A snapshot of illicit drug use in Sweden acquired through sewage water analysis -Ostman et al (2014)

Introduction

  • Monitoring illicit drug use is crucial for improving public health.
  • Traditional methods include population surveys, medical records, and seizure data.
  • Daughton (2001) proposed back-calculating illicit drug use from sewage water levels.
  • Zuccato et al. (2005) confirmed the feasibility, leading to numerous studies.
  • Sewage water analysis offers a faster, less labor-intensive complement to traditional methods.
  • Back calculation from sewage to community usage is still evolving.
  • Castiglioni et al. (2013) investigated the entire process, highlighting uncertainties.
  • Khan and Nicell (2011, 2012) studied mass balances of cocaine, heroin, ecstasy, methamphetamine, amphetamine, and THC.
  • Excretion patterns of major metabolites are essential for accurate back calculations.
  • Knowing which metabolite and its excretion extent is crucial.
  • Population size significantly impacts results, but methods like biological oxygen demand (BOD) and chemical oxygen demand (COD) can reflect population fluxes.
  • Differentiating industrial vs. domestic loads poses a challenge.
  • Adsorption of drugs to particulate matter can influence results.
  • Baker and Kasprzyk-Hordern (2011a) found minimal adsorption (less than 10%) for most drugs, except methadone and EDDP.
  • Monitoring studies are often labor-intensive due to manual pre-treatment steps.
  • Semi-automated on-line SPE techniques can minimize these issues.
  • On-line methods are generally faster, more precise, and efficient.
  • They eliminate evaporation and reconstitution steps, requiring less sample handling and solvent use.
  • On-line SPE has lower pre-concentration factors, leading to higher limits of quantification (LOQs).
  • However, it has been successfully used for various analytes in biological and environmental samples.
  • Previous studies have used on-line methods for illicit drugs but not extensively for incoming wastewater.
  • Postigo et al. (2008) and Fontanals et al. (2013) are exceptions, employing on-line methods for sewage water analysis of illicit drugs.
  • Drug use varies with location, as shown by previous sewage water analyses (Thomas et al., 2012).
  • Multivariate data analysis can provide an overview of illicit drug consumption patterns.
  • Aims of this study:
    • Develop an on-line SPE method for measuring 24 illicit and prescription narcotic substances and 7 metabolites in sewage water.
    • Acquire a snapshot of illicit drug usage in different Swedish municipalities/cities.
    • Investigate the dependence of drug use on city size and/or geographical location.

Materials and methods

  • Selection of illicit drugs was based on:
    • Information on established illicit drugs in Europe (European Monitoring Centre for Drugs and Drug Addiction, 2013b).
    • Previous findings from sewage water studies.
    • Plausible emerging illicit drugs from open access web forums.
  • Chemicals and reagents:
    • Milli-Q water (resistivity 18.2MΩcm−118.2 MΩ cm^{-1}) was used.
    • Methanol (Liqrosolve, Hypergrade) from Merck.
    • Formic acid (puriss. p.a.) from Fluka.
    • Various standards were purchased from Cerilliant, Sigma Aldrich, National Measurement Institute (Australia), and the Council of European Pharmacopoeia.
    • Isotopically labeled standards were purchased from Cerilliant and Cambridge Isotope Laboratories.
    • 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 supplier's instructions.
  • Sampling and pretreatment:
    • 33 STPs were selected for wide geographical distribution across Sweden; most served >10,000 people.
    • 24-hour flow proportional composite samples of incoming sewage water were collected by plant staff in January 2012.
    • Target collection date was January 17, 2012 (Tuesday), with some exceptions.
    • Only one sample per STP was collected due to practical constraints.
    • Samples were collected on a weekday to avoid weekend variations.
    • Samples were collected in 250 mL HDPE bottles and frozen for transport to the lab.
    • Prior to analysis, samples were thawed, syringe-filtered (0.45 μm), spiked with internal standards (ISs) to a final concentration of 500ngL−1500 ng L^{-1}, and acidified to pH 3 with formic acid (FA).
  • Analytical instrumentation:
    • Thermo TSQ Quantum Ultra mass spectrometer was used.
    • PAL HTC auto sampler (CTC Analytics AG) injected samples into a 1 mL stainless steel loop.
    • Surveyor LC pump (Thermo Fisher Scientific) loaded samples from the loop onto an on-line SPE Oasis HLB column (2.1 × 20, 15 μm; Waters, Ireland) for enrichment with water.
    • After 1.5 min, the valve was switched, and the Accela pump (Thermo Fisher Scientific) eluted compounds from the Oasis column using a methanol gradient.
    • Analytes were separated on a Thermo Hypersil GoldAQ analytical column (50 × 2.1 mm, 5 μm + guard column) using the Accela pump gradient program.
    • The flow passed through a PEEK capillary and was ionized by electrospray ionization (ESI) before mass spectrometer entry.
    • Mass spectrometer parameters were optimized semi-automatically.
    • ESI probe temperature was 325 °C; sheath gas flow was 60, auxiliary flow 25 (arbitrary units); source voltage was 3.0 kV.
    • Individual collision energies and tube lens values are shown in Table S3.
    • Cycle time was 1.0 s; Q1 detection width (FWHM) was 0.70; collision gas pressure was 1.5 mTorr.
  • QA/QC
    • Thirteen isotopically labeled ISs were used for quantification, matched based on analyte structure and retention time (Table 2).
    • Positive identification was based on two transitions (precursor ion and two product ions), with ratio deviation limited to +/−30% from the calibration standard.
    • Retention times had to be within +/−2.5% of the calibration standard, yielding four identification points.
    • The limit of quantification (LOQ) was determined from standard curves based on repeated measurements of low-level spiked water (MilliQ and wastewater); signal/noise ratio N10 defined the LOQ.
    • A seven-point calibration curve (1–1000 ng L−1) was used, prepared in Milli-Q water and spiked with ISs.
    • Recovery tests investigated the effect of different syringe filters.
    • Six syringe filters were tested: 0.45 μm Filtropur S, 0.20 μm Filtropur S, 0.45 μm MF Millex, 0.22 μm MF Millex, 0.45 μm PTFE-membrane, and 0.20 μm PTFE-membrane.
    • Milli-Q water was spiked with all compounds, filtered in triplicate with each filter, spiked with IS to 300ngL−1300 ng L^{-1}, and compared to unfiltered water.
    • A bench-top stability test assessed analyte stability under varying conditions (matrix and temperature).
    • Purified and filtered (0.45 μm Filtropur S) sewage water were spiked with analytes to 900ngL−1900 ng L^{-1} and analyzed in triplicate.
    • Samples were spiked with IS, acidified with FA to 0.1%, and stored at room temperature and 3 °C for 24 h.
    • Matrix effects were evaluated by comparing signal areas of standard solutions at 900ngL−1900 ng L^{-1} in sewage and Milli-Q water.
    • Carryover effects were assessed by injecting standards at 1200ngL−11200 ng L^{-1} followed by two mobile phase blanks.
    • Linearity was evaluated using a seven-point calibration curve (1–1000 ng L−1 ).
  • Impact of geography and city size by multivariate data analysis
    • Multivariate data analysis was performed to investigate the correlation between the STP catchment area or geographic location and the use of illicit drugs.
    • SIMCA 13.0 (Umetrics, Umeå, Sweden) was used for multivariate data analysis. All data were mean-centered and scaled to unit variance prior to modeling.
    • Principal component analysis (PCA) was initially performed to obtain an overview of the data.
    • Orthogonal partial least squares discriminant analysis (OPLS-DA) was carried out by grouping the samples into classes based on geographical location.
    • Internal model validation was performed with full (seven-fold) cross-validation.
    • For the two-class OPLS-DA models, cross-validated ANOVA p-values were calculated.

Results and discussion

  • QA/QC results
    • The Filtropur S filter (0.45 μm) had the highest average recovery (95%) and was therefore used in subsequent experiments.
    • 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.
    • 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%).
      *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.
    • 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.
    • Some carryover effects (N0.1%) could be observed for all analytes, except codeine, tramadol, and oxycodone, confirming the need to always perform a mobile phase blank run after high standards.
    • The linearity was good for all compounds; the correlation coefficient (R2R^2) 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%).
    • Based on the results of the filter test, intra-day variation, and the LOQ, TCH-COOH was excluded from the study.
  • Main study results and discussion
    • The study provides a snapshot of illicit drug usage in Sweden because only one sample was taken for each sample location.
    • The use of illicit drugs varies over time (both short term and long term).
    • Measurements conducted in Gothenburg and Umeå showed a relative standard deviation (RSD) of around 40% for most illicit drugs.
    • Castiglioni et al. (2013) estimated the uncertainties of the following: population size estimation, stability of drug biomarkers, sampling, chemical analysis, and back calculation.
    • Staff at the STPs should be consulted when making population estimations to minimize the uncertainty.
    • 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).
    • Complete back calculations (including corrections for metabolism and degradation) were not performed as part of the main results.
    • The results for all compounds were normalized against the flow in the STP and the number of inhabitants connected and expressed as mg (1000 inhabitants)^{-1}$ day$^{-1}$.
    • Complete back calculations for cocaine, amphetamine, and methamphetamine, as well as all the measured concentrations of all compounds included in the study, are provided in the Supplementary Information.
    • 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 (+ 6-acetyl-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.
    • 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.
    • LSD and MDMA are known to be less common in Sweden compared to many of the other illicit drugs.
    • As sampling was performed on a Tuesday, drugs primarily associated with weekend usage (like MDMA) were less likely to be detected.
    • Benzodiazepines:
      • Oxazepam was detected in all STPs (33/33) in the range from 20 to 200 mg (1000 inhabitants)^{-1}$ day$^{-1}$.
      • 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.
    • Stimulants:
      • Amphetamine was detected in 39% of the STPs (13/33) in the range from 10 to 140 mg (1000 inhabitants)^{-1}$ day$^{-1}$.
      • The highest loads were detected at the sampling locations Söderhamn (STP 11) and Gothenburg (STP 23).
      • Methamphetamine was detected in 48% of the STPs (16/33) in the range from 1 to 32 mg (1000 inhabitants)^{-1}$ day$^{-1}$.
      • Methylphenidate was detected in 94% of the STPs (31/33) in the range from 1 to 25 mg (1000 inhabitants)^{-1}$ day$^{-1}$.
      • 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$^{-1}$.
    • Opioids:
      • 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).
      • 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).
      • 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).
      • Methadone was detected in 88% of the STPs (29/33), and its metabolite EDDP was detected in 94% (31/33).
      • 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$^{-1}$.
    • Other compounds:
      • 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).
  • Correlation between geography, population, and drug consumption
    • Multivariate data analysis was performed 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.
    • Possible correlations between drug consumption and people connected to the STPs were investigated using an OPLS model.
    • Separation based on geographic location was investigated by first dividing the STPs into six classes according to their geographical location.
    • The OPLS-DA model revealed a trend in the geographical location of the STPs with the grouping of the classes, but no clear separation was obtained.
    • 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).
    • This 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.

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.
  • 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.