Angela
Yu-Chen Lin
*,
Sri Chandana
Panchangam
and
Huan-Yo
Chen
Graduate Institute of Environmental Engineering, National Taiwan University, 71-Chou-Shan Road, Taipei, 106, Taiwan. E-mail: yuchenlin@ntu.edu.tw; Fax: +886-2-3366-9828; Tel: +886-2-33663729
First published on 15th June 2009
Contamination of natural aquatic environments with human-use pharmaceuticals poses a significant potential threat to local ecosystems. However studies of the risk assessment of this contamination are limited, in part because currently the environmental concentration prediction is logistically and technically difficult to perform. In this study, 1) a strategic method to determine occurrence and risk of pharmaceutical compounds in an aquatic environment is proposed, and 2) this method is applied to study ten human-use pharmaceuticals in Taiwan's Sindian river, which traverses and supplies drinking water for Taipei, a major metropolitan center. In river-water samples collected over a three-week period, concentrations of NSAIDs (acetaminophen, diclofenac, ibuprofen, naproxen, and ketoprofen), steroids (estrone, 17α-ethinylestradiol and 17β-estradiol), the anti-hypertensive agent propranolol, and the lipid regulator gemfibrozil were found at ng/L to µg/L levels; night-time concentrations were often doubled or even greater, compared to day-time levels. With the exception of the estrogens, predicted environmental concentrations (PECs) of the target pharmaceuticals agreed well with measured environmental concentrations (MECs). Conclusions: In the Sindian river, which is impacted by both regional and hospital discharges, the pattern of pharmaceutical occurrences and concentrations observed here is due to variations in source release rather than to natural attenuation phenomena. In addition, the environmental risk posed by acetaminophen, ibuprofen, and the estrogens was observed to be imminent, while risk from the other drugs targeted was found to be lower, even when maximum MECs were considered.
The sources and pathways of pharmaceuticals into the environment are well understood.7,15 Pharmaceuticals consumed by humans may undergo biochemical modifications during hepatic or enteric metabolism before excretion in urine or feces16 and eventual arrival into the aquatic environment. Direct release via household toilet flushing is another non-negligible route of drug disposal. Bound et al.13 concluded that the chosen disposal method was unaffected by the human perception on risk associated with pharmaceuticals. Of non-household pharmaceutical contamination sources such as hospitals, sewage treatment plants, wastewater treatment plants (WWTPs), production facilities, animal husbandries and aquacultures – WWTPs, hospitals, and livestock operations have been shown to be the dominate sources.17,18 In Taipei, Taiwan, we previously conducted surveys of pharmaceutical removal efficiencies in WWTPs which revealed that NSAIDs were the main constituents (61–69%) entering WWTP influents, while certain antibiotics restrained the secondary treatment processes.19 In addition, a sweep of 97 pharmaceuticals in 23 Taiwanese potential pharmaceutical contamination source water sites were studied previously.10 The results showed the presence of several pharmaceuticals at µg L−1 level exceeding the predicted no effect concentration (PNEC) values warranting further investigations.
Although regulating the entire assembly of pharmaceuticals is impractical, a clear need exists for targeted risk assessment studies of compounds with the potential for environmental harm. Risk assessment guidelines proposed by the European Medicines Agency (EMEA: formerly the Agency for the Evaluation of Medicinal Products) calculate the predicted environmental concentration (PEC) and signal potential risk when the PEC-to-PNEC ratio (risk quotient) exceeds one.20 With these guidelines in mind, several recent attempts have been made to estimate PEC values for the common pharmaceuticals with high prescription.14,18,21 However, inconsistent findings for rates of pharmaceutical metabolism severely limit the reliability of currently reported PECs.22,23 In addition, only limited data are available for pharmaceuticals' eventual fate; their dilution in the receiving water bodies depend on various factors such as regional climatic conditions, catchment area and flow of the streams.21–24 This may cause underestimation of the predicted concentrations comparing to measured environmental concentrations (MECs), eventually resulting in false-negative error as in the case of carbamazepine and propranolol in the study by Ferrari et al.24 Even though the pharmaceuticals are present in low concentrations in surface waters, the level of toxicity they exhibit is significant as witnessed in the case of Pakistan where the decline in vulture population was attributed to diclofenac.25 The pharmaceuticals are expected to pose more risk due to their mixture/combined toxicity.
Although guidelines for risk assessment of newly registered pharmaceuticals were introduced in the US/EU, they are unlikely to affect the environmental occurrence of already licensed pharmaceuticals that are not regulated by any guidelines.26,27 In addition to the constraints discussed above, the analytical techniques which were available to detect pharmaceutical compounds in various complex environmental matrices were also limited. However, major advances in analytical chemistry have introduced reliable and accurate techniques such as liquid chromatography tandem mass spectrometry (LC-MS/MS) and solid phase extraction (SPE), which have minimized solving previous analytical limitations, lowered detection limits, and increased sensitivity to target compounds.6
This study investigated the occurrence of ten pharmaceuticals (five NSAIDs, three estrogens, an anti-hypertensive and a lipid-regulator) in Taiwan's Sindian river using LC-MS/MS. Risk assessment for the pharmaceuticals detected was performed and PECs were evaluated using a method based on removal rates by various routes. PECs and MECs were compared to validate the method itself and its usefulness for risk assessment in an aquatic environment.
Compound | CAS No. | Log Kowa | MW | Structure |
---|---|---|---|---|
a The log Kow values are from the references 28, 33, 46, and 47. | ||||
Acetaminophen | 103-90-2 | 0.46 | 151.2 |
![]() |
Ibuprofen | 15687-27-1 | 3.97 | 206.3 |
![]() |
Naproxen | 22204-53-1 | 3.18–3.24 | 230.3 |
![]() |
Ketoprofen | 22071-15-4 | 3.12–3.16 | 254.3 |
![]() |
Diclofenac | 15307-79-6 | 4.51 | 318.13 |
![]() |
Estrone | 53-16-7 | 3.13 | 270.37 |
![]() |
17α-Ethynylestradiol | 57-63-6 | 3.67 | 296.4 |
![]() |
17β-Estradiol | 50-28-2 | 4.01 | 272.39 |
![]() |
Propranolol | 525-66-6 | 1.20–3.48 | 295.8 |
![]() |
Gemfibrozil | 25812-30-0 | 4.77 | 250.3 |
![]() |
![]() | ||
Fig. 1 Map of sampling site. |
A high-performance liquid chromatography tandem mass spectrometry (HPLC-MS/MS) system (Applied Biosystems API 4000) was used to analyze the samples. The HPLC module included a degasser (Agilent 1100 Series Micro Vacuum Degasser), pump (Agilent 1100 Series Binary Pump) and autosampler equipment (CTC Analytics HTC PAL System). Chromatographic separations were performed with ZORBAC Eclipse XDB-C18 columns (Agilent, Palo Alto, CA, USA, 150 × 4.6 mm, 5 µm). The autosampler was operated at room temperature. The column was equilibrated for 5 min before injection of 20 µL samples. The analyzing time was 10 min for each sample. A binary gradient program was used with 0.1% formic acid in DI water as mobile phase A (MPA) and 0.1% formic acid in 100% methanol as mobile phase B (MPB) in both positive and negative ion modes. The HPLC solvent gradient was operated at 1000 µL min−1 flow rate, starting at 90%:10% MPA:MPB for 0.2 min, gradually approaching 30%:70% MPA:MPB, and finally reaching 10%:90% MPA:MPB at 4.5 min. The gradient reached 5% MPA:95% MPB at 5.0 min, continued until 8.0 min, and by 8.5 min reached 90% MPA:10% MPB which was held constant until the end of the analysis for up to 10 min.
A quadrupole mass spectrometer equipped with electrospray ionization (ESI) interface was used to make mass spectrometric measurements. The analyses were done in positive ion mode for acetaminophen, propranolol, estrone, 17β-estradiol, and 17α-ethinylestradiol, and in negative ion mode for ibuprofen, gemfibrozil, ketoprofen, naproxen, and diclofenac. Ions were acquired in multiple reaction monitoring (MRM) mode with a dwell time of 200 ms. The mass spectrometric conditions applied were: ion spray voltage of 5.5 and −4.5 kV, curtain gas at 10 L h−1, nebulizer gas at 60 L h−1 and turbo gas at 50 L h−1, heated capillary temperature of 550 °C and 450 °C (negative ion mode) and collisionally activated dissociation 5 with interface heater on. The precursor and MRM product ion pairs of the target compounds are shown in Table 2.
Compound | MRM pairs | Retention time | MDLs | Linearity, r | Accuracy (%) | Precision (%) | ||||
---|---|---|---|---|---|---|---|---|---|---|
m/z | (min) | (ng/L) | Day 1 | Day 2 | Day 3 | Day 1 | Day 2 | Day 3 | ||
Acetaminophen | 152 > 93 | 4.13 | 0.2 | 0.9998 | 17.1 | 4.4 | 8.4 | 2.7 | 1.0 | 3.0 |
152 > 110 | ||||||||||
Ibuprofen | 205 > 161 | 7.32 | 10 | 0.9993 | −3.3 | 9.0 | 9.8 | 1.8 | 6.1 | 3.8 |
Naproxen | 229 > 169.8 | 6.70 | 2 | 0.9922 | −13.3 | −4.6 | 7.4 | 4.2 | 3.7 | 2.5 |
229 > 184.9 | ||||||||||
Ketoprofen | 253 > 208.9 | 6.49 | 5 | 0.9997 | −4.9 | 6.6 | 5.0 | 3.7 | 4.8 | 6.7 |
253 > 197 | ||||||||||
Diclofenac | 294 > 250 | 7.21 | 5 | 0.9991 | 7.4 | 10.6 | 8.3 | 6.2 | 5.9 | 6.9 |
294 > 214 | ||||||||||
Estrone | 271 > 133 | 6.81 | 10 | 0.9985 | 8.3 | 10.6 | 3.2 | 11.0 | 5.2 | 3.7 |
271 > 253 | ||||||||||
17α-Ethynylestradiol | 279 > 133 | 6.67 | 10 | 0.9998 | −5.6 | 12.6 | 12.6 | 6.3 | 4.0 | 1.9 |
279 > 159 | ||||||||||
17β-Estradiol | 255 > 133 | 6.78 | 10 | 0.9999 | 11.9 | 7.2 | 5.4 | 13.9 | 8.7 | 4.2 |
255 > 159 | ||||||||||
Propranolol | 260 > 116 | 5.31 | 0.2 | 0.9999 | 18.4 | 16.2 | 14.8 | 2.6 | 3.9 | 6.0 |
260 > 183 | ||||||||||
Gemfibrozil | 249 > 120.9 | 7.74 | 0.2 | 0.9987 | −4.7 | 0.6 | 11.3 | 5.1 | 7.1 | 9.3 |
249 > 126.9 |
![]() | (1) |
In this study, the amount of pharmaceutical used (“A”, Table 3) was calculated using data from Taiwan's Bureau of National Health Insurance (NHI). The NHI is an exceptional database that tracks drug use patterns, with which about 91% of health care providers such as hospitals and clinics had contracted. Further, more than 97% of the Taiwanese population was enrolled by the end of 2005.29 For prescription medications, the most recent data available are from the year 2005 (Table 3), with the exception of ketoprofen, for which data from the year 2002 were used with a factor of 4 for the PEC estimation. The ‘A’ value in kg/yr was calculated as a product of prescribed dose to averaged drug content using the NHI medicinal usage data.
Pharmaceuticals | Prescription in Taiwan, NHI 2005 (× 103) | Amount of pharmaceutical used per annum, A (kg/yr) | Removal by transformation into metabolites, RTM (%) | Removal by natural attenuation, RNA (%), (half-life values from literature) | Net removal, Rnet (%) |
---|---|---|---|---|---|
a na: not available. | |||||
Acetaminophen | 568839 | 142278.0 | 9640 | 0, (>24 h)41 | 76.8 |
Ibuprofen | 131359 | 52543.8 | 9642 | 0, (14.8 ± 0.7 h)33 | 76.8 |
Naproxen | 22655 | 3685.9 | 3522 | 60, (1.4 ± 0.10 h)33 | 71.2 |
Ketoprofen | 7914 | 7914.0 | 2022 | 99, (4.1 ± 0.13 min)33 | 99.16 |
Diclofenac | 214672 | 6400.9 | 8543 | 60, (4 h)48 | 87.2 |
Estrone | 15 | 0.0189 | na | 60, (2.3 ± 0.07 h)33 | 60.0 |
17α-Ethinylestradiol | 801 | 0.0281 | 7440 | 60, (2.3 ± 0.11 h)33 | 83.68 |
17β-Estradiol | 19106 | 8.9821 | 9540 | 60, (2.0 ± 0.14 h)33 | 90.4 |
Propranolol | 156637 | 3916.0 | 9943 | 60, (1.1 ± 0.04 h)33 | 91.68 |
Gemfibrozil | 25015 | 11256.9 | 5043 | 0, (14.8 ± 0.70 h)33 | 40.0 |
Removal rate is a critical parameter that depends on factors including conversion of pharmaceuticals in human body, illicit disposal of drugs, and the fate of pharmaceuticals in the aquatic environment after being subjected to natural attenuation processes. Here, the method depicted in Fig. 2 was implemented to find an appropriate Rnet value. In general, human-use pharmaceuticals enter the aquatic environment via one of two major routes. A drug may be consumed and metabolized by a human, who eventually excretes a portion of the remaining parent compound along with its metabolites. Although some metabolites are shown to be as toxic as or even more toxic than the parent compouds,7 due to the fact that limited and sparse metabolite data is available, only parent compounds are considered in this study. Alternatively, unused drugs may be thrown away, having never been consumed or metabolized.
![]() | ||
Fig. 2 Methodological approach to calculate the net removal rates. |
In Taiwan, medications are available predominantly via prescription. In contrast, the quantity of over-the-counter (OTC) medications consumed is negligible relative to prescription medications and for the purposes of this study may be ignored. Medications are relatively inexpensive, making it easy for Taiwanese patients to obtain too many prescription drugs, which are then stored at home. Upon reaching their expiration dates, these surplus drugs are thrown away into household dustbins or flushed down toilets. This path (route 2, Fig. 2), which depends on local cultural and social patterns of medication use, may involve a significant medication volume; in fact it may be reasonable to conclude that a majority (nearly 80%) of human pharmaceutical release into aquatic streams followed route 1, while the remaining 20% appeared to follow route 2.30 Drugs which were released through route 1 are considered to be removed from the system via their transformation into metabolites (RTM). Pharmaceutical residues which remain in the form of the original parent compound would then be subjected to natural attenuation processes such as hydrolysis, photodegradation, biodegradation, and adsorption, and may undergo further degradation. In surface waters, pharmaceuticals and their residues are most likely to be removed by direct and indirect photolysis.31,32 Sorption and biotransformation are also significant natural attenuation mechanisms.7,22
Their removal rates (RNA) were assigned according to the half-lives reported in the literature. Since the rivers in Taiwan flow at comparatively faster rates and are exposed to little sunlight, a removal rate of 60% was assumed for a half-life <5 h, 30% for half-lives between 5 and 10 h, and no removal for a reported half-life >10 h. However, two exceptions must be taken into account when calculating Rnet: for ketoprofen, whose photolytic half-life is only a few minutes long, RNA is taken to be 99%,33 and for estrone, RTM is assigned as 100%, since metabolite data are unavailable. The net removal rate Rnet was then calculated using the following formula:
![]() | (2) |
Table 3 shows the RTM, RNA, and Rnet values along with the half-life values obtained from the literature. In order to calculate PEC values, the Rnet value obtained from equation 2 was substituted in to equation 1 (Table 4).
Pharmaceuticals | PEC | MEC (min–max) | MEC (Median) | PNEC | RQPEC (PEC/PNEC) | RQmin (Minimum MEC/PNEC) | RQmax (Maximum MEC/PNEC) | RQmed (Median MEC/PNEC) |
---|---|---|---|---|---|---|---|---|
a RQ was obtained by calculating PEC with half of the MDL value; nd. not detected; na. not available. | ||||||||
NSAIDs | ||||||||
Acetaminophen | 848.1 | 8.3–9170 | 1755.0 | 920021 | 0.0922 | 0.0009 | 0.9967 | 0.1908 |
Ibuprofen | 313.2 | nd–4350 | 231.5 | 500021 | 0.062 | <0.001a | 0.87 | 0.0462 |
Naproxen | 27.3 | 35.2–270 | 65.5 | 3700021 | 0.0007 | 0.0010 | 0.0073 | 0.0018 |
Ketoprofen | 1.7 | nd–45.0 | nd | 1560000014 | 0.00000011 | <0.00000016a | 0.0000028 | <0.00000016a |
Diclofenac | 21.1 | nd–56.5 | 12.9 | 1000021 | 0.0021 | <0.00025a | 0.0057 | 0.0013 |
Estrogens | ||||||||
Estrone | 0.0002 | nd–191 | 13.1 | 1844 | 0.000011 | <0.27a | 10.6 | 0.7278 |
17α-Ethynylestradiol | 0.0001 | nd–19.5 | nd | 0.0244 | 0.005 | <250a | 975 | <250a |
17β-Estradiol | 0.0222 | nd–64.7 | 15.6 | 0.0244 | 1.11 | <250a | 3235 | 780 |
β-Blocker | ||||||||
Propranolol | 8.4 | 5.8–39.1 | 13.1 | 50024 | 0.0168 | 0.5800 | 0.0782 | 0.0262 |
Lipid-regulator | ||||||||
Gemfibrozil | 173.5 | 66.9–279.0 | 115.2 | 10000045 | 0.001735 | 0.1520 | 0.0028 | 0.001152 |
![]() | ||
Fig. 3 Semi log plot of pharmaceutical contamination in the Sindian river at the six sampling sites. |
As shown in Fig. 3, acetaminophen, ibuprofen, gemfibrozil, and naproxen were most often detected at each sampling point and at the highest concentrations; in contrast, other pharmaceuticals were detected at levels well below 40 ng L−1. The high concentrations of these four compounds may reflect their usage patterns. Closer examination of pharmaceutical occurrences at different sampling points showed that the median concentrations of acetaminophen and naproxen decreased from sampling points A1 to A4 and increased up to sampling point A6, while ibuprofen concentrations displayed the reverse trend. Gemfibrozil concentration increased continuously from A1 to A6. Interestingly, acetaminophen alone displayed concentrations of thousands of ng L−1. Our findings on the pharmaceutical contamination in the Sindian river are likely to have been influenced by uncertainties in local customs of medication prescription, regional discharges, or dilution factors (the Sindian's streamway appears to widen beyond sampling location A3). Sites A5 and A6 may have been impacted more by nearby urban regional discharges (since hospitals were absent), while sampling locations A1 and A2 might have been influenced by both urban regional and hospital discharges. However, the distances between the sampling points are not so far away that any alarming differences in the contamination sources may be identified.
Pharmaceuticals | Day/night | First week | Second week | Third week | Median concentration |
---|---|---|---|---|---|
a n.d. = not detected. | |||||
Acetaminophen | Day | 1540.0 | 512.2 | 1730.0 | 1450.0 |
Night | 2370.0 | 4800.0 | 1930.0 | 2360.0 | |
Ibuprofen | Day | 92.7 | 303.0 | 365.5 | 119.5 |
Night | 125.0 | 237.0 | 1635.0 | 237.0 | |
Naproxen | Day | 50.4 | 63.2 | 65.7 | 57.8 |
Night | 49.4 | 79.9 | 156.0 | 72.65 | |
Ketoprofen | Day | n.d. | n.d. | n.d. | n.d. |
Night | n.d. | n.d. | n.d. | n.d. | |
Diclofenac | Day | 8.1 | 10.4 | 14.9 | 10.2 |
Night | 7.3 | 23.3 | 24.3 | 20.3 | |
Estrone | Day | 12.3 | n.d. | n.d. | 5.7 |
Night | n.d. | 56.5 | 29.6 | 21.3 | |
17α-Ethilnylestradiol | Day | n.d. | n.d. | n.d. | n.d. |
Night | n.d. | n.d. | n.d. | n.d. | |
17β-Estradiol | Day | 7.9 | 5.3 | 24.4 | 11.0 |
Night | 5.0 | 44.5 | 24.6 | 24.6 | |
Propranolol | Day | 13.9 | 8.4 | 10.6 | 11.6 |
Night | 13.6 | 31.8 | 12.2 | 14.6 | |
Gemfibrozil | Day | 109.8 | 95.7 | 91.5 | 97.2 |
Night | 107.8 | 126.5 | 194.5 | 126.5 |
Values for PECs and minimum, maximum and median MECs are shown in Table 4. Almost all of the PECs fell within the experimentally detected concentration range, with almost perfect prediction of the median MECs of diclofenac, ibuprofen, gemfibrozil and propranolol (MECs: 12.9, 231.5, 115.2, and 13.1 ng L−1 and PECs: 21.1, 313.2, 173.5, and 8.4 ng L−1). PECs of naproxen and acetaminophen, popular over-the-counter (OTC) analgesics, were about half of their median MECs but still fallen within the detection range. This discrepancy may be explained by the fact that we did not consider OTC purchase of these widely available drugs. Ketoprofen was almost undetectable in many samples, but the PEC was one order of magnitude below the maximum MEC value, which was considered an acceptable prediction. Estrogen levels were underestimated in most of the cases. This could be due to underestimation of the quantity of drugs used or overestimation of removal by transformation (RTM) and natural attenuation processes (RNA). In this study, pharmaceutical removal by wastewater treatment was not considered, despite its importance in PEC estimation,14,22 as there were no wastewater treatment plants situated along the stretch of river near our sampling areas. In an earlier study, Lin et al.10 did not detect any estrogens in hospital effluents in Taiwan; this finding in combination with the fact that they are prescribed in considerable quantity indicates that estrogens are effectively removed by currently used treatment processes. Furthermore, the estrogens (with the exception of 17α-ethinylestradiol) occur naturally and could enter the river through natural excretion and plant decomposition processes. In general, with the exception of estrogens, our method for predicting environmental concentrations of these pharmaceuticals in Taiwan's aquatic environment yielded results which closely matched measured values, strongly supporting the methodology adopted in this study.
When RQs were calculated with minimum MECs, none of the pharmaceuticals would be predicted to pose risk, as the RQmins were well below one. When RQmed and RQPEC were analyzed, 17β-estradiol posed potential risk, with a calculated RQmed and RQPEC of 780 and 1.11, respectively. The RQmax of estrone (10.6), 17α-ethinylestradiol (975), and 17β-estradiol (3235) were alarmingly high, signifying a major potential risk. The RQmax of acetaminophen (0.9967) and ibuprofen (0.87) were close to one, indicating a non-negligible risk. When RQs were calculated using PECs, the estrogens appeared to pose little risk, with RQPEC of 0.000011 (estrone) and 0.005 (17β-estradiol). In certain cases PECs may underestimate the underlying risk of the pharmaceuticals, which can cause a false-negative error (as for estrone and 17β-estradiol). A false-negative error is a situation in which a chemical that poses significant risk to the environment is falsely identified as non-hazardous. Overall, acetaminophen, ibuprofen, and the three estrogens emerge in Sindian river as risk posing pharmaceuticals while other drugs in this study are far safer even by their maximum MEC. It is important to note that no pharmaceutical contaminant is present in isolation in the environment; instead the true state of affairs should always be viewed as an assembly of pharmaceuticals with unpredictable toxicities and complex synergistic or antagonistic interactions.
In conclusion, we have assessed the risk from the selected human pharmaceuticals in the Sindian river. A novel method for evaluating Rnet which is based on removal rates of multiple degradative processes, and predicts experimentally measured concentrations with a high degree of accuracy was proposed in this study. This approach may be implemented in the future to monitor the occurrence and risk assessment of other human pharmaceuticals in the Sindian river. Since no WWTPs were located in the study area, inclusion of a factor accounting for removal of pharmaceuticals by hospital treatment processes into the PEC calculation (equation 2) would improve the accuracy and broader applicability of this approach.
Footnote |
† Part of a themed issue dealing with water and water related issues. |
This journal is © The Royal Society of Chemistry 2010 |