics Antibiotics Antibiotics Pharmaceutical

Frequency of detection

– – – – – – – – – – – – – – – – – – –

– – – – – – – – – – – – – – – – – – –

– – – – – – – – – – – – – – – – – – –

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

3.2. Occurrence of emerging organic compounds in source and finished drinking water Of the 46 reportable EOCs, 23, 22 and 18 were detected (i.e. results above DL) in source, finished drinking water, or both, respectively on at least one occasion (Table 2). Table 2 also lists the 19 compounds that were not detected (ND). Of the 25 antibiotics analyzed, fourteen were detected in source water and 12 in finished drinking water. Only one hormone (equilin) was detected in one finished drinking water sample. Of the 11 pharmaceuticals analyzed, 7 were detected in source water and 8 in finished drinking water. Only five samples were taken from two DWSs using ground water. Ibuprofen was the only compound detected and in only one single sample. We observed that the highest percent detections (for EOCs analyzed) were from river (N90%) followed by lake sources, and therefore the results focus on DWSs that used either river or lakes as their source water (Tables 2 and 3). The antibiotics lincomycin, sulfamethoazole, trimethoprim, sulfamethazine, tylosin, and tetracycline were previously reported as being detected in two different surveys in the U.S. and Canada (Kolpin et al., 2002b; Lissemore et al., 2006); monensin sodium and erythromycin were reported in Canada (Lissemore et al., 2006). Bisphenol A was detected in both source and finished drinking water.

Table 3 Most frequently detected compounds (≥ 10%) in source (rivers and lake, N = 125) and finished drinking water (N = 123). References to the reported maximum were given following each value. Compound

DL, ng/L

Median 95th percentile

Maximum Reported maximum

ng/L (parts-per-trillion) Source water Carbamazepine#

1

3

152

749

Gemfibrozil# Bisphenol A# Naproxen# Ibuprofen#

1 2 2 0.5

0.7 2.1 1 0.98

6 44 58 24

9 87 199 79

Lincomycin 0.5 Sulfamethoxazole# 2 a Monensin Na 10 # Acetaminophen 2

0.12 0.17 0.59 0.1

15 28 67 95

143 284 810 298

Benzafibrate Trimethoprim

0.2 0.4

2 11

0.5 1

3.6 25

Erythromycina Sulfamethazine# Tylosina Tetracyclinea Enrofloxacina Roxithromycina

10 1 5 5 5 5

0.40 0.06 0.10 0.05 NC NC

19 4.5 6 4 NC NC

145 34 39 35 13 66

Drinking water Carbamazepine# Ibuprofen# Gemfibrozil# Bisphenol A# Monensin Naa Tylosina Tetracycline a Erythromycina Enrofloxacina Lincomycin Roxithromycina Bezafibrate Sulfamethoxazole# Acetaminophen# Trimethoprim

1 0.5 1 2 10 10 10 10 5 0.5 5 0.5 2 2 1

0.21 0.33 0.5 0.14 0.84 2.14 5.93 0.03 0.30 0.00 0.12 NC NC NC NC

37 12 2 17 22 13 11 12 3 0 6 NC NC NC NC

601 25 4 99 76 31 15 155 13 1413 41 1 2 17 15

650(19), 7100(20), 190(21) 710(15), 66(19) 12,000(15) 551(19), 150(25) 2700(14), 790(19), 270(21) 730(15), 355(16) 510(15) 220(16) 10,000(15), 160(21) 3100(18), 200(19) 710(15), 134(19), 160(21) 6.9(16) 220(15), 38(16) 280(15) 110(15) – –

24(22), 258(23) 930(24), 112(25) 70(22) 420(24) – – – – – – – – – – –

DL: Detection Limits; NC: Not calculated as the total detected number was b 2; a: semiquantitative data, for reference only; and #: IDMS data.

1485

Table 2 shows that the 13 most frequently detected compounds (in ≥10% of the samples analyzed) in the source waters were: carbamazepine (50%), gemfibrozil (33%), BPA (22%), ibuprofen (21%), naproxen (21%), lincomycin (19%), sulfamethoxazole (18%), acetaminophen (11%), monensin (11%), and benzafibrate, trimethoprim, erythromycin and sulfamethazine at 10%. The %R QA data of monensin sodium and erythromycin indicated that their frequency of detections might be biased high and should be used for reference only (as designated by an * in Table 3). Some compounds were detected frequently (in ≥ 10% of the samples) in source water but infrequently in drinking water. For example, naproxen and sulfamethazine were detected in 21% and 10% in source waters samples respectively, but were undetected in drinking water. Five compounds – lincomycin, sulfamethoxazole, acetaminophen, benzafibrate and trimethoprim – also showed lower frequency of detection (2% or less) in the finished drinking water samples following treatment. The most frequently detected four compounds in finished drinking water were: carbamazepine (CBZ, 25%), ibuprofen (IBU, 15%), gemfibrozil (GFB, 15%) and bisphenol A (BPA, 12%). Four compounds – sulfachloropyridazine, clofibric acid, diclofenac and equilin – were detected only once in finished drinking water, but were not detected in the source water. This may be an indication that the sample collection did not account for retention time in the DWS or the cleavage of conjugates of these compounds during the treatment (Ternes et al., 2004) and needs be confirmed in future studies. Table 2 shows the most frequently detected (≥10%) compounds in source water and their respective detections in finished drinking waters. 3.3. Concentrations of EOCs in source and finished drinking water 3.3.1. Source water Table 3 shows the concentrations (median, 95th percentile and maximum values) of the 17 EOCs that were detected in at least 2% of the source water samples. In comparing the point estimate summary statistics in this survey (Table 3) with those published from Spain, Germany, Norway, United States and Canada, the concentrations are generally lower (Hernando et al., 2004; Kolpin et al., 2002b; Lissemore et al., 2006; Ternes, 1998b; Metcalfe et al., 2003; Weigela et al., 2004; Focazio et al., 2008). Carbamazepine was the most frequently detected compound in this MOE survey. It was detected in 50% of the samples collected from 10 different sites, with maximum and median concentrations of 749 and 3 ng/L for all source water samples (N = 125). The mean and median values in detected samples was 45 and 6 ng/L respectively (N = 63). These concentrations are comparable to the 650 ng/L and 190 ng/L previously reported in Canada (Metcalfe et al., 2003) and the U.S. (Focazio et al., 2008), but an order of magnitude lower than the maximum reported concentration of 7100 ng/L in Germany (Weigela et al., 2004). Gemfibrozil, naproxen, IBU and BPA were all detected in N20% of samples collected in this survey. Gemfibrozil was detected in 33% of the samples in this MOE survey with corresponding maximum and median concentrations on 9 and 0.7 ng/L (N= 125). The percent detection and detected mean concentration (3 ng/L, N = 41) are consistent with those reported in earlier studies in Ontario (Hernando et al., 2004). However, the maximum concentration ranged from about 3 to 12 times lower in all but one of the previous studies (Weigela et al., 2004). The maximum concentration reported for naproxen in this MOE survey is similar to earlier studies (Hernando et al., 2004; Weigela et al., 2004), as are the mean and median concentrations for detectable samples only (40 ng/L and 20 ng/L, N = 26). Ibuprofen had a much lower maximum concentration reported in this MOE study than in earlier studies (Hao et al., 2007a; Hernando et al., 2004; Lissemore et al., 2006). Bisphenol A was detected at a higher frequency in this MOE survey than others (Lissemore et al., 2006). However, the maximum and

1486

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

median concentrations (87 and 2.1 ng/L, respectively) are lower than earlier reports (Hao et al., 2007b). These results indicate that source water in Ontario may have lower EOC loadings than those previously reported. The median (in detected samples only) and maximum concentrations reported in this MOE survey for naproxen (21 ng/L, 199 ng/L), acetaminophen (60.1 ng/L, 298 ng/L) and trimethoprim (9.6 ng/L, 25 ng/L) are consistent with those concentrations previously reported in Ontario and the United States (Hernando et al., 2004; Lissemore et al., 2006; Weigela et al., 2004). Whereas, the maximum reported concentrations for lincomycin (143 ng/L), sulfamethoxazole (284 ng/L), benzafibrate (3.6 ng/L) and sulfamethazine (34 ng/L) in this MOE study were much lower than previous reported values (Hao et al., 2007b; Pfeifer et al., 2002; Hernando et al., 2004). Also reported in Table 3 are the first reported detections and concentrations of two antibiotics roxithromycin and enrofloxacin in environmental samples. Due to the lack of corresponding ILCs to carry our IDMS analysis and inferior QA data, values of these antibiotics were semi-quantitatively determined to be at a concentration maximum of 13 and 66 ng/L, and should only be used as an occurrence reference. 3.3.2. Drinking water Table 3 shows the concentrations (median, 95th percentile and maximum values) of the 11 EOCs that were detected in at least 2% of the drinking water samples. Of the 48 EOCs analyzed, only 4 (i.e. CBZ, IBU, GFB and BPA) were detected in ≥10% of the drinking water samples. These results matched well in comparison with reported occurrence data in drinking water (Tauber, 2003; Stackelberg et al., 2004; Loraine and Pettigrove, 2006; Servos et al., 2007) from other jurisdictions. The remaining 7 EOCs – monensin, tylosin, tetracycline, erythromycin, enrofloxacin, lincomycin, roxithromycin and benzafibrate – – were detected in 2 to 9% of the drinking water samples. Carbamazepine was the most frequently detected compound in drinking water. It was detected in 25% of the samples from eight different sites, had a median, 95th percentile, and maximum value of 0.21, 37, and 601 ng/L, respectively (N = 123). The maximum concentration of carbamazepine reported in this MOE survey was higher than those previously reported in Canada (Tauber, 2003) and the U.S. (Stackelberg et al., 2004). Ibuprofen, GFB and BPA were detected in 15, 15 and 12% of the samples analyzed, respectively. Levels of ibuprofen, gemfibrozil and BPA measured were also lower than those previously reported (Tauber, 2003; Loraine and Pettigrove, 2006; Servos et al., 2007). 3.4. Effect of treatment technologies and concentrations of EOCs Field samples were collected from various DWSs. The selected DWSs used different treatment technologies including disinfection with chlorination, chemically assisted filtration using GAC or GAC filtration followed by UV treatment (Table 1). Due to the limited number of samples with measurable concentrations from all 17 DWSs, it was only possible to quantify the removal efficiency of the four most frequently detected compounds (i.e., CBZ, GFB, IBU and BPA) in drinking water from five individual DWSs. These five DWSs used either GAC or GAC and UV to treat source waters from river to obtain finished water. The median RE of carbamazepine from untreated source to finished drinking water was statistically significant for all plants (75 to 93% respectively) (Table S2). These results are similar to those using ozone and oxidation agent (Ternes et al., 2003) for removal of EOCs. For GFB, the two DWSs that used GAC and UV treatment achieved a median, statistically significant RE of 38 and 82%; the three DWSs using GAC alone achieved a median RE of 0, 44 and 55% from the source to finished waters, with the RE value of 44% being statistically significant. For IBU we observed RE values from 21 to 15% from source to finished waters; however, none of these RE could be calculated statistically to be significant. For BPA, we observed

a statistically significant RE of 99% BPA from source to finished water from one of the two DWSs using GAC and UV for disinfection; and two statistically insignificant RE of 98% and 80% from two DWSs that used, respectively, GAC only and GAC and UV treatment processes. By pooling data from all DWSs, regardless of treatment type, the general ability of DWSs in Ontario to reduce the levels of the four most frequently detected compounds from source to finished drinking water, was estimated (Fig. 1). Based on a comparison of log normal distributions of the concentrations of these four EOCs in source water and in drinking water (Figs. 2 and 3), it was clear that measured concentrations in drinking water samples were observed to be less than those observed in source water for these four EOCs, and at most sites. This is an indication that current drinking water treatment technologies used in Ontario can reduce these parent compounds consistently. For the remaining EOCs that were detected in source water but not in ≥ 10% of the drinking water samples (naproxen, lincomycin, sulfamethoxazole, acetaminophen, benzafibrate and trimethoprim), the reduced frequency of detection and concentrations in drinking water, may also be indicative of reduction through drinking water treatment (i.e., ≥ 10% detection in source water versus ≤ 2% in drinking water). There was also an indication that treatment process may result in the generation of target compounds because of the deconjugational process (Richardson and Bowron, 1985). This is illustrated in Table 2 in which four compounds, i.e. clofibric acid, diclofenac, equilin and sulfachloropyridazine were detected only in drinking water but not in source water. Further, there was a greater detection (40 to 80%) for the antibiotics enrofloxacin, norfloxacin, roxithromycin, tetracycline and tylosin in the finished drinking water with results that were relatively greater than those obtained from the source water. This phenomenon may be attributed more to the fact that the sampling did not take into account the hydraulic retention times at the individual treatment plants and less to the potential for inaccuracies in the average percent recoveries in the analytical method. However, none of these results was statistically significant and in view of the inferior QC/QA data, these results should be treated as a qualitative instead of quantitative statement. The fact that a compound can be detected in drinking water does not mean there is a risk to human health. In comparing the maximum concentration for carbamazepine that was detected in this MOE survey of 601 ng/L to a combined ‘predicted no effect concentration’ (PNEC) for drinking water and fish consumption of 0.226 × 106 ng/L (Focazio et al., 2008), carbamazepine would still have a calculated margin of safety (MOS) of 340 (Cunninghama et al., 2010). This PNEC is about 370 to 6000 times higher than the maximum and median concentrations that are reported in this MOE survey. Similarly, the

Fig. 1. Most frequently detected (≥10%) compounds in untreated source and finished drinking waters in Ontario.

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

Fig. 2. Log normal distribution of the concentration of carbamazepine (top) in source (detects = 64, non-detects = 61) and finished drinking waters (detects = 31, nondetects=92); and gemfibrozil (bottom) in sources waters (detects=41, non-detects=84) and treated waters (detects=18, non-detects=105) from different DWSs. Solid red circle: source water; Solid blue circle: finished drinking water.

PNECs for ibuprofen and gemfibrozil are 1.6 × 106 and 0.8 × 106 ng/L (Schwab et al., 2005) — both values are approximately 5 orders of magnitude higher than the maximum concentration reported in this survey. Further, a recent publication has estimated the total allowable concentration of BPA for an adult human weighing 70 kg, who ingests 2 L of drinking water per day, to be 100 μg/L (Whillhite et al., 2008). This value is 1000 times higher than the maximum concentration of 99 ng/L detected or 106 times the median value of 0.14 ng/L detected in this MOE survey. Acknowledgement The authors acknowledge the technical assistance of Nazma Khan, Imran Iftakhar and Yixun Shen, critical review of the manuscript by Renee Luniewski and Ray Clement and the support of the participating municipalities of the drinking water systems for their cooperation in this survey. Appendix A. Supplementary Data Supplementary data to this article can be found online at doi:10.1016/ j.scitotenv.2011.01.010. References Anderson PD, D' Aco VJ, Shanahan P, Chapra SC, Buzby ME, Cunningham VL, et al. Screening analysis of human pharmaceutical compounds in U.S. surface waters. Environ Sci Technol 2004;38:838–49. Cunninghama VL, Perinob C, D'Acoc VJ, Hartmannd A, Bechterd R. Human health risk assessment of carbamazepine in surface waters of North America and Europe. Regul Toxicol Pharmacol 2010;56(3):343–51. Daughton CG, Ternes T. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ Health Perspect 1999;107:907–41.

1487

Fig. 3. Log normal distribution of the concentration of bisphenol A (top) in source (detects = 27, non-detects = 98) and finished drinking waters (detects = 15, nondetects=110); and ibuprofen in source (detects=25, non-detects=100) and finished drinking waters (detects=18, non-detects=105) from different DWSs. Solid red circle: source water; Solid blue circle: finished drinking water.

Focazio MJ, Kolpin DW, Barnes KK, Furlong ET, Meyer MT, Zuagg SD, et al. A national reconnaissance for pharmaceuticals and other organic wastewater contaminants in the United States — II) Source drinking water sources. Sci Total Environ 2008;402 (2–3):201–16. Hao C, Zhao X, Yang P. GC-MS and HPLC-MS analysis of bioactive pharmaceuticals and personal-care products in environmental matrices. Trends Anal Chem 2007a;26 (6):569–80. Hao C, Clement R, Yang P. Liquid chromatography–tandem mass spectrometry of bioactive pharmaceutical compounds in the aquatic environment—a decade's activities. Anal Bioanal Chem 2007b;387:1247–57. Hao C, Zhao X, Tabe S, Yang P. Optimization of a multi-residual method for the determination of waterborne emerging organic pollutants using solid phase extraction and liquid chromatography/tandem mass spectrometry and isotope dilution mass spectrometry. Environ Sci Technol 2008;42(11):4068–75. Helsel DR. Fabricating data: how substituting values for nondetects can ruin results, and what can be done about it. Chemosphere 2006;65:2434–9. Hernando MD, Petrović M, Fernández-Alba AR, Barceló D. Analysis by liquid chromatography–electrospray ionization tandem mass spectrometry and acute toxicity evaluation for β-blockers and lipid-regulating agents in wastewater samples. J Chromatogr 2004;1046:133–40. Kolpin DW, Furlong ET, Meyer MT, Thurman EM, Zaugg SD, Barber LB, et al. Pharmaceuticals, hormones, and other organic wastewater contaminants in U.S. streams, 1999–2000: a national reconnaissance. Environ Sci Technol 2002a;36: 1202–11. Kolpin DW, Furlong ET, Meyer MT, Thruman EM, Zaugg SD, Bareber LB, et al. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999–2000: a national reconnaissance. Environ Sci Technol 2002b;1067: 153–60. Lissemore L, Hao C, Yang P, Sibley PK, Mabury S, Solomon KH. An exposure assessment for selected pharmaceuticals within a watershed in Southern Ontario. Chemosphere 2006;64:717–29. Loraine GA, Pettigrove ME. Seasonal variations in concentrations of pharmaceuticals and personal care products in drinking water and reclaimed wastewater in Southern California. Environ Sci Technol 2006;40(3):687–95. Metcalfe CD, Miao XS, Koenig BG, Struger J. Distribution of acidic and neutral drugs in surface waters near sewage treatment plants in the lower Great Lakes, Canada. Environ Toxicol Chem 2003;22:2881–9.

1488

S. Kleywegt et al. / Science of the Total Environment 409 (2011) 1481–1488

Ministry of the Environment. The determination of emerging organic pollutants in environmental matrices by LC-MS-MS analysis. Laboratory Services Branch method E3454; 2010. Ministry of the Environment (MOE), Laboratory Services Branch Procedures for Processing and Reporting Drinking Water Samples, version 2.0 (Operating Procedure #39 (LSBSOP.039). (as updated, 2008). Pfeifer T, Tuerk J, Bester K, Spiteller M. Determination of selected sulfonamide antibiotics and trimethoprim in manure by atmospheric pressure chemical ionization tandem mass spectrometry. Rapid Commun Mass Spectrom 2002;16:663–9. Richardson ML, Bowron JM. The fate of pharmaceutical chemicals in the aquatic environment. J Pharm Pharmacol 1985;37:1-12. Richardson SD, Ternes TA. Water analysis: emerging contaminants and current issues. Anal Chem 2005;77:3807–38. Rowney NC, Johnson AC, Williams RJ. Cytotoxic drugs in drinking water: a prediction and risk assessment exercise for the Thames Catchment in the United Kingdom. Environ Toxicol Chem 2009;28(12):2722–43. Schwab BR, Hayes EP, Fiori JM, Mastrocco FJ, Roden NM, Cragin D, et al. Human pharmaceuticals in US waters: a human health risk assessment. Regul Toxicol Pharmacol 2005;42:296–312. Servos MR, Smith M, McInnis R, Burnison BK, Lee BH, Seto P, et al. The Presence of selected pharmaceuticals and the antimicrobial triclosan in drinking water in Ontario, Canada. Water Qual Res J Can 2007;42(2):130–7. Stackelberg PE, Furlong ET, Meyer MT, Zauggb SD, Hendersond AK, Reissmand DB. Persistence of pharmaceutical compounds and other organic wastewater contaminants in a

conventional drinking-water-treatment plant. Sci The Total Environ 2004;329(1–3): 99-113. Tauber R. Quantitative analysis of pharmaceuticals in drinking water from ten Canadian cities — report prepared for Mark Stevenson (CTV News, Canada). Enviro-Test Laboratories/Xenos Division, Ontario, Canada; 2003 http://www.ctv.ca/servlet/ ArticleNews/story/CTVNews/1044053088271_39462288/. 20 January. Ternes TA. Occurrence of drugs in German sewage treatment plants and rivers. Water Res 1998a;32:3245–60. Ternes TA. Occurrence of drugs in German sewage treatment plants and rivers. Water Res 1998b;32:3245–60. Ternes TA, Stuber J, Herrmann N, McDowell D, Ried A, Kampmann M, et al. Ozonation: a tool for removal of pharmaceuticals, contrast media and musk fragrances from wastewater? Water Res 2003;37:1976–82. Ternes TA, Joss A, Siegrist H. Scrutinizing pharmaceuticals and personal care products in wastewater treatment. Environ Sci Technol 2004;38(20):392A–9A. Weigela S, Bergerb U, Jensenc E, Kallenbornb R, Thoresenc H, Hühnerfuss H. Determination of selected pharmaceuticals and caffeine in sewage and seawater from Tromsø/ Norway with emphasis on ibuprofen and its metabolites. Chemosphere 2004;56(6): 583–92. Whillhite CC, Ball GL, McLellan CJ. Derivation of a oral reference dose - bisphenol a and drinking-water equivalent concentration. J Toxicol Environ Health B 2008;11(69): 69-146.