Occurrence and Human Health Risk Assessment of Pharmaceuticals and Hormones in Drinking Water Sources in the Metropolitan Area of Turin in Italy

Pharmaceuticals and hormones (PhACs) enter the aquatic environment in multiple ways, posing potential adverse effects on non-target organisms. They have been widely detected in drinking water sources, challenging water companies to reassure good quality drinking water. The aim of this study was to evaluate the concentration of sixteen PhACs in both raw and treated drinking water sources in the Metropolitan Area of Turin—where Società Metropolitana Acque Torino (SMAT) is the company in charge of the water cycle management—and evaluate the potential human health risks associated to these compounds. Multivariate spatial statistical analysis techniques were used in order to characterize the areas at higher risk of pollution, taking into account the already existing SMAT sampling points’ network. Health risks were assessed considering average detected concentrations and provisional guideline values for individual compounds as well as their combined mixture. As reported in the just-issued Drinking Water Directive 2020/2184/UE, in order to establish priority substances, a risk assessment of contaminants present in raw drinking water sources is required for monitoring, identifying potential health risks and, if necessary, managing their removal. The results showed negligibly low human health risks in both raw water sources and treated water.


Introduction
The majority of European countries rely on surface and groundwater sources for their drinking water needs. However, the distribution of safe drinking water is one of the most important-although intricate-issues nowadays since these sources can often be contaminated. Surface and groundwater quality may be affected by both natural and anthropogenic factors [1]. Metals, single organic ions, more complex organic molecules, and biological components can derive from various sources, such as natural disasters, rural run-off, industrial and sewage discharge, population, and economic growth [1,2]. Water utilities and the scientific community are called to find efficient contaminants' remediation systems in order to improve the performance of treatment plants and deliver safe drinking water to the population. The techniques usually in place include conventional methods such as precipitation, activated carbon adsorption, biological processes, and innovative methods such as advanced oxidation processes, membrane filtration using reverse osmosis, nano-and ultrafiltration processes, and biochar [3,4].
Pharmaceuticals and hormones (PhACs) represent one major category of anthropogenic contaminants present in the aquatic environment, degrading water quality [1]. In Europe, their use is continuously increasing, with 3000 compounds currently being active on the market [5]. Due to their large consumption, pharmaceuticals and hormones can reach the aquatic environment through different routes, including animal and human

Selection of Compounds
A priority list containing different pharmaceutical compounds and hormones was prepared based on the EU watch list, the just-issued European Drinking Water Directive (2020/2184/UE) [27], the Regional Environmental Protection Agency (ARPA Piemonte) analytical protocol and the NORMAN prioritization framework of emerging substances (Table A1). In this way, our conclusions led to sixteen different compounds, including antibiotics, beta-blockers, non-steroidal anti-inflammatory drugs, and hormones: Ketoprofen, Atenolol, Trimethoprim, Ofloxacin, Azithromycin, Ciprofloxacin, Cyclophosphamide, Toxics 2021, 9, 88 3 of 13 Sulfamethoxazole, Erythromycin, Clarithromycin, Diclofenac, Carbamazepine, Ibuprofen, Caffeine, Estrone, and 17-beta estradiol. Caffeine was included in this study as a tracer of anthropogenic pollution.

Study Area and Sampling
The present study focused on the occurrence of pharmaceuticals and hormones in the Metropolitan Area of Turin (Piemonte, Italy), where SMAT is in charge of water distribution to 2.3 million inhabitants, supervising 293 municipalities. Within the context of Green Analytical Chemistry and for avoiding the costs, efforts and environmental impact of chemical analyses at a large-scale blind monitoring assessment, the selection of the sampling points based on the prioritization of the sites at major risk was done according to a geographical model, built in a previous study [28]. Spatial and multivariate statistical analysis tools were used in order to predict potential pollution levels and classify "hotspot" areas for monitoring. In this case, 44 hospitals and care houses and 24 major WWTPs in the territory were taken into account as possible pollution sources. As a result, 270 out of the 683 already existing sampling points in the catchment area ( Figure 1)-used by SMAT for routine analyses-were found to be at highest risk and were selected for monitoring. Map of the study area and its geographical position in the Italian territory, including all the SMAT existing sampling points in the catchment areas, WWTPs, hospitals/care houses taken into account as potential pollution sources.

Sample Preparation and Analysis
For the extraction of the analytes (mainly acidic), the pH of water samples was adjusted to 2.0 with HCl after the addition of 500 mg of Na4EDTA to each of them. The 1 L Figure 1. Map of the study area and its geographical position in the Italian territory, including all the SMAT existing sampling points in the catchment areas, WWTPs, hospitals/care houses taken into account as potential pollution sources. The sampling/monitoring campaign lasted one year and was carried out between October 2019 and October 2020. In total, 328 samples were collected, according to the specifications and requirements of ISO 5667 accreditation [29], including groundwater, surface and drinking water. As surface water were considered the samples taken at the drinking water treatment plant's intake, as drinking those taken after the last step of the whole treatment line, fountains, and tanks, and as groundwater those taken from pumps at each wellhead. Amber glass bottles (1 L)-previously decontaminated and rinsed with MeOH, according to the EPA 1694 method [30]-were used for water collection. The samples were refrigerated throughout their transport (10-15 • C), stored at 4 • C prior to their analysis, and analyzed within seven days from their sampling.

Sample Preparation and Analysis
For the extraction of the analytes (mainly acidic), the pH of water samples was adjusted to 2.0 with HCl after the addition of 500 mg of Na 4 EDTA to each of them. The 1 L samples were loaded to Oasis-HLB (200 mg) solid-phase extraction (SPE) cartridges (Waters, Milford, MA, USA)-which were preconditioned with 12 mL MeOH followed by 6 mL MilliQ and 6 mL MilliQ with pH 2.0-with a flow rate of 10 mL/min. The analytes were extracted from the sorbent material with 12 mL MeOH and reconstituted to 1 mL MilliQ after evaporating the solvent with a rotary evaporator (BUCHI Rotavapor R-114). Chromatographic analyses were carried out using a triple quadrupole SCIEX QTRAP ® 6500 system (SCIEX, Framingham, MA, USA) connected to a Thermo Scientific Dionex UltiMate 3000 HPLC system equipped with a Kinetex ® C18 HPLC column (1.7 µm particle size, 100 mm × 2.1 mm; Phenomenex Inc., Torrance, CA, USA). The QTRAP system operated in both Positive and Negative Electrospray Ionization Mode (ESI) using Multiple Reaction Monitoring (MRM) scan mode. Considering the heterogeneity among the compounds, three subsequent methods were developed. For the Positive ESI substances (Table A2), a volume of 12 µL of the sample was injected at a mobile phase consisted of a mixture of 0.1% Formic Acid in MilliQ Water and Methanol, following a gradient profile in a total run time of 10 min. For the Negative ESI compounds (Table A2), a volume of 10 µL of sample was injected at a mobile phase consisted of a mixture of 0.02% Ammonia in MilliQ Water and Methanol, following a gradient profile with a total run time of 10 min, while for Ibuprofen a volume of 10 µL was injected into a mixture of 0.1% Formic Acid and 0.1% Ammonium Acetate in MilliQ Water and Methanol, following a gradient profile in a total run time of 10 min (Table A2).

Validation Study
For reassuring the developed method's applicability, a validation study was necessary and carried out according to ISO/IEC 17025 accreditation requirements [31]. Six-point calibration curves of final concentrations 1000, 2000, 4000, 6000 and 10,000 ng/L were built for each target compound and used for quantification taking into account the SPE preconcentration factor of 1000. For each point, fifteen replicates were analyzed and used for testing uncertainty, trueness, linearity, recovery and limits of Detection (LOD) and Quantification (LOQ). In addition, blank and quality control samples were analyzed to ensure the instrument's best performance during the analysis. The quality control samples had a final concentration of 4000 ng/L, and their analysis was processed after every ten samples. The quantitation was performed using the MultiQuant TM 3.0.3 software (SCIEX, Framingham, MA, USA).

Average Concentrations in Water
In order to avoid the wrong estimation of the average detected concentrations of the compounds, non-detects were considered at a value of 1 4 of the individual LOD of each target molecule as proposed by Houtman et al. [9]. This method was adopted since Toxics 2021, 9, 88 5 of 13 removing samples with non-detected compounds or setting their value as zero would have over or underestimated the average concentrations.

Human Health Risk Assessment
As this study focused on the determination of selected pharmaceuticals and hormones in surface and groundwater for drinking water production, a human health risk assessment was necessary and was done by comparing the pharmaceuticals' detected concentrations to guideline values. As a first step, we obtained the n-octanol-water partition coefficient (log Kow) for each compound using the KOWWIIN algorithm of the EPI Suite 4.11 software [32]. Compounds with log Kow > 3 were not included in the risk assessment study as there is a slighter possibility for them to pass through all the steps of the drinking water treatment line [33]. The Risk Quotient (RQ i ) (Equation (1)) for each compound was then calculated as the ratio between the Mean Detected Concentration (MEC i ) and the corresponding guideline value or, where it did not exist, the provisional guideline value ((p)GLV) [33]. The pGLVs were calculated using Equation (2), where ADI is the Acceptable Daily Intake (µg/kg bw/day); BW is the body weight set at a default value of 70 kg, as it is the closest to the average European bodyweight value of 70.8 kg [34]; DWI is the drinking water intake (L/day) set at a default value of 2 L/day as reported from WHO 2006 and a 10% of drinking water allocation factor was taken into account, as drinking water is not the only exposure way for humans [9,33,35]. The ADI values for the detected compounds were obtained from literature, and when they did not exist, they were derived from N(L)OAEL values by dividing them with an uncertainty factor of 100 [36]. RQ values ≥ 1 indicate the possibility of risk if the compound is ingested by drinking water consumption considering a lifelong exposure, while for RQ values ≤ 0.2 the risk for adverse human health effects is negligibly low [9,33]. Since in the majority of the samples more than two compounds occurred, a mixed health Risk Quotient (RQ mix ) was calculated as a sum of individual RQs taking into account the Concentration Addition (CA) concept, as proposed by Qin et al. [37].

Validation Results
Six-point calibration curves of a final range of 1000-10,000 ng/L-taking into account the preconcentration factor-were built and used for quantification and for defining linearity, trueness, uncertainty, recovery, LOD and LOQ for each target compound (Table 1).
Good coefficient results were obtained for all the molecules (range 0.9951-0.9999), indicating a good linear correlation. Concerning the systematic and random errors, for uncertainty, accepted values were RSD ≤ 20%, for recovery within a range of 70-120% and for trueness ≤ 30%, following the ISO/IEC 17025 requirements [31]. Satisfying results within the required ranges were obtained for each point of the calibration curve, and those obtained for 4000 ng/L are reported in Table 1 (as an example). The recovery of the compounds after the off-line SPE treatment was checked in 4 different real water and 2 Milli-Q water samples spiked with the mix of the target compounds at two different concentrations (4 ng/L and 10 ng/L) and resulted in a range of 85.5-128% for all the compounds. Regarding the limits of Detection and Quantification, the guidelines of the ICH (International Conference on Harmonisation) Method [38] were followed. For calculating the LOD the ratio between the standard deviation of the y-intercepts of 15 replicates of the six-point calibration curve (taking into account the preconcentration factor of 1000) and the slope of the calibration curve was multiplied by 3.3, while the LOQ by multiplying 10 times the same ratio (Table 1).

Screening Assessment in the Study Area
Within the context of Green Analytical Chemistry and in order to avoid a blind monitoring, a correlation study between the already existing sampling points in the area and the potential pollution sources was done based on a geographical model developed in another study [28]. Of note, 270 sampling points among the Metropolitan Area of Turin, including both surface and groundwater, resulted at a higher risk based on spatial regression, which correlated their geographical position with WWTPs, hospitals and care houses within a radius of 5 km, taking into account the nearest-neighbor points as well. In total, 325 samples were analyzed, 287 were groundwater and 24 were surface water. For raw samples-including both surface and groundwater matrices-in which the highest PhACs' concentrations were detected, treated or finished water samples from the same areas were analyzed as well. In this way, 14 treated water samples were analyzed as well in order to reassure their good quality, and take the appropriate measurements if necessary. The average concentration detected in the area as a sum of the sixteen target compounds was 28.32 ng/L (ranging from 2.02 to 523.36 ng/L) in groundwater and 18.54 ng/L (2.02-82.05 ng/L) in surface water. In 40 samples, none of the target compounds was detected above their individual LOQs. Only one compound was detected in 52 samples, indicating that a mix of them was present in the majority of the samples. The maximum number of coexisting compounds was 11, and was detected only in one sample of groundwater. This sampling point is close to two WWTPs and one care house, indicating and confirming the high risk of pollution in this area again. The range of the individual detected concentrations in the study area was between 0.08 ng/L and 483.94 ng/L. Table 2 and Figure 2 summarize the occurrence concentrations for all the target compounds.
From the sixteen target compounds included in this study, only two of them, ofloxacin and erythromycin, were not detected in any of the samples in concentrations higher than their individual LOQs (1.64 ng/L for ofloxacin and 0.81 ng/L erythromycin). The lack of detection results for these two compounds could depend on the human consumption trends in the area or on the compounds' physicochemical characteristics, enabling them to be adsorbed or biodegraded. These results are in accordance with a study from Verlicchi et al. [10], that did not detect ofloxacin and erythromycin above their Caffeine is generally reported as one of the most abundant compounds in the aquatic environment worldwide. However, the concentrations found in this study are significantly lower than those reported in other studies (in the scale of µg/L) [39]. Moreover, ketoprofen was detected in 143 groundwater samples with an average concentration of 6.51 ng/L (0. .98 ng/L), and in 21 surface water samples at an average concentration of 5.84 ng/L (0.43-71.84 ng/L). The wide range of the detected concentrations for ketoprofen is also confirmed from other studies in highly urbanized areas in Italy [12,39] and could be correlated to socioeconomic aspects.  From the sixteen target compounds included in this study, only two of them, ofloxacin and erythromycin, were not detected in any of the samples in concentrations higher than their individual LOQs (1.64 ng/L for ofloxacin and 0.81 ng/L erythromycin). The lack of detection results for these two compounds could depend on the human consumption trends in the area or on the compounds' physicochemical characteristics, enabling them to be adsorbed or biodegraded. These results are in accordance with a study from Verlicchi et al. [  mix of the two hormones 17-beta estradiol and estrone was present, while none of the two was detected in 148 of them.
The highest detected concentrations in this study were for atenolol 483.94 ng/L, estrone 125.97 ng/L, carbamazepine 183.49 ng/L, ketoprofen 152.88 ng/L, and diclofenac 121.46 ng/L. All of them were detected in groundwater samples around WWTPs, highlighting the need to implement new removal techniques. In general, the findings of this study were in accordance with the literature reporting occurrence patterns in Italy [11][12][13]39] and in other countries as well [9,15,18,[40][41][42][43]. However, in order to better estimate the impact of PhACs in the studied area, further information on their occurrence through time is needed. Additional monitoring campaigns are already planned in order to better assess the risks these molecules can cause.

Occurrence of Pharmaceuticals in Treated/Drinking Water
Even if a limited number of studies examining the occurrence of PhACs in drinking water are available, their existence has been confirmed in tap water around the globe [19][20][21]40]. Moreover, these studies claimed that conventional treatments in DWTPs like flocculation and sedimentation are unable to remove PhACs from water completely, especially when present in trace levels (in the order of ng/L) [18]. An improvement of treatment lines including steps, like ozonation and adsorption with activated carbon, is necessary in order to remove them. Hence, the best solution for drinking water companies in order to reassure safe and good quality of drinking water is a combination of them in their treatment lines [15,44]. SMAT as a drinking water company has incorporated this approach in particular for more vulnerable water resources. In such cases, DWTPs include multiple steps such as pre-settling, ozonation, clarification-flocculation, oxidation, filtration with activated carbon and/or ultrafiltration and final disinfection.
Treated samples originating from both surface and groundwater have been included in this study as well and the results showed that atenolol, azithromycin, erythromycin, sulfamethoxazole, trimethoprim, diclofenac, ofloxacin, ciprofloxacin, cyclophosphamide, clarithromycin, estrone, and 17-beta-estradiol were not quantified at all, even if most of them were present in the raw water. The reasons for their absence could be due to different phenomena: biodegradation, adsorption on the carbon filters and oxidation, mainly chlorination [15,18]. On the other hand, even if carbamazepine, caffeine, ibuprofen and ketoprofen have been detected in some treated samples in concentrations above their individual LOQs, they were still at a very low level. One parameter-other than the consumption trends in the area-that could explain their existence in treated water, is their hydrophilic behavior (log Kow < 3.0), since those with higher log Kow values are expected to be adsorbed on the particles and removed through the treatment line steps. These results have been confirmed by other studies as well, which report the occurrence of these compounds after the treatment lines [15,36]. Figure 3 highlights the differences between the sum of detected concentrations in raw and treated water samples and the good degree of efficiency of the applied treatments.

Individual Compounds
Taking into account the calculation of the log Kow values, only ten out of the sixteen target compounds of this study were considered as potential threats to human health if present in drinking water. This assumption was confirmed by analyzing samples after treatment with ozonation, GAC filtration and chlorination, which resulted in negligible low or zero concentrations of PhACs. However, ketoprofen and ibuprofen-even if with log Kow values higher than 3-were included in the risk assessment as well since they were detected in treated samples.
pGLV values could not be derived from toxicological data in the literature-confirming knowledge gaps in PhACs risk assessment estimation [45]-and they were calculated using ADI values and, where not existing, N(L)OAEL values. All the ADI and N(L)OAEL values were obtained from literature and the most restrictive value was used ( Table 3). The pGLVs ranged from 0.07 µg/L for ofloxacin to 5285 µg/L for the psychoactive compound caffeine. The RQ average for every compound was calculated as the ratio between the derived pGLV value and the mean detected concentration in raw water sources, while the RQ max as the ration between the pGLV value and the maximum detected concentration for each compound. All determined RQis were lower than 0.2 ( Table 3), indicating that none of the target compounds could potentially pose a risk of adverse health effects to humans even after a lifelong exposure. Even if most of the target compounds were present in surface and groundwater samples in the area, their quantification frequency was low, indicating low probability of threat to human health. These outcomes are in accordance with other studies, which report that the majority of the detected contaminants in drinking water sources do not pose individually a risk to human health [6,9,33,35,46]. sary in order to remove them. Hence, the best solution for drinking water companies in order to reassure safe and good quality of drinking water is a combination of them in their treatment lines [15,44]. SMAT as a drinking water company has incorporated this approach in particular for more vulnerable water resources. In such cases, DWTPs include multiple steps such as pre-settling, ozonation, clarification-flocculation, oxidation, filtration with activated carbon and/or ultrafiltration and final disinfection.
Treated samples originating from both surface and groundwater have been included in this study as well and the results showed that atenolol, azithromycin, erythromycin, sulfamethoxazole, trimethoprim, diclofenac, ofloxacin, ciprofloxacin, cyclophosphamide, clarithromycin, estrone, and 17-beta-estradiol were not quantified at all, even if most of them were present in the raw water. The reasons for their absence could be due to different phenomena: biodegradation, adsorption on the carbon filters and oxidation, mainly chlorination [15,18]. On the other hand, even if carbamazepine, caffeine, ibuprofen and ketoprofen have been detected in some treated samples in concentrations above their individual LOQs, they were still at a very low level. One parameter-other than the consumption trends in the area-that could explain their existence in treated water, is their hydrophilic behavior (log Kow < 3.0), since those with higher log Kow values are expected to be adsorbed on the particles and removed through the treatment line steps. These results have been confirmed by other studies as well, which report the occurrence of these compounds after the treatment lines [15,36]. Figure 3 highlights the differences between the sum of detected concentrations in raw and treated water samples and the good degree of efficiency of the applied treatments.  Among the target compounds, the two hormones, estrone and 17-beta estradiol, are well known for their endocrine disruptive activity. Their presence in surface and groundwater sources could result in severe risks to human health. To prevent these negative health effects from their possible occurrence in drinking water, chlorination and ozonation have been reported as efficient remediation technologies [47]. In this study, 17-beta estradiol and estrone have not been detected in the treated water samples for drinking water consumption, as in the studied DWTP both treatment techniques occur, confirming the literature's results. Hence, human health risks from the two hormones are not reported in the study.

Risk Assessment of Combined Exposure
Since in the majority of the samples more than one PhAC was present, an estimation of the risk only for individual compounds could result in risk underestimation [36,51]. However, since toxicological data of mixtures are limited, in this study we calculated the RQ of the mixtures as a sum of the individual RQs of the detected compounds according to the concentration addition (CA) concept [51]. This concept is widely used for calculating the combined risks of exposure and it assumes that the different compounds in the mixture will not interact among them, since they share the same mechanism of action and the same toxicity target [25,37]. All the components in a mixture contribute to the total toxicity depending on their concentration, resulting to the expectation that even if the individual compounds do not pose a risk, the mixture could pose it due to the addition effect [25]. This assumption was confirmed by the results of this study, that showed a risk of the combined exposure higher than the individual, but negligibly low as well (lower than 0.2). Although only some PhACs were taken into account in this study, usually a larger number of pollutants exists in the aquatic environment-even in trace levels-highlighting the fact that the mixture risk assessment is incomplete and further research is needed in order to find new ways of estimating it [37,44].

Conclusions
This study has addressed, in the first place, the presence of pharmaceuticals and hormones in surface and ground water in the Metropolitan Area of Turin (Italy, Piemonte). Prior to the screening assessment, a correlation study was performed in order to identify the areas at higher contamination risk and the good quality of the criteria employed was confirmed by the results obtained. Fourteen out of the sixteen compounds analyzed have been detected at low concentration ranging from tens to hundreds of ng/L. Since these water resources are used as catchment areas for drinking water production, a human health risk assessment was included. The results showed that risk for adverse human health effects was negligibly low-both for individual compounds and the mixture of them-in water sources before treatment, and almost non-existent in treated/finished drinking water. Nevertheless, the results of this study can be relevant for the prioritization of hazardous substances (as reported in the just-issued Drinking Water Directive (2020/2184/UE) in order to address suitable monitoring campaigns and any necessary countermeasures to be adopted for safeguarding these essential resources. Finally, they could be used for filling the knowledge gaps and attract attention to the need for regulations aimed at reducing the spread of pharmaceuticals and hormones in the environment and, in particular, in natural water resources, which is a major concern worldwide. Author Contributions: Conceptualization, R.B. and D.P.; methodology, D.P. and S.M.; software, D.P. and S.M.; validation, D.P.; formal analysis, D.P.; investigation, D.P. and S.M.; resources, G.B.; data curation, D.P. and S.M.; writing-original draft preparation, D.P.; writing-review and editing, D.P., R.B. and P.C.; supervision, R.B.; funding acquisition, P.C. and R.B. All authors have read and agreed to the published version of the manuscript.
Funding: This paper is part of a project that received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No. 765860 (AQUAlity).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Conflicts of Interest:
The authors declare no conflict of interest.