Sediments accumulated in sewer settling basins as recorders of historical use of drugs: Potential and limitations Open Access

The occurrence of drugs (i.e. pharmaceutically active compounds and illicit drugs) in numerous environmental compartments raises serious concerns about their potential hazard for living beings (Richmond et al. 2018). A significant part of the current literature focuses on both assessing the removal of organic contaminants in classical wastewater treatment plants (Wilkinson et al. 2017) and optimizing their removal via tertiary treatments (Altmann et al. 2015).

However, wastewater treatment plant influents (WWIs) could also represent a tremendous source of information because of the link between their chemical composition and the excretion of various chemical compounds by the population in the catchment (Castiglioni et al. 2014; Choi et al. 2018). Several studies have demonstrated the consistency of such analyses on illicit drugs and pharmaceutically active compounds to back-calculate the consumption of the population (Baker et al. 2014). This new scientific discipline has been called wastewater-based epidemiology and currently investigates not only the chemicals consumed as pharmaceuticals but also passive exposure to other markers, such as industrial chemicals and pyrethroid pesticides (Rousis et al. 2017; Lopardo et al. 2019), among many examples. Such studies provide near real-time dynamics of practices, whereas the assessment of past uses is not accessible.

However, several studies have highlighted the potential of sedimentary archives to record such historical use or historical environmental contamination through analyses of the vertical distributions of drugs and other contaminants (Klaminder et al. 2015; Thiebault et al. 2021a). Among these latter studies, few described the distribution of contaminants within silt traps (Flanagan et al. 2021; Wei et al. 2023), directly in the sewer system, as sampled sediments are generally coarse and therefore considered unfavourable for the trapping of organic molecules owing to limited reactivity (Crabtree 1989). One of the major advantages of settling basins, considering consumption/excretion amount assessment, is their position upstream in the sewer network and wastewater treatment plant and close to excretion areas (Karlsson et al. 2010; Spahr et al. 2020). As a result, settling basin sediments are potentially able to record a wider variety of contaminants than are riverine sediments because of the significant degradation or removal of some contaminants during wastewater treatment (Castiglioni et al. 2006; Behera et al. 2011).

A common prerequisite to assess the historical distribution of organic molecules within sedimentary cores is also to be able to build a robust age-depth model that establishes a relationship between depth and age. Whereas ¹³⁷Cs and ²¹⁰Pb are classically used to date the sediments of the 20th century (Kirchner & Ehlers 1998), very recent sediments retrieved in settling basins necessitate new dating processes (Kaizer et al. 2020).

In this study, we analysed the occurrence of drugs in a core sampled from a settling basin with the following objectives: (i) investigate the occurrence and vertical distribution of drugs in this record; (ii) date such archives for which curation operations were not recorded and propose an age-depth model highlighting the first detection of drugs; and (iii) evaluate the potential of such archives to perform sediment-based epidemiology, considering that settling basins are close to excretion sources, before the wastewater treatment plant removal effect, and therefore potentially record a wide diversity of drugs.

In 2014, a renovation of the CSA was performed by local authorities to facilitate cleaning operations and modify flow management to limit the direct discharge of wastewater into the Loire River during intense precipitation events. This represented a unique opportunity to perform the coring of this settling basin that has presumably never been cleaned since its building. The coring occurred in February 2015, prior to the major cleaning that removed ∼7 m of sediment in May 2015 (Figure 1).

Among the several cores sampled in this settling basin (Thiebault et al. 2021b), core CSA-02-2015-A is 2.54 m long and was collected with a UWITEC gravity corer, whereas the coring potential was expected to be 14 m. The top sediment was recorded 33.5 cm from the top of the coring tube, which corresponds to an altitude of −2.9425 m from the technical platform (Figure 1). The depths given below are homogeneously relative to the top of the tube.

After drying, granulometry analysis of all the samples (n = 72) was conducted by sieving 20 g of sediment through four screens of different square mesh sizes (>1 mm; >200 μm; >50 μm and <50 μm). The rejects were collected and weighted.

The 50–200 μm of fraction was used to determine the total organic carbon (TOC) (in %) via Rock-Eval6 (Vinci Technologies, Rueil-Malmaison, France) pyrolysis (Behar et al. 2001).

Standards for acetaminophen (ACM), atenolol (ATE), bezafibrate (BZB), carbamazepine (CBZ), codeine (COD), diazepam (DIA), diclofenac (DCF), ibuprofen (IBP), ketoprofen (KET), metoprolol (MET), oxazepam (OXA), salicylic acid (SCA), sulfamethoxazole (SMX), tramadol (TRA), and trimethoprim (TMP) were purchased from Sigma‒Aldrich (Saint Quentin-Fallavier, France) with purities >98%. The standards for benzoylecgonine (BZE), cocaine (COC), 3,4-methylene-dioxy-N-methylamphetamine (MDA), methadone (MED) and morphine (MOR) were purchased from LGC Standards (Wesel, Germany). Further information about the selected contaminants and their physicochemical properties is given in Table 1. The extraction and separation solvents dichloromethane (DCM), methanol (MeOH) and acetonitrile (AcN) were purchased from Thermo Fisher-Scientific (Illkirch-Graffenstaden, France) and were of analytical grade (purity up to 99.95%).

One gram of the 50–200 μm of granulometric fraction was extracted via pressurized liquid extraction via an accelerated solvent extractor (ASE-200, Dionex, Sunnyvale, CA, USA). The extraction mixture was MeOH/H₂O (1:1 v/v); the operating temperature and pressure were 100 °C and 6,895 kPa, respectively. The extracts were then fully dried at 60 °C under nitrogen flow and stored at −18 °C. The extracts were recovered in ultrapure water acidified with 0.1% formic acid before liquid chromatography analysis. Standards and controls were prepared with surface waters sampled from the Loiret River (filtered at 0.22 μm), which, after analysis, were found to be deprived of drugs.

Drug separation was achieved at 30 °C and a flow rate of 0.4 mL min⁻¹ with a Nucleodur C₁₈ Gravity column (150 mm × 2 mm × 1.8 μm, Macherey-Nagel, Hoerdt, France) supplemented by a guard column using an Ultimate 3,000 RSLC (Thermo Fisher-Scientific, San Jose, CA, USA) ultrahigh-performance liquid chromatography system equipped with a binary pump. The injection volume was 3 μL. Ultrapure water (A) and AcN (B), acidified with 0.1% formic acid, were used as the mobile phase. The elution gradient was a transition from 95 to 5% A in 16.2 min followed by 0.3 min of 100% B and then a return to the initial conditions (95% A) for 3.35 min for a total analysis time of 19.85 min. The chromatography system was coupled to a TSQ Endura triple quadrupole mass spectrometer equipped with a heated electrospray ionization (H-ESI) interface (Thermo Fisher-Scientific, San Jose, CA, USA) at a flow rate of 0.3 mL min⁻¹.

An electrospray ionization source was used for quantification, operating in positive mode (except for SCA), with a vapourizer temperature of 425 °C, an ion transfer temperature of 325 °C, an electrospray voltage of 3,600 V, an auxiliary gas of 20 Arb, a sheath gas of 50 Arb, and a sweep gas of 1 Arb. Xcalibur 2.2 software was used to evaluate the qualitative and quantitative analysis of the selected drugs.

The following procedure was systematically used for the analysis validation: a calibration curve (six standards from 0.5 to 100 ng L⁻¹), followed three times by four samples, one quantification control (a standard at 5 ng g⁻¹) and one blank (pure water), and finally, another calibration curve. This procedure allowed the assessment of the drift of the analysis between each calibration curve, as described in other studies (Thiebault et al. 2017b). According to classical organic analysis, the limits of quantification (LOQ) and limits of detection (LOD) were calculated with signal-to-noise ratios of 3 and 10, respectively. The linearity considers the three calibration curves that frame the sample analysis and was determined by the linear correlation coefficient r². Recovery values were calculated by first extracting three randomly selected sampled sediments with a DCM/MeOH (9:1, v/v) mixture to avoid all the drugs (Thiebault et al. 2021b). Then, a concentration of 100 ng g⁻¹ for each drug was spiked onto these sediments prior to extraction, which was carried out following the same procedure (MeOH/H₂O, 1:1, v/v) and compared to the standard injection value.

Nineteen ¹⁴C datings (Table 2) were obtained from the LSCE (Laboratoire des Sciences du Climat et de l'Environnement;GifA/ECHo lab identification) and Poznań Radiocarbon Laboratory (Poz lab identification). The organic materials visible in the core were sampled, identified and selected for ¹⁴C dating according to their nature. Short-lived materials were selected, giving preference to seeds. The idea is to integrate the shortest possible period of time and thus achieve a fine dating in time. They have been sent for dating either to the LSCE for the smaller ones, or to Poznan for the larger ones. The LSCE operates a MICADAS that can run samples down to a few tens of μg of carbon (Thil et al. 2024). In the laboratory, samples are chemically treated to remove potential contaminants from the sand chambers (e.g. Hatté et al. 2023). Those for which several hundred μg of carbon were accessible were transformed into CO₂ and reduced in the form of C graphite. At LSCE, these operations are combined and carried out using an automatic system, AGE3; at Poznan, the two operations are carried out separately. Measurements are carried out using the ECHoMICADAS solid source or the Poznan AMS. The smallest samples (after chemistry) were introduced as CO₂ into the ECHo gas source via a gas interface system connected to an elemental analyzer (EA-GIS).

In both laboratories, normalization standards, blanks and reference materials of equivalent matrix and surrounding age are measured at the same time as the samples. These are used to standardize results and check the reliability of all sample processing steps. In line with community recommendations, results are reported as F¹⁴C (fraction of modern carbon). Throughout the manuscript, the core depths correspond to the sediment-depth from the core top (see Figure 2), with the top sediment being ∼ 2.943 m below the platform (Figure 1(c)). The depth error corresponds to the thickness of the layer in which the material was found. If zero, it corresponds to the exact depth of the sample.

F¹⁴C data were computed via Bayesian chronological modelling with ChronoModel (Lanos & Dufresne 2024) 3.2.7 software. Since the CSA was built after 1945, F¹⁴C data have been calibrated by restricting the range of possible ages to the period post 1945. The bomb13 calibration curve for the Northern Hemisphere zone 1 was used (Hua et al. 2013), considering that the grape whose seeds were dated came from this geographical area. This curve was extended for 2010–2015 with data from Graven et al. (2017). The 19 F¹⁴C data and their depths were considered stratigraphic successions of events incorporated into a single phase that terminated on top of the sedimentary sequence (Supplementary material, Table S2), dated from February 2015 (coring date). The study period was considered to be from 1947 to 2016. The Markov chain Monte-Carlo settings were as follows: three chains, 1,000 iterations during the burn-in phase, 200 iterations during the adaptation phase, and 1,000,000 iterations during the acquisition phase.

The age-depth model was then used in the ChronoModel (‘ChronoCurve’ tool) as a calibration curve for determining the date of deposition of the 72 samples, taking into account their mean depth and thickness for depth and depth error (Supplementary material, Table S3). The same approach and calibration curve were applied to evaluate the date of appearance of specific molecules (TRA, MED, MOR, paracetamol, and ATE). The depth at which the first occurrence of a molecule occurred was considered the mean depth between the deepest sample with > LOQ concentration followed by quantification and the sample below. The depth error was half the difference between the two sample depths.

As expected in view of the role of the settling basin, the sediments were very coarse with very limited fine particle contents (i.e. <50 μm), which were mostly approximately 1% (with minimum and maximum values of 0.05 and 16.1%, respectively) all along the profile, with a main contribution of 0.2 < 1 mm fractions with a median of 74.8% (Figure 2).

Five distinct facies (F1, F3, F5, and less frequent F2 and F4) can be distinguished and are repeated across the core into 39 distinct and successive units (Figure 2, full description in Supplementary material, Table S1). F1 corresponds to brownish/black organic layers, F2 to ochre fine sands, F3 to beige fine sands, F4 corresponds to grey fine sands, and F5 corresponds to ochre coarse sands.

Glass and plastic particles, shells, brick pieces, and numerous seeds were found over the core.

The TOC content varied widely, with values ranging from 0.4 to 9.2%. Higher TOC contents were generally observed in specific layers (Figure 2), particularly the F1 facies.

According to this age-depth model, the oldest sediments in the core were deposited in approximately the 1950s, and the youngest were deposited in approximately 2009. More recent sediments (from 2009 to 2015) correspond to the rework interval (Figure 2). The age-depth model is consistent with the fact that this structure has not been cleaned since it was installed.

Among the 20 targeted drugs, 18 were quantified at least once within the CSA-02-2015-A core, with the highest concentrations of CBZ, ATE, and SCA. Stimulants were sparsely quantified in the sediment, with only one quantification and three additional detections for COC, whereas MDMA and BZE were not detected (Table 3). Except for SCA, with a quantification frequency of 70%, the quantification frequencies of the drugs varied from several percent to 37% for CBZ. Drug concentrations range from hundreds of pg g⁻¹ to hundreds of ng g⁻¹.

The calibrated dates of the first occurrence of the targeted drugs are given in Supplementary material, Table S4 by using the same process.

Depending on the drug, the first occurrence depth is very different, from 40.5 cm for MED and TRA to 2.575 m for SCA (Supplementary material, Table S4). Based on the above-mentioned age-depth model, this corresponds to first sedimentary occurrences between 1952 and 1998 with most first detections in the 1970s (Figure 5). It is worth mentioning that age-depth model precision decreases with depth, with 10 years of interval for the top core and 25 years interval in the bottom part of the core.

The 10-year time lag observed for certain molecules (e.g. DIA, ACM, SMX) can be due to their unfavourable physicochemical properties. Such molecules are besides rarely observed in sediments, whereas for less hydrophilic molecules (e.g. CBZ, TRA), first quantification is almost simultaneous with MAD as already observed in riverine sediments (Thiebault et al. 2017a; Kerrigan et al. 2018).

This shows that the detection of drugs in this type of sediment allows them to be dated at a minimum, and therefore reinforces the robustness of the age-depth model.

In line with studies that have investigated fluvial or lacustrine sedimentary archives, the molecules appear to be trapped on particles, settled in the sedimentary column, and therefore not mobile. In addition, no in-situ degradation in the changes in concentrations was observed, attesting to good preservation of the recorded signal. However, in contrast to environmental archives, the highly heterogeneous nature of sediment geochemistry makes it difficult to assess general trends in contamination.

Beyond the knowledge of the first uses detected in the study area, the potential of these archives for estimating historical consumption arises. Indeed, one of the prerequisites for carrying out sediment-based epidemiology is to be able to estimate a dissolved concentration (which is used in wastewater-based epidemiology) from the sediment content. As demonstrated in a previous study from the same site (Thiebault et al. 2021b), the main factor that controls the adsorption of dissolved molecules onto particles (which can then settle in sediments) is organic carbon (Al-Khazrajy & Boxall 2016), particularly for noncationic molecules. It is therefore possible to estimate historical dissolved concentrations from K_oc (organic carbon partition coefficient) values (Table 1). To compare these estimated concentrations with the actual dissolved concentrations, it is possible to use 85 consecutive days of monitoring of the same drugs in the WWI in the same area (Thiebault et al. 2017b).

A first parameter that can specifically impact the calculated concentration for each drug is the value of K_oc. Most published values are estimated or experimental data, but the conditions behind each result are not systematically relevant for the sampled system. It would therefore be preferable to adapt the K_oc value to current conditions in the settling basin to increase this potential. The second parameter, which will impact all the content homogeneously, is the characterization of the TOC. Such analyses were conducted on 50 < 200 μm fractions because <50 μm was very limited and coarser fractions were not adapted for such analyses. The hypothesis of a weak contribution of coarser fractions to the sorption of drugs was also made, resulting in this choice. If it is difficult to conclude here on the automatic use of sediment for the estimation of historical concentrations and consumption, one can reasonably be optimistic as the breakdowns of the concentration ranges are close between the estimated and measured values.

Once these estimates have been made, two key points still need to be resolved to be able to calculate historical usage: (i) To know more about the sources and dynamics of sediment deposition in this settling basin (e.g. event-driven or long-term?), and (ii) to determine the historical or approximate cumulative annual discharge.

This study reveals for the first time that coarse sediments from settling basins contain a wide diversity of drugs generated by excretion, although some of them were never detected in river sediments. This diversity of drugs makes it possible to estimate the consumption/excretion history once the sediments have been dated. In contrast to those from environmental archives, sediment cores from settling basins show contrasting accumulation patterns, with highly heterogeneous organic and mineral contents due to distinct sediment sources (i.e. stormwater and wastewater). This heterogeneity makes it difficult to use conventional age modelling, whereas an event-based model has been used here for the first time for this type of sediment. The good agreement between the first detection of each drug, what is known about their historical use, and the age model was a further clue to confirm the potential of these sediments to record the history of the city. Beyond this chronological assessment, a second objective of this study was to attempt to quantitatively estimate the historical amount of drug excreted. Using the solid/water partition coefficient equation, the back-calculated wastewater concentrations were found to be mostly lower than the actual concentrations, indicating that further work remains necessary to fully understand the solid–water partitioning of the drug and the parameters impacting the achievement of a quantitative assessment.

The authors gratefully acknowledge C. Morio for sampling authorization. The authors also wish to thank C2FN (L. Augustin, A. de Moya and L. Ardito) for their support during coring operations. E. Destandau and L. Fougère are thanked for providing access to HPLC-MS. Postgraduate students, specifically A. Thibault and M. Réty, are acknowledged for their participation in data acquisition. R. Boscardin is acknowledged for her long-standing technical support.

This work was supported by the CNRS/INSU/EC2CO-/BIOHEFECT Golden Spike French national program, the French National Agency for Research (ANR-21-CE03-0005) and the Labex VOLTAIRE (NR-10-LABX-100-01).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.