Doi:10.1016/j.chroma.2004.04.005

Journal of Chromatography A, 1041 (2004) 171–180 Determination of pharmaceutical compounds in surface- and ground-water samples by solid-phase extraction and high-performance liquid chromatography–electrospray ionization mass spectrometry Jeffery D. Cahill , Edward T. Furlong , Mark R. Burkhardt , Dana Kolpin , a National Water Quality Laboratory, US Geological Survey, P.O. Box 25046, MS 407, Denver Federal Center, Denver, CO 80225-0046, USA b Department of Chemistry, University of Colorado at Denver, P.O. Box 173364, Denver, CO 80217-3364, USA Received 18 July 2003; received in revised form 26 March 2004; accepted 8 April 2004 Abstract
Commonly used prescription and over-the-counter pharmaceuticals are possibly present in surface- and ground-water samples at am- bient concentrations less than 1 ␮g/L. In this report, the performance characteristics of a combined solid-phase extraction isolation andhigh-performance liquid chromatography–electrospray ionization mass spectrometry (HPLC–ESI-MS) analytical procedure for routine deter-mination of the presence and concentration of human-health pharmaceuticals are described. This method was developed and used in a recentnational reconnaissance of pharmaceuticals in USA surface waters. The selection of pharmaceuticals evaluated for this method was based on us-age estimates, resulting in a method that contains compounds from diverse chemical classes, which presents challenges and compromises whenapplied as a single routine analysis. The method performed well for the majority of the 22 pharmaceuticals evaluated, with recoveries greaterthan 60% for 12 pharmaceuticals. The recoveries of angiotensin-converting enzyme inhibitors, a histamine (H2) receptor antagonist, and antihy-poglycemic compound classes were less than 50%, but were retained in the method to provide information describing the potential presence ofthese compounds in environmental samples and to indicate evidence of possible matrix enhancing effects. Long-term recoveries, evaluated fromreagent-water fortifications processed over 2 years, were similar to initial method performance. Method detection limits averaged 0.022 ␮g/L,sufficient for expected ambient concentrations. Compound-dependent matrix effects on HPLC/ESI-MS analysis, including enhancement andsuppression of ionization, were observed as a 20–30% increase in measured concentrations for three compounds and greater than 50% increasefor two compounds. Changing internal standard and more frequent ESI source maintenance minimized matrix effects. Application of themethod in the national survey demonstrates that several pharmaceuticals are routinely detected at 0.010–0.100 ␮g/L concentrations.
2004 Elsevier B.V. All rights reserved.
Keywords: Water analysis; Environmental analysis; Drugs 1. Introduction
use of pharmaceuticals has been the inadvertent introductionof these compounds or their metabolites into surface wa- The discovery, production, and medicinal use of phar- ter and ground water, initially identified in European studies maceuticals, synthetically derived or extracted from natu- The presence of pharmaceuticals in aquatic environ- ral materials, are one of our society’s greatest medical as- ments typically results from human excretion of metabolized sets Human and animal health has been improved sub- and unmetabolized drug passing into sewage or septic sys- stantially by the introduction of antibiotics, analgesics, and tems and subsequent discharge of wastewater perco- condition-specific formulations for heart disease, hyperten- lation of septic-system leachate into surface water or ground sion, and other episodic and chronic health problems. An un- water. Animal waste in pasturage or confined animal-feeding surprising but unintended consequence of the near universal operations is another potential source of pharmaceuticals.
Direct discharge to ground from excretion by livestock can ∗ Corresponding author. Tel.: +1-303-2363941; fax: +1-303-2363499.
contaminate surface water, or collection of waste in lagoons E-mail address: [email protected] (E.T. Furlong).
0021-9673/$ – see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2004.04.005 J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 Recent data from Europe indicate that the normal op- ter contaminants in surface water and ground water in the eration of sewage-treatment plants results in the incom- USA. Some of the goals of the study are as follows: (1) to plete removal of pharmaceuticals, hence as much as 80% of develop a list of compounds that reflect the prescription and the total load of pharmaceuticals entering sewage-treatment over-the-counter pharmaceuticals most likely found in USA plants may be discharged into surface water phar- surface water and ground water; (2) to develop a method maceutical concentrations measured in wastewater prior to that would be sufficiently sensitive and selective to measure and after treatment have been about several micrograms per a variety of classes of compounds in the low ng/L range; (3) liter. The concentrations measured in surface-water samples to apply the method to samples from a range of surface- and downstream from sewage-treatment plant discharges typi- ground-water sites throughout the USA; and (4) to describe cally have been in the tens of nanograms per liter, although the long-term performance of the method and assess its suit- concentrations in the ␮g/L range are possible. Although ability to routine monitoring. Goals 1, 2, and 4 are the focus these concentrations are much lower than typical maximum of this report. Goal 3 is the subject of a separate report contaminant concentrations (in the tens of micrograms perliter) reported for other industrial contaminants, the effectsof continuous exposure to mixtures of pharmaceuticals on 2. Experimental
High-performance liquid chromatography coupled to mass spectrometry (HPLC–MS) and particularly tandemmass spectrometry (HPLC–MS–MS) have been the primary The prescription pharmaceuticals included in the method techniques used to determine ultratrace concentrations of (were selected in part based on data estimating the many pharmaceuticals in wastewater, surface water, and number of prescriptions written per year the typical ground water The long-term method performance dose, and typical number of doses per prescription. An esti- of HPLC–MS and HPLC–MS–MS for routine monitoring mate of the annual mass prescribed was calculated by using of environmental pharmaceutical concentrations previously In this report, the development of a combined solid-phase extraction (SPE) isolation and HPLC–electrospray ioniza- tion (ESI) MS analytical procedure for routine determina-tion of pharmaceuticals, primarily those associated with hu- man health, in surface- and ground-water samples, is de-scribed. This method development was part of a larger study There is uncertainty in these estimates because of varia- to assess the presence and distribution of organic wastewa- tions in dosing for individuals, variations in the size of the Table 1Pharmaceutical compounds determined by this method in order of elution CNS: central nervous system; NSAID: non-steroidal anti-inflamatory drug.
J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 dose, and the limited availability of proprietary prescription standard, at a concentration of 0.010 ␮g/␮L. The recon- data. In some cases the excreted metabolite is the primary stituted sample was filtered through a polytetrafluoroethy- form in wastewater; therefore, primary metabolites also were lene/Teflon (PTFE) 0.2-␮m syringe filter (Acrodisc [13C]R, considered in developing the list of pharmaceuticals ana- Pall Corp., East Hills, NY, USA). Samples were extracted in sets of up to 10 environmental samples, with two labo- Standards were obtained from commercial sources or ratory quality-control (QC) samples: an HPLC-grade water were provided by the manufacturer and typically were at sample containing only the performance surrogate and an purities of 95% or greater. An injection internal standard, HPLC-grade water sample containing the performance sur- 13C-labeled caffeine (Cambridge Isotope Labs., Cam- rogate and fortified at 1.0 ␮g/L of all compounds measured bridge, MA, USA), was used to quantify all compounds.
in this study. These laboratory samples were used to assess 13C-labeled Phenacetin (Cambridge Isotope Labs.) was possible laboratory contamination of the sample, formation used as the method-performance surrogate. Stock solutions of analytical artifacts, and the recovery of pharmaceuticals of all compounds, fortification solutions, and surrogate so- lutions were dissolved in HPLC-grade methanol (Burdick& Jackson, Muskegeon, MI, USA). Calibration and for- 2.4. Chromatography and mass spectrometry tification stock solutions were diluted in 10 mM, pH 3.7aqueous formic acid/ammonium formate buffer.
HPLC–MS was applied using a Hewlett-Packard (now Agilent Technologies, Palo Alto, CA, USA) Series 1100 HPLC/MSD system. An ammonium formate/formic acidbuffer (10 mM, pH 3.7) aqueous phase and acetonitrile Surface-water samples were collected using standard US were used to produce a multistep binary elution gradient Geological Survey (USGS) procedures for trace organic (The flow rate was 0.200 mL/min, and all flow was analysis Samples were shipped to the laboratory by directed to the mass spectrometer. Separations were made overnight express. Samples that could not be filtered in using a Metasil Basic 3 ␮m, 150 mm × 2.0 mm, C18 ana- the field were filtered in the laboratory, typically within lytical column coupled to either a Metasil Basic Safeguard 48 h. All samples were filtered using baked (450 ◦C for (MetaChem Technologies), 3 ␮m, 2.0 mm guard column, or 8 h) Whatman 0.7 ␮m GF/F (glass-fiber filters) (Whatman, NewGuard RP-18, 7 ␮m, 15 mm × 3.2 mm guard column The HPLC system was interfaced with the mass spec- trometer using ESI in the positive ionization mode. TheESI source conditions were as follows: source temperature Several SPE stationary phases were tested under a 150 ◦C, nebulizer gas pressure of 100 kPa, drying gas (ni- range of elution conditions for optimum recovery (data trogen) flow rate of 9 L/min, and drying gas temperature of not shown). Upon selection, pharmaceuticals were iso- 350 ◦C. The potential difference between the source and the lated from 1-L water samples using Waters (Milford, MA, capillary was held at 3500 V. Programmed capillary exit volt- USA), Oasis HLB, 0.5 g, 6-mL SPE cartridges with a age changes were used to produce sufficient fragmentation positive-displacement pump (Fluid Metering, Syosset, NY, of each compound (so that characteristic fragments USA). Prior to extraction, a performance surrogate con- were produced. In the instrument used in this study, the cap- taining 1.0 ␮g of [13C]1-ethoxy-phenacetin in 100 ␮L of illary exit voltage is referred to as the fragmentor voltage.
methanol was added to each filtered sample. The SPE car- A series of flow injection analyses were used to optimize tridges were conditioned with 6 mL of methanol, followedby vacuum drying at 1.6 kPa on a vacuum manifold for10–15 s, followed by 6 mL of HPLC-grade water. Samples were processed through the SPE cartridge at 15 mL/min.
Elution gradient used for high-performance liquid chromatographic sep-aration of pharmaceutical compounds, flow rate of 0.200 mL/min Following isolation, 1 mL of 5% methanol in reagent-gradewater was passed through the cartridge using a vacuum manifold, and the water discarded. The cartridge then was sequentially eluted with two 3-mL aliquots of methanol fol- lowed by two 2-mL aliquots of methanol acidified to pH 3.7 with trifluoroacetic acid (Sequanal grade, Pierce Biotech- nology, Rockford, IL, USA). The combined aliquots were (Zymark, Hopkinton, MA, USA) evaporative concentra- tor, using nitrogen at a pressure of 34.5 kPa, in a 40 ◦C bath. Samples were reconstituted with 800–900 ␮L of for- mate buffer and 100 ␮L of [13C]caffeine-3-methyl internal J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 Table 3Instrumental parameters for pharmaceutical compounds determined by using high-performance liquid chromatography–mass spectrometry under selected-ionmonitoring conditions 114.0, 208.0
98.1, 177.1
166.1, 222.1, 240.1
84.1, 246.1, 406.1
355.1, 373.2, 391.2
268.0, 284.0, 345.0
163.1, 251.1, 309.1
205.2, 233.2, 273.0
Quantitation ions in bold; confirmation ions in normal font. [13C]caffeine-3-methyl—internal standard. Phenacetin [13C]—surrogate.
a Mass + hydrogen.
b Mass + sodium.
the fragmentor voltage for each compound. For each com- market-tracking sources was obtained from literature pound, the optimal detection conditions for the protonated The source of the prescription data was an audit molecular ion and at least one confirming fragment ion were that tracked 2.486 × 109 prescriptions dispensed by 35 000 used when collecting data in the selected-ion monitoring community pharmacies from December 1997 through (SIM) mode, thereby increasing the sensitivity of detection November 1998, and projected to provide national esti- mates. Selected pharmaceuticals were drawn from the top A multipoint internal standard calibration, from 0.010 to twenty most commonly prescribed drugs and the most 2.0 ␮g/L, was used for each sample set analyzed. Calibra- commonly used nonprescription drugs Except for al- tion was monitored through the use of continuing calibra- buterol and digoxigenin, all compounds were selected on tion verification (CCV) samples, and if the calibration was the basis of these usage criteria. Albuterol and digoxigenin within ±20%, analysis of environmental and laboratory QC were included in the selected compound list as a negative samples was continued. Instrument blanks to monitor poten- control, because it was hypothesized that these compounds tial carryover between injections were interspersed between would not be detected in surface-water samples because of every 10 injections prior to a CCV. The HPLC–MS analy- ses were quantified using commercial quantitation software An effective chromatographic separation (was (Target 4.0; Thru-Put Systems, Boca Raton, FL, USA).
achieved by using the stepped gradient listed in response of the individual compounds varied substantially(This variation was expected because of the diversity 3. Results and discussion
of chemical classes among the selected pharmaceutical com-pounds under electrospray ionization conditions. However, Information for the most frequently prescribed drugs overall response for all compounds was sufficient for anal- for the year 1998 from a compilation by commercial ysis at the expected ambient environmental concentrations.
J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 Fig. 1. Representative high-performance liquid chromatography/mass spectrometry chromatogram of a standard solution of the pharmaceuticals determinedin this study. The amount of each compound injected was 0.050 ␮g, equivalent to a 1-␮g/L concentration in a 1-L sample.
Flow injection analysis of each compound was 0.1 min of standard retention times. Fragmentor voltages used to determine the fragmentor voltage required to pro- were set for groups of compounds because closely eluting duce a definitive mass spectrum at optimal sensitivity. The compounds could not always be separated. For some com- characteristics of a definitive mass spectrum for qualitatively pounds this procedure resulted in a compromise fragmen- identifying compounds consisted of a protonated molecu- tor setting, which had the effect of either yielding frag- lar ion and at least one confirming fragment ion in excess ment ions in less than 20% relative abundance or fragment of 20% abundance of the molecular ion, and elution within ions in greater abundance than the molecular ion. The final Fig. 2. The effect of fragmentor voltage on the fragmentation of acetaminophen, a representative pharmaceutical compound.
J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 Table 4Percent recoveries and relative standard deviation of pharmaceutical compounds from solid-phase extraction compared to long-term recoveries ofpharmaceuticals determined from fortified reagent-water samples analyzed with environmental samples Solid-phase extraction trial and 1999 spikes fortified with 1.0 mg of each compound and 2000 spikes fortified with 1.0 or 0.5 ␮g/L. Five unrecoveredcompounds not listed.
fragmentor voltages, quantitation and confirmation ions, and only slightly, 45–88%, but with poor reproducibility using their relative abundances used to confirm identification are C18 SPE sorbent. Subsequent studies on the effect of sam- ple pH did not reveal any significant improvements to over- The variety of chemical classes represented in the ana- all extraction efficiency (data not shown). The final choice lyte list provided varying results among the solid-phase me- for the HLB cartridges was based on overall recovery and dia tested in preliminary trials and required compromises in consistency of recoveries over multiple months. In addition, making a final choice of extraction media. The best overall some compounds were recovered using HLB cartridges that solid-phase extraction recoveries were achieved using Oa- were not recovered using other SPE sorbents tested, such sis HLB extraction cartridges with no pH adjustment of the as acetaminophen, which is suspected to be ubiquitous in sample. The average recoveries for the compounds tested are wastewater-effected surface water. Compounds that were re- listed in Twelve of the 22 compounds tested were covered less than 60% were retained in the method to inves- recovered at greater than 60%. Highly polar compounds, tigate possible sample matrix effects and to compare with such as the histamine (H2) receptor antagonists ranitidine related methodologies applied to different matrices, such as and cimetidine, were recovered at less than 50%. The ex- sediments and other solids, planned for subsequent projects.
tremely polar compound metformin (C4H11N5) and the un- These compounds are qualified as estimated concentrations stable ␤-lactam antibiotic amoxicillin were not recovered at and any compounds that are recovered less than 30% are all. Low recovery for polar compounds is believed to be a reported only as qualitatively being present and are only re- result of poor retention on the polymeric sorbent as a result tained in the method if they are deemed sufficiently impor- of not adjusting pH of the sample for extraction. Cimetidine tant in regards to possible environmental impact or human and ranitidine and metformin did however have recoveries greater than 75% from octadecyl (C-18) SPE cartridges. All Future strategies in SPE development would include use three compounds have a more linear molecular shape, which of tandem or mixed modes of extraction sorbents to en- may better suit retention on C18. Amoxicillin, with an unsta- hance multiclass methodology. In addition, the development ble ␤-lactam structure, is known to degrade in solution. Re- of newer mixed mode phases with ion-exchange capability coveries of less than 30% were observed for the amphoteric provide a possible means to isolate specific classes of com- angiotensin-converting enzyme (ACE) inhibitors, lisinopril pounds and better eliminate interferences from dissolved or- and enalaprilat, and for paroxetine metabolite. Enalaprilat and lisinopril likewise are believed to be poorly retained The method detection limit (MDL) was determined ac- in the polymeric sorbent without pH adjustment. Enalapri- cording to US Environmental Protection Agency guidelines lat did recover greater than 90% using a graphitized carbon MDL is defined as the minimum concentration of SPE sorbent. Enalaprilat and lisinopril recoveries improved a substance that can be measured and reported with a 99% J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 HPLC–MS–MS, which range between 0.010 and 0.050 ␮g/L Method detection limits (MDL) for pharmaceutical compounds, in micro- grams per liter (␮g/L) determined in this method and listed in order of Recoveries from an initial validation of the method were compared to the recoveries from quality-control samples processed with environmental samples. The quality-control sample was a reagent-water sample amended to a concen- tration of 1.0 ␮g/L of each compound. Sixteen of these reagent-spike samples were extracted in 1999. In 2000, 28 reagent-spike samples were amended at 0.50 ␮g/L, ex- tracted, and analyzed. These set spikes are used to evaluate method performance over an extended time. Several po- tential sources of variation are built into these long-term recoveries, including multiple operators and instruments (Recoveries for SPE trials, extracted and ana- lyzed in triplicate, ranged from 8 to 127%, with a mean of 77%. The relative standard deviation (R.S.D.) ranged from 1 to 20.8%, with a mean of 9.1%. Recoveries for the year 1999 set spikes ranged from 11 to 87%, with a mean of 62%. The R.S.D. for year 1999 set spikes ranged from 3.6 to 19.1%, with a mean of 8.1%. Lower recover- ies for the year 1999 reagent set spikes, compared to the initial SPE trials, likely are attributable to the addition of a syringe-filtering step. This step was applied consistently to all sample extracts and set spikes to prevent clogging of theHPLC column from precipitates that formed in environmen- a Not detected under experimental conditions.
tal sample extracts during concentration. The decrease inR.S.D. observed between the SPE trials and the year 1999 confidence that the compound concentration is greater than set spikes is not substantial. These results indicate that the zero, and is determined from at least seven replicate analyses recoveries are sufficiently similar before and after addition of samples containing the compounds of interest. Eight 1-L of the required syringe-filtration step and that this method water samples, collected upstream from known or suspected is suitable for short- and long-term application.
sources of contamination of the compounds of interest, were Set-spike recoveries for year 1999 also were compared amended to a concentration of 0.10 ␮g/L for each compound.
to data for the year 2000 set spikes to indicate long-term The samples were processed as a single set through the en- performance (The mean recovery for all set spikes tire extraction and analytical procedure. The MDL for each in 2000 for all compounds was 62%. Mean set-spike recov- compound was determined from the standard deviation of eries for individual compounds ranged from 12 to 166%.
concentration for the replicate measurements, which is mul- The R.S.D.s of year 2000 set spike recoveries for individual tiplied by the Student’s t-value for (n − 1) degrees of free- compounds ranged from 7.8 to 61.4%, with a mean of all dom. The resulting MDL for each of the compounds is listed R.S.D.s of 16.7%. Mean recovery changes between 1999 in few compounds, such as metformin, are poorly and 2000 varied by compound. Mean recovery decreases of recovered, but have relatively low MDLs (0.0034 ␮g/L for greater than 20% between 1999 and 2000 were observed metformin) because the MDL is determined using the abso- for five compounds, and mean recovery increases of greater lute standard deviation of concentration. This result indicates than 20% were observed for three compounds. Mean recov- that the recovery and MDL need to be evaluated together to ery changes between 1999 and 2000 were less than 20% determine the performance of the compound in the method.
for nine compounds. These data indicate that there is no The mean MDL for all compounds is 0.022 ␮g/L and the systematic change in recoveries between years, given the median 0.014 ␮g/L. The MDLs for compounds previously multiple potential sources of variation in the data. However, identified as chromatographing poorly, such as lisinopril, the there is evidence in this data set that the effects leading metabolite of paroxetine, and compounds that were not re- to signal enhancement, which are thought to occur in the covered by SPE, were not determined, and are not listed HPLC–MS, are producing enhanced recoveries for some in The high MDL for enalaprilat, 0.15 ␮g/L, is compounds. These enhanced recoveries are observed as likely caused by variability in recovery, resulting in a high greater than expected spike recoveries and continuing cali- standard deviation of concentration and the resulting high bration verification (CCV) standard concentrations (data not shown) for some compounds. The compounds that demon- tween 0.006 and 0.030 ␮g/L, comparing favorably with pub- strate substantial signal enhancement are acetaminophen, lished limits of detection (LODs) for other studies that use 1,7-dimethylxanthine, digoxigenin, and gemfibrozil (in- J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 creased CCV concentration). Enhancement was seen only that the internal standard is, other compounds are not and after running surface-water sample extracts, and no com- therefore produced significant differences in observed con- pounds were detected in instrument blanks and method centrations In order to eliminate or at least min- blanks. Peak areas and recoveries fell within expected lim- imize such affects another internal standard was used that its for CCV standards when calibration standards were would not be affected by complex matrices. Isotopically la- analyzed just prior to a CCV standard and prior to sample beled nicotinamide was substituted as an internal standard extracts, whereas enhanced peak areas were measured in because it elutes much earlier in the chromatogram and is the same standards when analyzed after sample extracts.
therefore much less affected by later eluting interferences.
Signal enhancement was as low as 20–30% increase for It was tested against labeled caffeine as an internal standard acetaminophen, digoxigenin, and gemfibrozil, and as high in serial dilutions of surface water with high concentrations as 50–100% for acetaminophen and 1,7-dimethylxanthine.
of DOM measuring concentration of an isotopically labeled Replacing the analytical HPLC guard column and analyt- surrogate, ethyl nicotinate. The observed concentration re- ical HPLC column improved the apparent enhancement mained within ±20% using the labeled nicotinamide inter- slightly. Decreasing sample analyses to only one set of 10 nal standard while the observed concentration using labeled sample extracts, cleaning the MS source, and recalibrating caffeine rose to 160% recovery from undiluted water (data between each sample set resulted in the most important im- provement, as evidenced in normal or only slightly elevated Recoveries of pharmaceuticals from surface-water sam- ples prepared as a single set and spiked at 1.000 ␮g/L were Compound specific matrix interferences have been noted compared to mean spike recoveries for 16 set spikes pre- by measuring concentrations of spiked compounds in serial pared using the same method for about 6 months ( dilutions of surface water, containing high organic matter Recoveries ranged from 0 to 123% with a mean of 57%, in- concentrations, with reagent grade water. Some compounds cluding zero recoveries, and 73% for all nonzero recoveries.
were seen to increase in measured concentration by up to Matrix enhancement as well as suppression of recoveries oc- 100% while others remained within ±20% of the average curred although the enhancement is neither as marked nor as measured concentration using an isotopically labeled caf- consistent as that observed in similar methodology applied feine internal standard. The response of the caffeine internal to detection of herbicides Lisinopril and enalaprilat standard in the MS detector was seen to decrease as dilu- are amphoteric compounds that chromatograph poorly under tion of surface water decreased. The variability of effect on present conditions. Lisinopril as well as paroxetine metabo- the MS signals of compounds as affected by the aqueous lite were not quantifiable greater than the baseline signal.
matrix reflects the variability in measured concentrations.
Enalaprilat had a substantially enhanced signal and recovery While some compounds are affected in the same manner from environmental matrices. Cimetidine, which is highly Measured Concentration
Relative Response
Relative Response at Detector
Measured Concentration ug/L
% High Matrix Water
% High Matrix Water
Fig. 3. (a) Measured concentration and (b) relative response of selected pharmaceuticals in serial dilutions of surface water high in dissolved organicmatter spiked at 0.5 ␮g/L (codeine not included in initial study).
J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 Percent
n
Recovery, i
Fig. 4. Effect of natural environmental matrix components in surface-water (SW) samples on the high-performance liquid chromatography/mass spectrometricdetermination of pharmaceutical compounds as compared to organic-free reagent-water (RW) samples.
polar, also shows substantial enhancement, which appears to maceuticals in USA streams in 1999 and 2000 The be the trend among the more polar basic compounds. Am- pharmaceuticals measured by this method were detected photeric and neutral and acidic compounds show an over- frequently, with individual compounds present in as many all tendency toward matrix-related suppressed recoveries.
as 61.9% of samples tested. The most frequently detected Furosemide was suppressed the most as the signal could not compounds were nonprescription compounds, such as ac- etaminophen (23.8%), caffeine (61.9%), and cotinine (a Five pairs of duplicate field samples were analyzed to nicotine metabolite, 38.1%). Median concentrations mea- characterize method precision at ambient environmentalconcentrations and in the presence of interferences. Dupli-cate samples were analyzed within a sample set (intrasetreplicates) or in separate sample sets (interset replicates), and results for both were combined for data analysis.
Boxplots of the relative percent difference (RPD) for six pharmaceuticals that were detected in four or more of the duplicate samples, and the performance surrogate are shown in With the exception of caffeine, all compounds averaged 39% RPD or less. The performance surrogate ([13C]1-ethoxy-phenacetin) that was added at a concen- tration of about 1.000 ␮g/L had a RPD of 9% and was accounted for in all 10 samples. The concentrations of the pharmaceuticals detected ranged from 0.009 to 1.02 ␮g/L with a median concentration of 0.082 ␮g/L. The low con- centrations detected, together with a low number of samples for comparison, likely result in the somewhat higher RPDsobserved in this report when compared to those reported for sulfonylurea, imidazolinone, and sulfonamide herbicides 1,7-DimethylxanthineTrimethoprim Caf eine The method described herein has been applied to sam- Fig. 5. Relative percent differences between individual pharmaceutical ples collected as part of a national reconnaissance of phar- compounds determined in duplicate environmental water samples.
J.D. Cahill et al. / J. Chromatogr. A 1041 (2004) 171–180 sured were 0.110 ␮g/L for acetaminophen, 0.081 ␮g/L cate what chemical characteristics are amenable to existing for caffeine, and 0.024 ␮g/L for cotinine; maximum con- centrations for these compounds were 10, 6, and 9 ␮g/L,respectively.
Prescription pharmaceuticals were less frequently de- Acknowledgements
tected, with the highest frequencies of detection fortrimethoprim (27.4%) and sulfamethoxazole (19%), two The authors wish to thank the numerous USGS District antibiotics used in combination for treating urinary tract Office staff who provided samples for analysis of ambient and ear infections. Median concentrations for these antibi- pharmaceutical concentrations. We gratefully acknowledge otics were 0.013 ␮g/L for trimethoprim and 0.066 ␮g/L insights provided in reviews from Susan Glassmeyer and for sulfamethoxazole, with maximum concentrations of Colleen Rostad of the USGS. This research was performed 3 and 5.2 ␮g/L, respectively. Other pharmaceuticals de- through the Toxic Substances Hydrology Program, USGS.
tected included diltiazem, 1,7-dimethylxanthine (a caffeine Any mention of trade, product, or firm names is for identi- metabolite), dehydronifedipine, cimetidine, ibuprofen, met- fication purposes only and does not imply endorsement by formin, and gemfibrozil. Median concentrations for these compounds were all less than 0.20 ␮g/L. The detection ofmetformin suggests recovery improvements related to envi-ronmental sample matrix because reagent-spike recoveries References
[1] R. Porter, The Greatest Benefit to Mankind: A Medical History of Humanity from Antiquity to the Present, W.W. Norton, London, 4. Conclusions
[2] H.R. Buser, T. Poiger, M.D. Muller, Environ. Sci. Technol. 32 (1998) The routine determination of multiple classes of pharma- [3] H.R. Buser, M.D. Muller, N. Theobald, Environ. Sci. Technol. 32 ceuticals in surface- and ground-water samples at ambient environmental concentrations using SPE and HPLC–MS [4] H.R. Buser, T. Poiger, M.D. Muller, Environ. Sci. Technol. 33 (1999) has been shown to be practical and effective. The pre- [5] T.A. Ternes, R. Hirsch, J. Mueller, K. Haberer, Fresenius J. Anal.
scription and over-the-counter pharmaceuticals most likely found in USA surface- and ground-waters was evaluated [6] T.A. Ternes, P. Kreckel, J. Mueller, Sci. Total Environ. 228 (1999) from prescription data and resulted in a potential analyte list consisting of multiple chemical classes. The method [7] T.A. Ternes, M. Stumpf, J. Mueller, K. Haberer, R.D. Wilken, M.
performance and quality-control data presented indicate Servos, Sci. Total Environ. 228 (1999) 87.
[8] D.W. Kolpin, E.T. Furlong, M.T. Meyer, E.M. Thurman, S.D. Zaugg, that the techniques applied to routine analysis of surface- L.B. Barber, H.T. Buxton, Environ. Sci. Technol. 36 (2002) 1202.
and ground-water samples for pharmaceuticals is robust and [9] E. Mohle, C. Kempter, A. Kern, J.W. Metzger, Acta Hydrochim.
sensitive for the majority of compounds tested, at concen- trations less than 0.10 ␮g/L, with method detection limits [10] M. Stumpf, T.A. Ternes, R.D. Wilken, S.V. Rodrigues, W. Baumann, averaging 0.022 ␮g/L. Highly polar and amphoteric com- [11] N. Sandow, Rx List—The Internet Drug Index, RxList LLC, accessed pounds were the most difficult to isolate and detect. Preci- sion under controlled conditions, as reflected in mean spike [12] L.R. Shelton, US Geological Survey Open-File Report 94-455, Sacra- recovery R.S.D. of 16.7% is good, but the limited amount of data obtained from replicate environmental samples indicate [13] T. Heberer, U. Dunnbier, C. Reilich, H.J. Stan, Fresenius Environ.
that a larger data set is required to characterize precision at [14] T. Heberer, H.J. Stan, Int. J. Environ. Anal. Chem. 67 (1997) 113.
ambient environmental concentrations. Extensive QA/QC [15] T. Heberer, K. Schmidt-Baumler, H.J. Stan, Acta Hydrochim. Hy- data collection and interpretation, particularly of replicate samples, are necessary to validate the performance of this [16] J. Zoeller, Am. Druggist 215 (1999) 46.
method and to assess matrix effects on recoveries of the [17] US Environmental Protection Agency, in US Code of Federal Reg- selected compounds. SPE recovery data in this report are ulations, Title 40, revised as of 1 July 1992, p. 565.
[18] R. Hirsch, T. Ternes, K. Haberer, K.L. Kratz, Sci. Total Environ.
useful in describing the directions for further exploration in developing methods to extract more specific classes of com- [19] E.T. Furlong, M.R. Burkhardt, P.M. Gates, S.L. Werner, W.A.
pounds and enhance extraction recoveries, as well as to indi- Battaglin, Sci. Total Environ. 248 (2000) 135.

Source: http://clasfaculty.ucdenver.edu/landerso/Pharmaceuticals2004.pdf