Science of the Total Environment 472 (2014) 862–871

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

A snapshot of illicit drug use in Sweden acquired through sewage water analysis Marcus Östman ⁎, Jerker Fick, Elin Näsström, Richard H. Lindberg Department of Chemistry, Umeå University, SE-901 87 Umeå, Sweden

H I G H L I G H T S • • • •

An online-SPE LC–MS/MS method for illicit drugs was developed. Incoming water from 33 sewage treatment plants was analyzed for illicit drugs. Regional drug consumption differences were highlighted by multivariate data analysis. The differences were supported by prescription statistics for pharmaceuticals.

a r t i c l e

i n f o

Article history: Received 25 September 2013 Received in revised form 16 November 2013 Accepted 16 November 2013 Available online xxxx Keywords: Illicit drugs Wastewater Analysis LC–MS/MS Multivariate data analysis Sweden

a b s t r a c t Analytical measurements of sewage water have been used many times to estimate the consumption of specific drugs in an area. This study measured a large number of illicit drugs and metabolites (N30) at a large number of sewage treatment plants (STPs) distributed across Sweden. Twenty-four illicit and prescription drugs, classified as narcotic substances in Sweden, and seven selected metabolites were included in the study. A 24 hour composite sample of incoming sewage water was collected from 33 different municipalities at various geographic locations across Sweden. Species were analyzed using an on-line solid-phase extraction–liquid chromatography electrospray tandem mass spectrometry method. The method proved to be rapid with minimum need for sample work up and was able to detect 13 compounds above their respective limits of quantification. The results for all compounds were presented as per capita loads. Multivariate data analysis was used to relate drug consumption to geographical location and/or population of cities. The results showed that geographical differences in drug consumption were apparent across the country. For the narcotic pharmaceuticals, the geographical differences suggested by the multivariate model were supported by prescription statistics. © 2013 Elsevier B.V. All rights reserved.

1. Introduction The monitoring of illicit drug use in a given community is an important tool for improving public health. Traditionally, this has been carried out by indirect methods, such as population surveys, medical records, and data on seizures (European Monitoring Centre for Drugs and Drug Addiction, 2002). Daughton (2001) suggested that it should be possible to back calculate the use of illicit drugs from levels measured in incoming sewage water. Zuccato et al. confirmed that this was feasible in 2005 and since then, several studies have been published (Baker and Kasprzyk-Hordern, 2013, 2011a; Bijlsma et al., 2012; Chiaia et al., 2008; de Voogt et al., 2012; Fontanals et al., 2013; Karolak et al., 2010; Lin et al., 2010; Postigo et al., 2011, 2008; Reid et al., 2011; Terzic et al., 2010; Thomas et al., 2012; van der Aa et al., 2013; van Nuijs et al., 2011a, 2011b, 2009a, 2009b; Zuccato et al., 2005).

⁎ Corresponding author. Tel.: +46 90 786 9320. E-mail address: [email protected] (M. Östman). 0048-9697/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.scitotenv.2013.11.081

These types of measurements are a valuable complement to classic methods as they are generally faster and less labor intensive. Back calculation from measured levels in sewage water to usage in a community is an approach that is still under development and several factors need further investigation. Castiglioni et al. (2013) have conducted an investigation of the whole process from sampling to back calculation to evaluate the uncertainty of all steps. Khan and Nicell (2012, 2011) have recently made in-depth studies of the mass balances of cocaine, heroin, ecstasy, methamphetamine, amphetamine and tetrahydrocannabinol (THC). For these illicit drugs, the excretion pattern for the major metabolites is known, which is an essential factor in back calculations. To make a good back calculation, it is crucial to know which metabolite to use and to what extent it is excreted from the body. Another important factor that may bias the results is the number of people served by each plant. Since this may differ dramatically, e.g., due to commuting, special events, or seasonal changes, a static approach has several disadvantages, as pointed out by Van Nuijs et al. (2011b). Alternative methods, e.g., using parameters such as the biological oxygen demand (BOD) and chemical oxygen demand (COD) to calculate the number

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871

of people connected each day, have the advantage of reflecting relevant fluxes in the population. Unfortunately, it is impossible to differentiate the industrial load from the domestic load. Adsorption of drugs to particulate matter in sewage water may also influence the results. However, Baker and Kasprzyk-Hordern (2011a) have recently shown that many illicit drugs adsorb by less than 10% to particulate matter, with the exceptions of methadone and 2-ethylidene-1,5-dimethyl-3,3diphenylpyrrolidine (EDDP). Monitoring studies can be labor-intensive and time-consuming, mostly due to the manual steps involved in the pre-treatment, e.g., filtration and solid phase extraction (SPE). This problem can be minimized by using semi-automated on-line SPE techniques. On-line methods are generally more rapid, precise and efficient than conventional techniques because they eliminate time-consuming evaporation and reconstitution steps and require minimal sample handling. On-line SPE is also more environmentally friendly since it requires less solvent, has a high sample throughput and utilizes comparatively small sample volumes. One major drawback of on-line SPE methods is that they generally have lower pre-concentration factors in comparison to off-line SPE of aqueous environmental samples, and therefore slightly higher method limits of quantification (LOQs). Nevertheless, on-line SPE has been used successfully in the analysis of both biological and environmental samples for various analytes (Fontanals et al., 2013, 2011; Khan et al., 2012; Lopez-Serna et al., 2010; Postigo et al., 2008; Segura et al., 2007). Several on-line methods have also been developed and used in studies of illicit drugs (Fontanals et al., 2013, 2011; Ghassabian et al., 2012; Postigo et al., 2008; Saussereau et al., 2012; and also reviewed by Vazques-Roig et al., 2013). However, most previous studies were not developed to study illicit drugs in incoming wastewater, and to the best of our knowledge, the studies of Postigo et al. (2008) and Fontanals et al. (2013) are the only ones to employ online methods for sewage water analysis of illicit drugs. The use of illicit drugs often varies with location and has been investigated with sewage water analysis before (Thomas et al., 2012). We propose that a better way to detect patterns in the consumption of illicit drugs and prescription narcotic substances is to subject the data to multivariate data analysis, providing an easy way of obtaining an overview of illicit drug consumption. The aims of this study were (a) to develop a method for measuring 24 illicit and prescription narcotic substances and seven of their metabolites in incoming sewage water using an efficient on-line SPE method, (b) to acquire a snapshot of the usage of illicit drugs in different municipalities/cities in Sweden, and (c) to investigate if drug use is dependent on the size of the city and/or its geographical location. 2. Materials and methods

863

acid (puriss. p.a.) was obtained from Fluka (Steinheim, Germany). Amphetamine, benzoylecgonine, cathinone HCl, 6-acetylmorphine, cocaine, heroin, ketamine HCl, n-methyl-1,3-benzodioxolylbutanamine (MBDB) HCl, 3,4-methylenedioxyamphetamine (MDA), methylenedioxyethylamphetamine (MDEA), 3,4-methylenedioxy-N-methylamphetamine (MDMA), methamphetamine, methylphenidate, midazolam, norketamine HCl, lysergic acid diethylamide (LSD), 2-oxo3-hydroxy-LSD, oxycodone, 2-ethylidene-1,5-dimethyl-3,3-diphenylpyrrolidine (EDDP) perchlorate, methadone, norbuprenorphine glucuronide, 11-nor-9-carboxy-Δ9-tetrahydrocannabinol (THC-COOH), 6-acetylmorphine-D6, amphetamine-D5, benzoylecgonine-D8, cocaine-D3, heroin-D9, MDA-D5, MDMA-D5, methadone-D9, methamphetamine-D5 and THC-COOH-D9 were bought from Cerilliant (Texas, USA) as 0.1 or 1 mg/mL standards in either methanol or acetonitrile. Alprazolam, buprenorphine HCl, codeine, fentanyl, flunitrazepam, morphine sulphate, oxazepam and zolpidem were purchased from Sigma Aldrich. Mephedrone was bought from the National Measurement Institute (Australia). Tramadol was bought from the Council of European Pharmacopoeia (Strasbourg) as 0.1 or 1 mg/mL standards in either methanol or acetonitrile. Codeine-D6, oxazepamD5 and tramadol-C13-D3 were purchased from the Cambridge Isotope Laboratories (Andover, MA, USA). All substances were classified as N99% pure except for 2-oxo-3-hydroxy-LSD (N97.7%) and LSD (N98.9%). All substances were stored according to the supplier's instructions. 2.3. Sampling and pretreatment Sewage treatment plants (STPs) were selected for sampling to achieve a wide geographical distribution across Sweden. In total, 33 plants were included, with all but two having more than 10,000 people connected to the STP (see Fig. 1). Samples, i.e., 24 hour flow proportional composite samples of incoming sewage water, were collected by the staff of each plant in January 2012. All samples were supposed to be collected on the same date – the 17th January 2012 (a Tuesday) – but some exceptions were made (see sampling dates listed in Table 1). Additional parameters measured in the wastewater are detailed in Table S14 of the Supplementary Information. Only one sample per STP was collected for practical reasons. Samples had to be collected during a weekday since it was difficult to obtain weekend samples from such a large number of STPs. Samples were collected in 250 mL high density polyethylene bottles (HDPE) and were sent frozen to the laboratory for analysis. Prior to injection on the analytical instrumentation, samples were thawed and subjected to syringe filtration (0.45 μm Filtropur S, Sarstedt, Nümbrecht, Germany), addition of internal standards (ISs) (to a final concentration of 500 ng L−1) and acidification to pH 3 using formic acid (FA).

2.1. Selection of illicit drugs 2.4. Analytical instrumentation The selection of illicit drugs included in this study was based on a) information on the established illicit drugs in Europe (European Monitoring Centre for Drugs and Drug Addiction, 2013b), b) previous findings from studies in sewage waters (Baker and Kasprzyk-Hordern, 2013, 2011a, 2011b; Bijlsma et al., 2012; Chiaia et al., 2008; de Voogt et al., 2012; Fontanals et al., 2013; Karolak et al., 2010; Lin et al., 2010; Postigo et al., 2011, 2008; Reid et al., 2011; Terzic et al., 2010; Thomas et al., 2012; van der Aa et al., 2013; van Nuijs et al., 2011a, 2011b, 2009a, 2009b), and c) plausible emerging illicit drugs (or metabolites thereof) mentioned in various open access web forums. 2.2. Chemicals and reagents For all standards and eluents, Milli-Q (resistivity 18.2 MΩ cm−1) water from a Millipore gradient ultrapure water system (Millipore, USA) was used. Methanol (Liqrosolve, Hypergrade) used in eluents were bought from Merck (Darmstadt, Germany) and formic

A Thermo TSQ Quantum Ultra (Thermo Fisher Scientific, San Jose, CA, USA) mass spectrometer was used. Samples were injected using a PAL HTC auto sampler (CTC Analytics AG, Zwingen, Switzerland) into a 1 mL stainless steel loop. The auto sampler was equipped with two valves and their configuration has been described elsewhere (Khan et al., 2012). A Surveyor LC pump (Thermo Fisher Scientific, San Jose, CA, USA) was used to load the sample from the 1 mL loop on to an online SPE Oasis HLB column (2.1 × 20, 15 μm; Waters, Ireland), where it was enriched using water. After 1.5 min, the valve was switched and the Accela pump (Thermo Fisher Scientific, San Jose, CA, USA) was used to extract the compounds from the Oasis column using a gradient flow of methanol. The gradient program for the Accela and Surveyor pumps is detailed in Tables S1 and S2, respectively, in the Supplementary Information. Following elution from the Oasis column, the analytes were separated on an analytical column (Thermo Hypersil GoldAQ, 50 × 2.1 mm, 5 μm + guard column), using the gradient program

864

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871 Table 1 Sampling sites, sampling dates and information about the sewage treatment plants (STP). Number City

Name of STP

Sampling date Flow rate at People sampling day connecteda (m3 d−1)a

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Gällivare Haparanda Luleå Piteå Skellefteå Lycksele Umeå Örnsköldsvik Östersund Härnösand Söderhamn Mora Borlänge Gävle Arvika Karlstad Örebro Köping Eskilstuna Nyköping Norrköping Stenungsund Gothenburg Bollebygd Borås Visby Oskarshamn Kalmar Karlskrona Halmstad

18,000 22,000 63,000 30,500 42,618 8950 92,118 11,500 50,300 23,500 14,500 14,300 44,700 86,000 14,884 65,000 112,214 21,700 82,700 34,000 115,900 14,500 658,114 4100 80,000 32,000 19,300 55,000 44,000 70,000

2012-01-17 2012-01-17 2012-01-24 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-23 2012-01-17 2012-01-24 2012-01-18 2012-01-17 2012-01-18 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17 2012-01-17

4979 8039 21,400 9716 14,375 2443 34,990 5365 16,934 8140 5587 4607 14,701 38,547 7343 24,540 42,260 9749 56,000 14,124 54,000 6000 352,512 1118 63,498 12,943 14,430 19,151 30,457 37,536

31 32 33

Helsingborg Hässleholm Trelleborg

Kavaheden Bottenviken Uddebo Sandholmen Tuvan Lycksele Ön Knorthem Göviken Kattastrand Granskär Solviken Borlänge Duvbacken Arvika Sjöstad Skebäck Norsa Ekeby Brandholmen Slottshagen Strävliden Rya Bollebygd Gässlösa Visby Enemar Kalmar Koholmen Västra stranden Öresund Hässleholm Trelleborg

121,000 29,997 28,107

2012-01-17 2012-01-17 2012-01-17

52,862 16,870 15,859

a

Fig. 1. Sampling sites. 1 = Gällivare, 2 = Haparanda, 3 = Luleå, 4 = Piteå, 5 = Skellefteå, 6 = Lycksele, 7 = Umeå, 8 = Örnsköldsvik, 9 = Östersund, 10 = Härnösand, 11 = Söderhamn, 12 = Mora, 13 = Borlänge, 14 = Gävle, 15 = Arvika, 16 = Karlstad, 17 = Örebro, 18 = Köping, 19 = Eskilstuna, 20 = Nyköping, 21 = Norrköping, 22 = Stenungsund, 23 = Gothenburg, 24 = Bollebygd, 25 = Borås, 26 = Visby, 27 = Oskarshamn, 28 = Kalmar, 29 = Karlskrona, 30 = Halmstad, 31 = Helsingborg, 32 = Hässleholm, 33 = Trelleborg.

provided by the Accela pump. The flow passed through a PEEK capillary and the analytes were ionized by electrospray ionization (ESI) before entering the mass spectrometer. The parameters of the mass spectrometer were optimized in a semi-automatic way using the manufacturer's MS software. For the ESI probe, the temperature of the capillary was held at 325 °C. The flow of sheath gas was set to 60 and the auxiliary flow to 25 in arbitrary units. The source voltage was kept at 3.0 kV for all analytes. Individual collision energies and tube lens values for all analytes are shown in Table S3. The cycle time was set to 1.0 s and the Q1 detection width (FWHM) was 0.70. The collision gas pressure was 1.5 mTorr. 2.5. QA/QC Thirteen different isotopically labeled ISs were used for quantification in this study. Matching of ISs was based on similarities in analyte structure and retention time (Table 2). Positive identification of analytes was made on the basis of two transitions i.e., involving one precursor ion and two product ions, using the criterion that the ratio between the transitions was not allowed to deviate from the ratio in the calibration standard by more than +/− 30%. Moreover, the retention times for all analytes had to be within +/− 2.5% of the calibration standard. Together, this gave four identification points (the highest possible number),

Data provided by the STP.

as described in the Commission Decision 2002/657/EC (2002) concerning the performance of analytical methods and interpretation of results. The limit of quantification (LOQ) was determined from standard curves based on repeated measurements of low level spiked water (MilliQ and wastewater); the lowest point in the standard curve that had a signal/noise ratio N10 was considered to be equal to the LOQ. A seven point calibration curve over the range of 1–1000 ng L− 1 was used for quantification. The calibration curve was prepared in Milli-Q water and was spiked with the ISs. Recovery tests were also conducted to investigate the effect of using different syringe filters. Six syringe filters for sample preparation were used; 0.45 μm Filtropur S (Sarstedt, Nümbrecht, Germany), 0.20 μm Filtropur S (Sarstedt, Nümbrecht, Germany), 0.45 μm MF Millex (Millipore, Ireland), 0.22 μm MF Millex (Millipore, Ireland), 0.45 μm PTFE-membrane (VWR, USA) and 0.20 μm PTFE-membrane (VWR, USA). A batch of Milli-Q water was spiked with all compounds and then filtered in triplicate with each filter separately. After filtration, each sample was spiked with IS to a concentration of 300 ng L−1 and the results were compared with those with unfiltered water. A bench top stability test was performed to determine whether the analytes were stable under varying conditions, i.e., matrix and temperature. Purified water and filtered (0.45 μm Filtropur S, Sarstedt, Nümbrecht, Germany) sewage water were both spiked with analytes to 900 ng L−1 and analyzed in triplicate. Before analysis, all samples were spiked with IS and acidified with FA to a total concentration of 0.1%. Half of the spiked water (sewage and purified) was stored at room temperature and the other half at 3 °C for 24 h. Matrix effects were evaluated by comparing the area of the signal for standard solutions of the analyte at 900 ng L−1 in both incoming sewage water and Milli-Q water (with the area of non-spiked water of

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871 Table 2 Data of limits of quantification (LOQ), linearity (R2) from a seven point calibration curve (1–1000 ng L−1) and intra-day precision (n = 3 in Milli-Q water) for the method used in the study. Compound

Benzodiazepines Alprazolam Flunitrazepam Midazolam Oxazepam

LOQ (ng L−1)

R2

Intra-day precision RSD (%) n = 3

Internal standard

1.000 1.000 1.000 0.999

14 7.2 14 1.6

Oxazepam-D5 Oxazepam-D5 Oxazepam-D5 Oxazepam-D5

250

0.995

19

THC-COOH-D9

Hallucinogens LSD 2-Oxo-3-Hydroxy-LSD Ketamine Norketamine

10 5 5 5

0.999 1.000 1.000 1.000

1.4 4.3 4.5 3.5

Stimulants Amphetamine Cathinone Cocaine Benzoylecgonine MDA MDEA/MBDB MDMA Mephedrone Methamphetamine Methylphenidate

25 50 1 5 50 10 5 10 1 5

0.999 0.998 0.999 1.000 0.999 1.000 1.000 1.000 0.999 1.000

3.6 19 3.7 1.9 1.2 4.0 3.8 3.8 2.5 2.5

5 5

1.000 1.000

8.1 8.0

Oxazepam-D5 Benzoylecgonine-D8

50 1 15 5 10 1 50 30 5

0.999 1.000 0.997 0.999 0.999 0.998 0.999 1.000 0.999

4.5 3.1 2.1 4.3 1.9 4.2 9.0 1.8 5.7

Codeine-D6 Cocaine-D3 Heroin-D9 6-acetylmorphine-D6 Methadone-D9 Methadone-D9 Codeine-D6 Codeine-D6 Tramadol-C13-D3

1

1.000

2.0

Cocaine-D3

Opioids Buprenorphine Norbuprenorphine glucuronide Codeine Fentanyl Heroin 6-Acetylmorphine Methadone EDDP Morphine Oxycodone Tramadol Other drugs Zolpidem

aids the visualization and interpretation compared to techniques such as partial least squares (PLS). Internal model validation was performed with full (seven-fold) cross-validation (Stone, 1974). For the two-class OPLS-DA models, cross-validated ANOVA (Eriksson et al., 2008) p-values were calculated. 3. Results and discussion

10 10 5 10

Cannabinoids THC-COOH

865

Cocaine-D3 MDMA-D5 MDMA-D5 MDMA-D5

Amphetamine-D5 Codeine-D6 Cocaine-D3 Benzoylecgonine-D8 MDA-D5 MDMA-D5 MDMA-D5 MDMA-D5 Methamphetamine-D5 Cocaine-D3

Metabolites are written in italic.

each kind subtracted). The area for the respective IS was compared in the same way to see if it could be used to compensate for suppression/enhancement effects. Carryover effects were evaluated by injecting standards at 1200 ng L−1 followed by two mobile phase blanks. The linearity was evaluated by a seven point calibration curve over the range of 1–1000 ng L−1.

3.1. QA/QC results The results from the filter test are shown in Table S4 in the Supplementary Information. The Filtropur S filter (0.45 μm) had the highest average recovery (95%) and was therefore used in subsequent experiments. However, THC-COOH adsorbed strongly to this filter and could not be detected after filtration. Most analytes were shown to be stable in the bench top stability test (Supplementary Tables S5 and S6). However, seven analytes showed a recovery of less than 90% when stored in sewage water for 24 h at 4 °C (TCH-COOH 89%, LSD 82%, cathinone 85%, cocaine 82%, methylphenidate 64%, norbuprenorphine glucuronide 39%, codeine 82%, heroin 19%), and their degradation was more pronounced at room temperature. It should be noted that the decreased cocaine levels correlated with increased benzoylecgonine concentration when stored in sewage water for 24 h at room temperature. Matrix effects could be seen for all analytes, with a signal suppression varying from 45 to 78% in sewage water compared to Milli-Q water (Supplementary Table S7). The ISs were subjected to a similar suppression and the difference compared to the assigned analytes ranged from 0 to 5% units for most substances (Supplementary Table S7). Exceptions were morphine, cathinone, buprenorphine and the benzodiazepines, which were higher. The difference observed in the suppression between oxazepam and oxazepam-D5 (16% units) is difficult to explain, but in general, the ISs seemed to compensate very well for the signal suppression caused by the matrix. Some carryover effects (N 0.1%) could be observed for all analytes, except codeine, tramadol and oxycodone (data not shown) confirming the need to always perform a mobile phase blank run after high standards. Concentrations of analyte in samples were generally so low that it was not expected to cause problematic carryover effects. The linearity was good for all compounds; the correlation coefficient (R2) exceeding 0.99 in a seven point calibration curve. The average intra-day precision was 5.5%, with only four analytes having a variation above 10% (alprazolam 14%, midazolam 14%, THC-COOH 19% and cathinone 19%) (Table 2). Based on the results of the filter test, intra-day variation and the LOQ, TCH-COOH was excluded from the study. THC-COOH was also severely affected by matrix effects, which gave unreliable results even with the filters other than Filtropur S. 3.2. Main study results and discussion

2.6. Impact of geography and city size by multivariate data analysis Multivariate data analysis was performed to investigate whether any patterns related to catchment area of the STP and/or geographic location could be detected. The multivariate data analysis was carried out in SIMCA 13.0 (Umetrics, Umeå, Sweden). All data were mean-centered and scaled to unit variance prior to modeling. Due to large difference in concentration between the variables, scaling to unit variance was performed to ensure that all variables were weighted the same. Principal component analysis (PCA) (Wold et al., 1987) was initially performed to obtain an overview of the data. Orthogonal least squares discriminant analysis (OPLS-DA) (Bylesjö et al., 2006) was then carried out by grouping the samples into classes based on geographical location. The OPLS (Trygg and Wold, 2002) technique separates the variation in the X variables (here, drug concentrations) related to the Y variables (geographical location) from X variation uncorrelated (orthogonal) to Y, which

It is very important to note that this study only provides a snapshot of illicit drug usage in Sweden because only one sample was taken for each sample location. Several previous studies have demonstrated that the use of illicit drugs varies over time (both short term and long term) (Baker and Kasprzyk-Hordern, 2013; Reid et al., 2011; Thomas et al., 2012; van Nuijs et al., 2011b, 2009b). However, several previous measurements conducted by our laboratory during a 1–2 week period in Gothenburg and Umeå showed a relative standard deviation (RSD) of around 40% for most illicit drugs (data not shown). Several factors can lead to increased uncertainty and Castiglioni et al. (2013) have estimated the uncertainties of the following: population size estimation, stability of drug biomarkers, sampling, chemical analysis and back calculation. They recommended that as carried out here, staff at the STPs should be consulted when making population estimations to minimize the uncertainty. With regard to residence time in the sewer network,

866

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871

Gothenburg had a mean residence time of 2 h and all other STPs included in the study were substantially smaller. Sampling uncertainties are generally small (5–10%) in studies where flow proportional samples are used (Castiglioni et al., 2013). Note that due to a lack of knowledge about, e.g., the excretion pattern or stability in the sewer for many of the compounds included in this study, complete back calculations (including corrections for metabolism and degradation) were not performed as part of the main results. Instead, the results for all compounds were normalized against the flow in the STP and number of inhabitants connected, and expressed as mg (1000 inhabitants)−1 day−1. However, complete back calculations for cocaine (Table S9), amphetamine (Table S10) and methamphetamine (Table S11), as well as all the measured concentrations of all compounds included in the study are provided in the Supplementary Information (Table S12). Supplementary Table S8 also shows information about correction factors used for these calculations. Since back calculations were not used for the input data to the multivariate data analysis, one major uncertainty was eliminated. Of the 24 illicit and prescription narcotic substances and the seven metabolites included in the study, 13 were detected in the incoming sewage water samples. Loads, measured as mg (1000 inhabitants)−1 day−1 ranged from 0.1 for methadone to 700 for tramadol, and detection frequency ranged from 9.1% (3/33) for oxycodone to 100% (33/33) for oxazepam, codeine, morphine and tramadol. The highest loads and detection frequencies were observed for the prescription narcotic substances; 9 out of the 13 prescription narcotic pharmaceuticals were detected in the incoming sewage. Three of the 11 illicit drugs included in the study were detected, i.e., cocaine, amphetamine, methamphetamine and the cocaine metabolite benzoylecgonine. Several of the classic drugs (or the metabolites), such as heroin (+ 6acetyl-morphine), MDMA and LSD, were not found above their respective LOQ. Seizures of heroin by Swedish customs have decreased from 47.7 kg in 2010 to 5.9 kg in 2012 (Statistics of Seizures from the Swedish customs). This indicates that heroin use currently has a low prevalence in Sweden, which explains why it was not detected in this study. The major metabolite of heroin is morphine, which was detected but probably has a licit source from morphine or codeine used in the healthcare system (Khan and Nicell, 2011). LSD and MDMA are known to be less common in Sweden compared to many of the other illicit drugs (European Monitoring Centre for Drugs and Drug Addiction, 2013a). Furthermore, LSD is known to be used in certain regions of Sweden, such as Stockholm and Uppsala, which were not included in the study (European Monitoring Centre for Drugs and Drug Addiction, 2013a). As sampling was performed on a Tuesday, drugs primarily associated with weekend usage (like MDMA) were less likely to be detected. 3.2.1. Benzodiazepines Benzodiazepines are used therapeutically to treat anxiety, insomnia and alcohol withdrawal syndrome. Common illicit usage is to combine high doses of benzodiazepines with opiates to enhance the euphoriant effects, a combination that is a major risk factor in drug-related deaths (European Monitoring Centre for Drugs and Drug Addiction: Drug profiles). One of the four benzodiazepines included in this study, oxazepam, a short-acting benzodiazepine used legally in Sweden, was detected in all STPs (33/33) in the range from 20 to 200 mg (1000 inhabitants)− 1 day− 1 (Fig. 2). Detected levels of individual benzodiazepines are expected to correlate with usage patterns, which differ from country to country. However, these levels correspond well to previously measured levels in Europe (Baker and Kasprzyk-Hordern, 2013, 2011a). Oxazepam is also a metabolite of the common benzodiazepine, diazepam, which therefore also contributes to the detected levels of oxazepam (Ono et al., 1996). Diazepam is the most common benzodiazepine in Sweden, followed by alprazolam and oxazepam (Swedish National Board of Health and Welfare). Midazolam is the rarest of the benzodiazepines included in the study and was not detected (Ono et al., 1996). Measured concentrations of oxazepam were somewhat

elevated at STPs situated on the west coast, which correlates with the region traditionally displaying increased usage, as shown by the prescription statistics for oxazepam and diazepam (Swedish National Board of Health and Welfare). 3.2.2. Stimulants Stimulants are a class of compounds that act upon the central nervous system by increasing transmitter concentrations in both the noradrenergic and dopaminergic synapses, causing hypertension and tachycardia and feelings of increased confidence, sociability and energy (Wise, 1996). Stimulants also suppress appetite, fatigue and cause insomnia. Later, users may feel irritable, restless, anxious, depressed and lethargic (Wise, 1996). Four of the stimulants included in this study were detected: amphetamine, cocaine (and its metabolite benzoylecgonine), methamphetamine and methylphenidate. Methylphenidate, and to some extent amphetamine, are used legally in Sweden on prescription. Amphetamine was detected in 39% of the STPs (13/33) in the range from 10 to 140 mg (1000 inhabitants)−1 day−1 (Fig. 3). These levels are comparable to those reported in previous studies (Baker and Kasprzyk-Hordern, 2011a; Chiaia et al., 2008; Postigo et al., 2008; Terzic et al., 2010; Thomas et al., 2012). The highest loads were detected at the sampling locations Söderhamn (STP 11) and Gothenburg (STP 23); these loads were comparable to levels measured in Belgium and the Netherlands in a study performed in 19 European cities (Thomas et al., 2012). Comparable levels of amphetamine have also previously been measured in Sweden (Thomas et al., 2012), but only Umeå was a common sampling location to both the present and previous studies. Amphetamine is the second most used illicit drug in Sweden after cannabis, which explains the observed high detection frequency (European Monitoring Centre for Drugs and Drug Addiction, 2013a). Methamphetamine was detected in 48% of the STPs (16/33) in the range from 1 to 32 mg (1000 inhabitants)− 1 day− 1 (Fig. 3). These levels are somewhat higher than the average loads shown previously in several European studies but lower than loads measured in other Nordic countries (Norway and Finland) and in Eastern European countries (Baker and Kasprzyk-Hordern, 2013, 2011a; Postigo et al., 2008; Reid et al., 2011; Thomas et al., 2012; van Nuijs et al., 2011b). Methylphenidate is a psychostimulant drug used for the treatment of attention-deficit hyperactivity disorder (ADHD) and narcolepsy (Se ilir et al., 2013). Methylphenidate was detected in 94% of the STPs (31/33) in the range from 1 to 25 mg (1000 inhabitants)− 1 day −1 (Fig. 3). Methylphenidate is a drug that is gaining popularity. The number of prescriptions issued in 2012 was seven times higher than that in 2006 and now continues at a substantial level, which explains it being found at most of the STPs (Swedish National Board of Health and Welfare). Cocaine and its metabolite benzoylecgonine were detected in 36% of the STPs (12/33) in the range from 0.1 to 2 mg (1000 inhabitants)− 1 day− 1 (Fig. 4). These levels are low compared to levels reported in other European studies (Baker and Kasprzyk-Hordern, 2013, 2011a; Chiaia et al., 2008; Karolak et al., 2010; Lin et al., 2010; Postigo et al., 2008; Terzic et al., 2010; Thomas et al., 2012; van Nuijs et al., 2009a, 2009b; Zuccato et al., 2005). Even the highest loads, which were detected in Gothenburg (STP 23) and Helsingborg (STP 31), were well below the average recorded in previous studies (Baker and Kasprzyk-Hordern, 2013, 2011a; Chiaia et al., 2008; Karolak et al., 2010; Lin et al., 2010; Postigo et al., 2008; Terzic et al., 2010; Thomas et al., 2012; van Nuijs et al., 2009a, 2009b; Zuccato et al., 2005). This result correlates well with the fact that Sweden has a relatively low prevalence of cocaine usage among the general population (aged 15–64) and young adults (aged 15–34) compared to other European countries, and especially compared to the high prevalence countries Belgium, the Netherlands, Spain, the United Kingdom, Italy, Ireland and Denmark (European Monitoring Centre for Drugs and Drug Addiction, 2013b). Previous studies have seen a trend towards higher usage of cocaine in more urbanized municipalities/cities within a given country, and it

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871

867

Fig. 2. Load (mg (1000 inhabitants)−1 day−1) of oxazepam and oxycodone in Swedish STPs. 1 = Gällivare, 2 = Haparanda, 3 = Luleå, 4 = Piteå, 5 = Skellefteå, 6 = Lycksele, 7 = Umeå, 8 = Örnsköldsvik, 9 = Östersund, 10 = Härnösand, 11 = Söderhamn, 12 = Mora, 13 = Borlänge, 14 = Gävle, 15 = Arvika, 16 = Karlstad, 17 = Örebro, 18 = Köping, 19 = Eskilstuna, 20 = Nyköping, 21 = Norrköping, 22 = Stenungsund, 23 = Gothenburg, 24 = Bollebygd, 25 = Borås, 26 = Visby, 27 = Oskarshamn, 28 = Kalmar, 29 = Karlskrona, 30 = Halmstad, 31 = Helsingborg, 32 = Hässleholm, 33 = Trelleborg.

should be pointed out that all the municipalities included in this study were relatively small in comparison (Banta-Green et al., 2009; Thomas et al., 2012). However, the highest load was found in the largest city included in this study. 3.2.3. Opioids Opioids are narcotic analgesics that predominately act on the μ-opiate receptor. Apart from analgesia, they also generate drowsiness and euphoria. Common side effects include nausea, dizziness, vomiting, fatigue, headache and constipation (Benyamin et al., 2008). Serious interactions can occur when opioids are mixed with cocaine, alcohol and other CNS depressants, e.g., benzodiazepines (European Monitoring Centre for Drugs and Drug Addiction: Drug profiles). Five of the eight opioids and one opioid metabolite included in this study were detected: tramadol, oxycodone, codeine, morphine and methadone (and its metabolite EDDP). All detected opioids have legal usage in Sweden. Tramadol is a synthetic opioid used to treat moderate to severe pain (Budd, 1999). Tramadol was found in 100% of the STPs in the range from 130 mg (1000 inhabitants)−1 day−1 in Bollebygd (STP 24) to 700 mg (1000 inhabitants)−1 day−1 in Örnsköldsvik (STP 8) (Fig. 5). The number of prosecutions associated with tramadol has increased in Sweden, whereas heroin prosecutions have decreased (Swedish National Council for Crime Prevention, 2010). Codeine is an opiate, i.e., naturally occurring alkaloid in opium. It is primarily used as an analgesic but is also used to treat coughs, diarrhea, anxiety etc. (Wu et al., 2013). Codeine was detected in all 33 STPs in the

range from 110 mg (1000 inhabitants)−1 day−1 in Bollebygd (STP 23) to 520 mg (1000 inhabitants)−1 day−1 in Gothenburg (STP 24) (Fig. 5). Morphine, the most common naturally occurring alkaloid in opium, is used to relieve severe pain (Andersson et al., 2003). Morphine is also a major metabolite of heroin, which might also contribute to the levels in some areas. Morphine was detected in all 33 STPs in the range from 50 mg (1000 inhabitants)−1 day−1 in Bollebygd (STP 23) to 350 mg (1000 inhabitants)− 1 day−1 in Piteå (STP 4) (Fig. 5). In Sweden, the suspected low use of heroin together with high use of legal morphine suggests that the morphine contribution from metabolized heroin is probably negligible in most cases (Khan and Nicell, 2011; Swedish National Board of Health and Welfare). Methadone is a synthetic opioid that is commonly used as an analgesic and also to treat opioid addiction (Garrido and Trocóniz, 1999). Methadone was detected in 88% of the STPs (29/33) and its metabolite EDDP was detected in 94% (31/33) (Fig. 6). Levels of methadone ranged from 0.5 mg (1000 inhabitants−1 day)−1 in Lycksele (STP 6) to 29 mg (1000 inhabitants)−1 day− 1 in Trelleborg (STP 33) and EDDP was found in the range from 0.5 mg (1000 inhabitants)−1 day−1 in Lycksele (STP 6) to 29.3 mg (1000 inhabitants)−1 day−1 in Trelleborg (STP 33). Oxycodone is a semi-synthetic opioid that is used to treat moderate to severe pain (Kalso, 2005). Oxycodone was detected in 3 out of 33 STPs: Umeå (STP 7), Hässleholm (STP 32) and Trelleborg (STP 33) in the range from 20 mg (1000 inhabitants)−1 day−1 to 260 mg (1000 inhabitants)−1 day−1 (Fig. 2). The amount in Trelleborg was high considering that levels were below the LOQ at almost all the other STPs. Further, the prescription statistics do not suggest that legal oxycodone

Fig. 3. Load (mg (1000 inhabitants)−1 day−1) of amphetamine, methamphetamine and methylphenidate in Swedish STPs. 1 = Gällivare, 2 = Haparanda, 3 = Luleå, 4 = Piteå, 5 = Skellefteå, 6 = Lycksele, 7 = Umeå, 8 = Örnsköldsvik, 9 = Östersund, 10 = Härnösand, 11 = Söderhamn, 12 = Mora, 13 = Borlänge, 14 = Gävle, 15 = Arvika, 16 = Karlstad, 17 = Örebro, 18 = Köping, 19 = Eskilstuna, 20 = Nyköping, 21 = Norrköping, 22 = Stenungsund, 23 = Gothenburg, 24 = Bollebygd, 25 = Borås, 26 = Visby, 27 = Oskarshamn, 28 = Kalmar, 29 = Karlskrona, 30 = Halmstad, 31 = Helsingborg, 32 = Hässleholm, 33 = Trelleborg.

868

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871

Fig. 4. Loads (mg (1000 inhabitants)−1 day−1) of cocaine and benzoylecgonine in Swedish STPs. 1 = Gällivare, 2 = Haparanda, 3 = Luleå, 4 = Piteå, 5 = Skellefteå, 6 = Lycksele, 7 = Umeå, 8 = Örnsköldsvik, 9 = Östersund, 10 = Härnösand, 11 = Söderhamn, 12 = Mora, 13 = Borlänge, 14 = Gävle, 15 = Arvika, 16 = Karlstad, 17 = Örebro, 18 = Köping, 19 = Eskilstuna, 20 = Nyköping, 21 = Norrköping, 22 = Stenungsund, 23 = Gothenburg, 24 = Bollebygd, 25 = Borås, 26 = Visby, 27 = Oskarshamn, 28 = Kalmar, 29 = Karlskrona, 30 = Halmstad, 31 = Helsingborg, 32 = Hässleholm, 33 = Trelleborg.

usage is specifically high in Trelleborg since prescription rates are reportedly higher in other parts of the country (Swedish National Board of Health and Welfare). Abuse of oxycodone is common in many other countries, especially the USA (Cicero et al., 2005). 3.2.4. Other compounds Zolpidem is a short-acting imidazopyridine used for the treatment of insomnia (Priest et al., 1997). Zolpidem was found in 88% of the STPs (29/33), with the highest load of 5.6 mg (1000 inhabitants)− 1 day − 1 found in Hässleholm (STP 32) (Fig. 6). Zolpidem was prescribed at approximately 12 Defined Daily Doses (DDDs) (1000 inhabitants)− 1 day− 1 in Sweden during 2012, which is more than any of the benzodiazepines (Swedish National Board of Health and Welfare). 3.3. Correlation between geography, population and drug consumption Multivariate data analysis was performed in order to investigate whether any patterns in illicit drug consumption related to the number of people connected to and/or geographical location of the STP could be detected. All model diagnostics are presented in the Supplementary Information, Table S13. The multivariate data analysis was based on 33 observations (STPs) and 13 variables (illicit drugs) since the 16 remaining drugs were below LOQ in all STPs. For each variable, detected concentration below LOQ was set to 0 in the model. Note that interpretation of the

models should be made with caution since only one sample was taken at each location. Possible correlations between drug consumption and people connected to the STPs were investigated using an OPLS model, treating the population as a quantitative y variable. No clear separations based on the number of people connected to the STPs could be seen (data not shown), but Gothenburg (STP 23), being the only major city in the study, was clearly separated from the others. Increased usage of cocaine in more populated cities within the same country has been shown in previous studies (Banta-Green et al., 2009; Thomas et al., 2012). Separation based on geographic location was investigated by first dividing the STPs into six classes according to geographical location. The sample classes were as follows: northeast coast (Söderhamn, Härnösand, Örnsköldsvik, Umeå, Skellefteå, Piteå, Luleå, Haparanda), northern inland (Gällivare, Lycksele, Östersund, Mora, Borlänge), middle Sweden west (Borås, Arvika, Bollebygd, Karlstad, Örebro), middle Sweden east (Köping, Norrköping, Nyköping, Eskilstuna), west coast (Trelleborg, Helsingborg, Hässleholm, Halmstad, Stenungsund, Gothenburg) and southeast coast (Oskarshamn, Kalmar, Karlskrona, Visby). The OPLS-DA model revealed a trend in geographical location of the STPs with grouping of the classes, but no clear separation was obtained (Fig. S1 in the Supplementary Information). However, a two-class OPLS-DA model based on data for the northeast and west coasts, i.e., two geographical areas distinctly distant from each other, showed significant separation (p = 0.017) (Fig. 7A). This

Fig. 5. Load (mg (1000 inhabitants)−1 day−1) of codeine, morphine and tramadol in Swedish STPs. 1 = Gällivare, 2 = Haparanda, 3 = Luleå, 4 = Piteå, 5 = Skellefteå, 6 = Lycksele, 7 = Umeå, 8 = Örnsköldsvik, 9 = Östersund, 10 = Härnösand, 11 = Söderhamn, 12 = Mora, 13 = Borlänge, 14 = Gävle, 15 = Arvika, 16 = Karlstad, 17 = Örebro, 18 = Köping, 19 = Eskilstuna, 20 = Nyköping, 21 = Norrköping, 22 = Stenungsund, 23 = Gothenburg, 24 = Bollebygd, 25 = Borås, 26 = Visby, 27 = Oskarshamn, 28 = Kalmar, 29 = Karlskrona, 30 = Halmstad, 31 = Helsingborg, 32 = Hässleholm, 33 = Trelleborg.

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871

869

Fig. 6. Load (mg (1000 inhabitants)−1 day−1) of methadone, EDDP and zolpidem in Swedish STPs. 1 = Gällivare, 2 = Haparanda, 3 = Luleå, 4 = Piteå, 5 = Skellefteå, 6 = Lycksele, 7 = Umeå, 8 = Örnsköldsvik, 9 = Östersund, 10 = Härnösand, 11 = Söderhamn, 12 = Mora, 13 = Borlänge, 14 = Gävle, 15 = Arvika, 16 = Karlstad, 17 = Örebro, 18 = Köping, 19 = Eskilstuna, 20 = Nyköping, 21 = Norrköping, 22 = Stenungsund, 23 = Gothenburg, 24 = Bollebygd, 25 = Borås, 26 = Visby, 27 = Oskarshamn, 28 = Kalmar, 29 = Karlskrona, 30 = Halmstad, 31 = Helsingborg, 32 = Hässleholm, 33 = Trelleborg.

separation was attributed to the high usage of methamphetamine and methylphenidate on the northeast coast and high usage of codeine, oxazepam and benzoylecgonine (metabolized from cocaine) on the west coast (Fig. 7B). This result is supported by the prescription statistics for pharmaceuticals available online from the National Board of Health and Welfare (accessed 08/26/2013). In 2012, more DDDs (1000 inhabitants)− 1 of methylphenidate were prescribed in the region in the north of Sweden compared to the region located on the west coast (Swedish National Board of Health and Welfare). Similarly, codeine and oxazepam have a higher prescription rate

per capita in the region located on the west coast compared to the north of Sweden (Swedish National Board of Health and Welfare). The finding of higher concentrations of benzoylecgonine on the west coast is logical because it has larger cities, which have previously been shown to correlate with higher cocaine usage (Thomas et al., 2012). Even though this study was based on single day measurements with large uncertainties in many aspects (see earlier discussion), the results from the OPLS-DA model were found to be in good agreement with previous findings. With a better dataset (more samples from each plant,

Fig. 7. OPLS-DA model of west coast against northern east coast. A) OPLS-DA score plot (predictive, t[1], vs. orthogonal, to[1], score vectors) showing a significant separation (p = 0.017) between the west coast and the northeast coast along the predictive component. West coast includes: Trelleborg, Helsingborg, Hässleholm, Halmstad, Stenungsund, Gothenburg. Northeast coast includes: Söderhamn, Härnösand, Örnsköldsvik, Umeå, Skellefteå, Piteå, Luleå, Haparanda. B) OPLS-DA loading plot (predictive, w*[1] vs. orthogonal, wo[1], loading weights) showing the illicit drug pattern responsible for the separation between west coast and northeast coast. It can be seen that the consumption of codeine and oxazepam seems to be higher in the west coast, while mainly not only methamphetamine, but also methylphenidate, has higher values in the northeast coast.

870

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871

sampled under more controlled conditions) this method might be a valuable complement for detecting consumption patterns of illicit drugs. 4. Conclusion This study provides a snapshot of the use of narcotic substances in Sweden on one day in January 2012 through sewage analysis in 33 STPs with the following key results: • An on-line SPE–LC–MS/MS method capable of determining 31 illicit drugs or pharmaceuticals classified as narcotics was developed. • 13 out of 31 illicit drugs or pharmaceuticals classified as narcotics were detected in one or more STP. • Some of the classic drugs (or metabolites) were not detected above their respective LOQ. • Multivariate data analysis was successfully used to investigate regional differences in drug consumption across Sweden. Conflict of interest All authors declare no conflict of interest. Acknowledgment Staff at the included sewage treatment plants are greatly acknowledged for performing the sampling. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2013.11.081. References Andersson G, Christrup L, Sjøgren P. Relationships among morphine metabolism, pain and side effects during long-term treatment: an update. J Pain Symptom Manage 2003;25:74–91. Baker DR, Kasprzyk-Hordern B. Spatial and temporal occurrence of pharmaceuticals and illicit drugs in aqueous environment and during waste water treatment: new developments. Sci Total Environ 2013;454–455:442–56. Baker DR, Kasprzyk-Hordern B. Multi-residue analysis of drugs of abuse in wastewater and surface water by solid-phase extraction and liquid chromatography-positive electrospray ionization tandem mass spectrometry. J Chromatogr A 2011a;1218: 1620–31. Baker DR, Kasprzyk-Hordern B. Multi-residue determination of illicit drugs and pharmaceuticals to wastewater suspended particulate matter using pressurized liquid extraction, solid phase extraction and liquid chromatography coupled with tandem mass spectrometry. J Chromatogr A 2011b;1218:7901–13. Banta-Green CJ, Field JA, Chiaia AC, Sudakin DL, Power L, de Montigny L. The spatial epidemiology of cocaine, methamphetamine and 3,4-methylenedioxymethamphetamine (MDMA) use: a demonstration using a population measure of community drug load derived from municipal wastewater. Addiction 2009;104:1874–80. Benyamin R, Trescot AM, Datta S, Buenaventura R, Adlaka R, Sehgal N, et al. Opioid complications and side effects. Pain Physician 2008;11:105–20. Bijlsma L, Emke E, Hernandez F, de Vogt P. Investigation of drugs of abuse and relevant metabolites in Dutch sewage water by liquid chromatography coupled to high resolution mass spectrometry. Chemosphere 2012;89:1399–406. Budd K. The role of tramadol in acute pain management. Acute Pain 1999;2:189–96. Bylesjö M, Rantalainen M, Cloarec O, Nicholson JK, Holmes E, Trygg J. OPLS discriminant analysis: combining the strengths of PLS-DA and SIMCA classification. J Chemometr 2006;20:341–51. Castiglioni S, Bijlsma L, Covaci A, Emke E, Hernandez F, Reid M, et al. Evaluation of uncertainties associated with the determination of community drug use through measurement of sewage drug biomarkers. Environ Sci Technol 2013;47: 1452–60. Chiaia AC, Banta-Green C, Field J. Eliminating solid phase extraction with large volume injection LC/MSMS: analysis of illicit and legal drugs and human urine indicators in US wastewater. Environ Sci Technol 2008;42:8841–8. Cicero TJ, Inciardi JA, Muñoz A. Trends in abuse of OxyContin® and other opioid analgesics in the United States: 2002–2004. J Pain 2005;6:662–72. Commission Decision 2002/657/EC. concerning the performance of analytical methods and the interpretation of results. Off J Eur Communities 2002:L221. [17.8.2002]. Daughton CG. Illicit drugs in municipal sewage: proposed new non-intrusive tool to heighten public awareness of societal use of illicit/abused drugs and their potential for ecological consequences. In: Daughton CG, Jones-Lepp T, editors. Pharmaceuticals

and personal care products in the environment: scientific and regulatory issues. Washington: American Chemical Society; 2001. p. 348–64. de Voogt P, Bijlsma L, Emke E, van der Aa M. Illicit drugs in surface waters, drinking water and urban wastewaters. Abstracts of papers, 243rd ACS National Meeting & Exposition, San Diego, CA, United States; 2012. [March 25–29, ENVR-43]. Eriksson L, Trygg J, Wold S. CV-ANOVA for significance testing of PLS and OPLS models. J Chemometr 2008;22:594–600. European Monitoring Centre for Drugs and Drug Addiction. Handbook for surveys on drug use among the general population. EMCDDA project CT.99.EP.08 B. Lisbon: EMCDDA; 2002. [August]. European Monitoring Centre for Drugs and Drug Addiction. Drug profiles benzodiazepine. Available: http://www.emcdda.europa.eu/publications/drug-profiles/benzodiazepine. [Accessed 2013-05-06]. European Monitoring Centre for Drugs and Drug Addiction. National report 2012: Sweden. Lisbon: National focal point; 2013a. European Monitoring Centre for Drugs and Drug Addiction. The state of the drug problem in the European Union and Norway. Annual report 2012. Luxembourg: Publications Office of the European Union; 2013b. Fontanals N, Marcé RM, Borrull F. On-line solid-phase extraction coupled to hydrophilic interaction chromatography–mass spectrometry for the determination of polar drugs. J Chromatogr A 2011;1218:5975–80. Fontanals N, Borrull F, Marce RM. On-line weak cationic mixed-mode solid-phase extraction coupled to liquid chromatography–mass spectrometry to determine illicit drugs at low concentration levels from environmental waters. J Chromatogr A 2013;1286: 16–21. Garrido MJ, Trocóniz IF. Methadone: a review of its pharmacokinetic/pharmacodynamics properties. J Pharmacol Toxicol Methods 1999;42:61–6. Ghassabian S, Moosavi SM, Valero YG, Shekar K, Fraser JF, Smith MT. High-throughput assay for simultaneous quantification of the plasma concentrations of morphine, fentanyl, midazolam and their major metabolites using automated SPE coupled to LC– MS/MS. J Chromatogr B 2012;03:126–33. Kalso E. Oxycodone. J Pain Symptom Manage 2005;29:47–56. Karolak S, Nefau T, Bailly E, Solgadi A, Levi Y. Estimation of illicit drugs consumption by wastewater analysis in Paris area (France). Forensic Sci Int 2010;200:153–60. Khan U, Nicell JA. Refined sewer epidemiology mass balances and their application to heroin, cocaine and ecstasy. Environ Int 2011;37:1236–52. Khan U, Nicell JA. Sewage epidemiology mass balances for assessing the illicit use of methamphetamine, amphetamine and tetrahydrocannabinol. Sci Total Environ 2012;421–422:144–62. Khan GA, Lindberg R, Grabic R, Fick J. The development and application of a system for simultaneously determining anti-infectives and nasal decongestants using on-line solid-phase extraction and liquid chromatography–tandem mass spectrometry. J Pharm Biomed Anal 2012;66:24–32. Lin AYC, Wang XH, Lin CF. Impact of wastewaters and hospital effluents on the occurrence of controlled substances in surface waters. Chemosphere 2010;81: 562–70. Lopez-Serna R, Perez S, Ginebreda A, Petrovic M, Barcelo D. Fully automated determination of 74 pharmaceuticals in environmental and waste waters by online solid phase extraction-liquid chromatography-electrospray–tandem mass spectrometry. Talanta 2010;83:410–24. Ono S, Hatanaka T, Miyazawa S, Tsutsui M, Aoyama T, Gonzalez FJ, et al. Human liver microsomal diazepam metabolism using cDNA-expressed cytochrome P450s: role of CYP2B6, 2C19 and the 3A subfamily. Xenobiotica 1996;26: 1155–66. Postigo C, Lopez de Alda MJ, Barceló D. Fully automated determination in the low nanogram per liter level of different classes of drugs of abuse in sewage water by on-line solid-phase extraction-liquid chromatography-electrospray–tandem mass spectrometry. Anal Chem 2008;80:3123–34. Postigo C, de Alda MJ Lopez, Barceló D. Evaluation of drugs of abuse use and trends in a prison through wastewater analysis. Environ Int 2011;37:49–55. Priest RG, Terzano MG, Parrino L, Boyer P. Efficacy of zolpidem in insomnia. Eur Psychiatry 1997;12:5–14. Reid MJ, Langford KH, Mørland J, Thomas KV. Quantitative assessment of time dependent drug-use trends by the analysis of drugs and related metabolites in raw sewage. Drug Alcohol Depend 2011;119:179–86. Saussereau E, Lacroix C, Gaulier JM, Goulle JP. On-line liquid chromatography/tandem mass spectrometry simultaneous determination of opiates, cocainics and amphetamines in dried blood spots. J Chromatogr B 2012;885–886:1–7. Segura PA, Gagnon C, Sauve S. A fully automated on-line preconcentration and liquid chromatography–tandem mass spectrometry method for the analysis of anti-infectives in wastewaters. Anal Chim Acta 2007;604:147–57. Se ilir A, Schrier L, Bijleveld YA, Toersche JH, Jorjani S, Burggraaf J, et al. Determination of methylphenidate in plasma and saliva by liquid chromatography/tandem mass spectrometry. J Chromatogr B 2013;923–924:22–8. Statistics of Seizures from the Swedish customs. Available in Swedish:http://www. tullverket.se/pressrum/beslagsstatistik.4.4ab1598c11632f3ba9280001976.html. [Accessed at: 2013-05-16]. Stone M. Cross-validatory choice and assessment of statistical prediction. J R Stat Soc 1974;36B:111–33. Swedish National Board of Health and Welfare. Prescription statistics for pharmaceuticals in Sweden. Available on:http://www.socialstyrelsen.se/statistikdatabas/lakemedel. [Accessed 08/26/2013]. Swedish National Council for Crime Prevention. Narcotic Drugs Statistics 2009; 201016 [Report]. Terzic S, Senta I, Ahel M. Illicit drugs in wastewater of the city Zagreb (Croatia) — estimation of drug abuse in a transition country. Environ Pollut 2010;158:2686–93.

M. Östman et al. / Science of the Total Environment 472 (2014) 862–871 Thomas KV, Bijlsma L, Castiglioni S, Covaci A, Emke E, Grabic R, et al. Comparing illicit drug use in 19 European cities through sewage analysis. Sci Total Environ 2012;432:432–9. Trygg J, Wold S. Orthogonal projections to latent structures (OPLS). J Chemometr 2002;16:119–28. van der Aa M, Bijlsma L, Dijkman E, van Nuijs ALN, van de Ven B, Hernandez F, et al. Risk assessment for drugs of abuse in the Dutch watercycle. Water Res 2013;47:1848–57. van Nuijs ALN, Pecceu B, Theunis L, Dubois N, Charlier C, Jorens PG, et al. Cocaine and metabolites in waste and surface water across Belgium. Environ Pollut 2009a;157:123–9. van Nuijs ALN, Pecceu B, Theunis L, Dubois N, Charlier C, Jorens PG, et al. Spatial and temporal variations in the occurrence of cocaine and benzoylecgonine in waste- and surface water from Belgium and removal during wastewater treatment. Water Res 2009b;43: 1341–9. van Nuijs ALN, Castiglioni S, Tarcomnicu I, Postigo C, de Alda ML, Neels H, et al. Illicit drug consumption estimations derived from wastewater analysis: a critical review. Sci Total Environ 2011a;409:3564–77.

871

van Nuijs ALN, Mougel JF, Tarcomnicu I, Bervoets L, Blust R, Jorens PG, et al. Sewage epidemiology — a real-time approach to estimate the consumption of illicit drugs in Brussels, Belgium. Environ Int 2011b;37:612–21. Vazques-Roig P, Blasco C, Picó Y. Advances in the analysis of legal and illegal drugs in the aquatic environment. Trends Anal Chem 2013;50:65–77. Wise RA. Addictive drugs and brain stimulation reward. Annu Rev Neurosci 1996;19: 319–40. Wold S, Esbensen K, Geladi P. Principal component analysis. Chemom Intell Lab Syst 1987;2:37–52. Wu X, Zhang W, Bai Y, Guo T, Gu J. Simultaneous analysis of codeine and its active metabolites in human plasma using liquid chromatography–tandem mass spectrometry: application to a pharmacokinetic study after oral administration of codeine. J Pharm Biomed Anal 2013;78–79:261–8. Zuccato E, Chiabrando C, Castiglioni S, Calamari D, Bagnati R, Schiarea S, et al. Cocaine in surface waters: a new evidence-based tool to monitor community drug abuse. Environ Health 2005;4:14–20.