Investigation of the Environmental Transport of Human Pharmaceuticals to Surface Water : A Case Study of Persistence of Pharmaceuticals in Effluent of Sewage Treatment Plants and Hospitals in Malaysia

The present work reports the occurrence and monitoring of 11 pharmaceuticals (i.e., caffeine, prazosin, enalapril, carbamazepine, nifedipine, gliclazide, levonorgestrel, simvastatin, hydrochlorothiazide, diclofenac-Na and mefenamic acid) in surface water and in the influent and effluent of sewage treatment plants (STPs) and hospitals (HSPs). A total of 105 water samples were analyzed using solid-phase extraction combined with liquid chromatography-time-of-flight/ mass spectrometry (SPE-LC-TOF/MS). The mean concentrations of the detected pharmaceuticals in STP influent and effluent ranged from < limit of quantification (LOQ) to 3909 ng L and 12 to 577 ng L, respectively. The mean concentrations of the detected pharmaceuticals in the hospital influent and effluent ranged from 28 to 1644 ng L and 20 to 1540 ng L, respectively. The highest concentration detected in the sampling points was 9099 ng L for caffeine in influent STP. The presence of prazosin has never been reported before in literature. In this study, prazosin was detected in all studied samples, and the highest concentration was 525 ng L in influent STP. Chemometric analysis was used to assess the presence of pharmaceuticals in samples.


Introduction
Presence of a large number of pharmaceuticals in different bodies of water may affect the purity of drinking water.Given that surface water is the most affected, pharmaceuticals may initially pose a problem to water utilities that use surface water as a water resource for drinking-water production.Trace amounts of emerging organic contaminants (EOCs), such as pharmaceuticals and hormones, can enter the environment.[3][4][5][6][7][8][9][10][11] Presence of pharmaceuticals in STPs was discovered in the 1970's, in which two compounds namely, clofobric acid and salicylic acid, were detected in a municipal sewage. 12ince then, research in the field has expanded dramatically, and almost 100 pharmaceuticals or their metabolites have been detected in water samples. 134][15][16] Although some studies have detected pharmaceuticals or their metabolites in drinking water, these compounds were of minor concern Vol. 26, No. 6, 2015   to humans because of their extremely low concentrations (in the range of few ng L −1 ) compared with therapeutic doses (in the ranges of milligrams). 15Presence of these compounds in drinking water is still a sign of contamination originating from sewage, so the analysis of low concentrations (i.e., ng L −1 ) of pharmaceuticals found in water samples requires highly sensitive and selective analytical methods in detecting pharmaceuticals in samples.8][19] The structure, therapeutic classes and physicochemical properties for all studied pharmaceuticals are presented in Figure 1 and Table 1.
In Malaysia, research in this field is still at its infancy, so conducting such study is required to collect more information about Malaysian aquatic environment.

Study site
All samplings were performed in Negeri Sembilan, one of Malaysian's 13 states, which lies on the western coast of Peninsular Malaysia, immediately south of Kuala Lumpur and borders Selangor on the north, with Pahang in the east and Malacca and Johor to the south.The population of Negeri Sembilan is approximately 1.0 million, and the total area is approximately 6686 km 2 .Samples were collected from four STPs, three hospital (HSPs), and two receiving surface water (SW) in the state.A total of 13 sampling points, in the present study, include the SW as well as the influents and effluents of STPs and HSPs as presented in Figure S1 in the Supplementary Information (SI) section.

Sample collection
Table 2 shows some characteristics of the four STPs and three HSPs.Samples were also collected from the surface water SW at two points (SW STP 1 and SW HSP 2).The first three STPs (STP1, STP2, and STP3) are the main sources of sewage effluent in the Langat River, and STP4 is considered the biggest source in the Negeri Sembilan.Sampling was conducted from May to December 2013 (eight months).All samples were collected in 1 L amber glass bottles with Teflon-lined caps to ensure sample integrity using a high-density polyethylene plastic bucket previously rinsed with distilled water and MeOH.The bottle head space was kept to a minimum by completely filling the bottles.The bottles were rinsed in the field twice with the sample and completely filled on the third sampling.Disposable gloves were used by the sampler to prevent any personal care products from contaminating the sample bottles.The collection of samples from influent and effluent of sewage treatment plants is to evaluate the treatment performance and is to explain which compounds are still persistent after treatment.

Method of analysis
The pharmaceuticals present in the collected samples were analyzed using a previously developed and validated method based on SPE, followed by liquid chromatographytime-of-flight/mass spectrometry (LC-TOF/MS). 27Briefly; the samples were preserved by adding 1 g L −1 of sodium azide to prevent microbial degradation.Aliquots of 100, 250, 500, and 1000 mL were taken from the STP and HSP influent, STP and HSP effluent, SW, and deionized water (DIW), respectively.The aliquots were then passed through a 0.7 µm Whatman GF/F filter (UK) to remove particulate matter present in the water samples.Subsequently, the samples were stored at 4 °C to minimize the degradation of pollutants until the samples were extracted using SPE.The pharmaceuticals were extracted from the aqueous samples using Oasis HLB (3 cc, 60 mg) SPE cartridges by means of a 10-sample GAST SPE vacuum manifold (DOA-P504-BN, USA).Exactly 500 ng of the internal standard mixture, which was composed of caffeine-13 C 3 (CAF- 13 C 3 ), simvastatin-D 6 (SMV-D 6 ), and diclofenac-D 4 (DIC-D 4 ), was added to the samples performed and mixed thoroughly before the extraction was performed.The solid-phase adsorbent was pre-conditioned with 2 mL of methyl t-butyl ether (MTBE), 2 mL of MeOH, and 2 mL of deionized water before the samples were loaded in the SPE cartridges.The samples were loaded at a flow rate of 9 mL min −1 under vacuum conditions.After sample loading, the solid-phase adsorbent was washed with 2 mL of DIW.The cartridges were then dried under vacuum at 15 mL min −1 for 25 min to 30 min to remove the residual water.The pharmaceutical compounds were subsequently eluted in 12 mL-glass tubes by sequentially passing 5 × 1 mL of MTBE, 2 × 1 mL of acetone-MeOH (21:9, v/v), and 3 × 1 mL of acetone-MeOH (9:21, v/v).The combined eluents were evaporated to dryness under a gentle stream of N 2 gas.The dry extracts were reconstituted with 500 µL of MeOH-DIW (10:90, v/v) and transferred to a 250 µLdeactivated glass insert with polymer feet inserted in amber glass vial (Agilent Technologies, USA).Exactly 30 µL of the extract was automatically injected into LC-electrospray ionization (ESI)-TOF/MS system for analysis.

LC-TOF/MS analysis
Instrumental analysis was carried out using a Dionex Ultimate 3000/LC 09115047 system (USA) equipped with a vacuum degasser, quaternary pump, and autosampler.The ESI interface consisted of the standard Z-spray TM ion source fitted with an electrospray probe.
The analytes that were detected in ESI (+) and ESI (−) modes were separated on the same column, as previously mentioned.The eight compounds, namely, caffeine, prazosin, enalapril, carbamazepine, nifedipine, gliclazide, levonorgestrel and simvastatin, and two internal standards that include CAF-13 C 3 and SMV-D 6 were analyzed in positive ion (PI) mode as shown in Figure 2. The three compounds, including hydrochlorothiazide, diclofenac-Na, and mefenamic acid, as well as the internal standard DIC-D 4 were analyzed in negative ion (NI) mode as presented in Figure 3.

Quality control
Recoveries of the studied pharmaceuticals in the different matrices of water samples were determined by initially spiking the samples with a standard mixture solution.The samples were then enriched using SPE.Six replicates were evaluated on different days, and the spiking levels used were 0.5, 1, 2, and 5 µg L −1 as shown in Table S1.Instrumental quantification limits are defined as the lowest concentration with a signal to noise ratio (S/N) of 10.Table S2 shows the linearity and limit of quantification (LOQ) of all the water samples.
Selectivity of the proposed method was investigated by analyzing the chromatograms obtained from the individual  standards, standard mixture, and solvent without standards.Robustness of the method was studied by changing the mobile phase flow, location of instrument, and volume of injection.The retention time was constant under all the conditions, indicating that the method was robust (Table S3).

Results and Discussion
Monitoring of pharmaceuticals in surface water, influent, and effluent of STP and HSPs

Occurrence of pharmaceuticals in sewage treatment plants
A total of 64 samples (i.e., influent and effluent of STPs) were collected and analysed by using LC-TOF/MS (Tables 3 and 4).
The frequency of detection and the concentration were 32:32 and 9099 ng L −1 , respectively; these values were expected for caffeine in influent of sewage treatment plants.The mean concentration of caffeine in the effluent of sewage treatment plants was 577 ng L −1 , which is considered low, compared with the results of other previous studies in the United Kingdom and Ireland at 2048 and 2700 ng L −1 , respectively. 28,29Caffeine was moderately removed through the treatment process; hence, the high concentration might be related to irregular treatment timing, as observed in STP2.Consequently, the number of detection was 13:32 in the effluent of sewage treatment plants.
The maximum concentration of caffeine in the influent of the STPs was usually 9099 ng L −1 ; however, the high concentration of caffeine detected in untreated wastewater was not only because of the amount of caffeine present in the pharmaceuticals, but also because of the presence of this compound in some products, such as coffee, tea, chocolate, and sports drinks.
The antihypertensive pharmaceuticals: prazosin, enalapril, nifedipine and hydrochlorothiazide were abundant in influent STP samples and had frequencies of detection of 18:32, 9:32, 1:32 and 27:32 respectively, and the maximum concentration was 525, 200, < LOQ and 287 ng L −1 , respectively.The poor detection of nifedipine may be related to the low consumption of consumers or rapid photodegradation when exposed to the environment.This result was in agreement with studies provided in Malaysia and Germany. 12,29Three pharmaceuticals, namely carbamazepine, hydrochlorothiazide and gliclazide were the most persistent compounds after treatment with a maximum concentration of 344, 111 and 70 ng L −1 , respectively.In some cases, carbamazepine had higher effluent concentration than the STP influent.This finding has been noticed in other previous studies. 7,9,30In this study, diclofenac-Na was detected in the influent and effluent of STP with a maximum concentration of 5049 ng L −1 and < LOQ, whereas mefenamic acid had maximum concentration of 296 and 30 ng L −1 .
In comparison, STP 4 caters for 25-fold more than STP1 and STP3.Furthermore, the pharmaceuticals detected over the same 8 months (May 2013 to December 2013) were found to be slightly low.This result might be attributed to the tertiary treatment conducted on the STP4.

Occurrence of pharmaceuticals in hospitals
A total of 24 samples were collected from 3 hospitals.Eight samples of the influent were collected from one hospital (HSP1), and 16 samples of the effluent were collected from the other two hospitals (HSP2 and HSP3) (Table 5 and 6).Table 5 shows the concentrations of studied pharmaceuticals in the influent of hospital HSP1.This hospital discharges directly to STP4; hence, no effluent sample was collected from this hospital.The concentration of pharmaceuticals in the influent HSP1 ranged from non-detected to 2979 ng L −1 as the maximum concentration.The highest concentration was for caffeine, that is, 2979 ng L −1 at high frequency detection 7:8.Antihypertensive pharmaceuticals, namely, prazosin, enalapril and hydrochlorothiazide, were detected at frequency detection of 3:8, whereas nifedipine was undetected in all samples.Levonorgestrel and diclofenac were detected only once in all collected samples.The concentration of pharmaceuticals in the effluent of hospital HSP2 and HSP3 are summarized in Table 6.The highest maximum concentration of 4131 ng L −1 was for caffeine, and the frequency of detection was 16:16.The occurrence of synthetic hormone (levonorgestrel) in the effluent of hospitals is presented in Table 6.However, the maximum and minimum concentrations of levonorgestrel were 105 and 73 ng L −1 , respectively.The antidiabetic drug (gliclazide) was persistent in the effluent of hospitals and had a frequency of detection of 15:17 and a maximum concentration of 66 ng L −1 .Hydrochlorohtiaizde was the most frequent detected drug compared with other pharmaceuticals in similar therapeutic classes at a maximum concentration of 78 ng L −1 .Carbamazepine was found at a concentration range of 12 to 142 ng L −1 and the frequency of detection was 12:18.Nifedipine was detected only once at a concentration of 264 ng L −1 .The efficiency of treatment for sewage treatment plants compared with hospitals is considered better in terms of the number of detection and concentration of pharmaceuticals.

Occurrence of pharmaceuticals in surface water
A total of 15 samples were collected from two rivers (Table 7).Caffeine is one of the most widely used nervous system drugs (stimulant drug) and is available over the counter.Caffeine was continuously detected in the surface water at a high frequency number of detection (15:15) with the highest concentration of 1213 ng L −1 .This result might be due to the non-prescribed property of caffeine.However, caffeine was detected in surface water in Italy and Romania, with maximum concentration of 1056 and 3480 ng L −1 , respectively. 31,32The lowest concentration of caffeine detected in surface water was 116 ng L −1 ; this result was in agreement with our previous study, wherein the concentration of caffeine was 257 ng L −1 . 10ntihypertensive pharmaceuticals, prazosin, enalapril, nifedipine and hydrochlorothiazide, were detected at varying frequencies.However, nifedipine was undetected in surface water during all periods of the study.This phenomenon may be related to the fact that nifedipine is easily degraded and light sensitive.Hence, nifedipine did not show high persistency in the environment.Nifedipine was also seldom reported in the environment and was absent in 11 rivers and streams samples collected from Germany. 12 This was the first time that prazosin was reported in environment matrices and was found at a concentration of 2 to 30 ng L −1 .The : prazosin; ENL: enalapril; CBZ: carbamazepine; NFD: nifedipine; LNG: levonorgestrel; SMV: simvastatin; HDZ: hydrochlorothiazide; GLZ: gliclazide; DIC: diclofenac-Na; MEF: mefenamic acid; ND: not detected; < LOQ: the lowest limit at which analyte could be quantified.number of frequency detection of prazosin and enalapril was 6:15 and 3:15, respectively.This result might be due to the rapid metabolization of enalapril by liver esterases to enalaprilat.However, a few previous studies reported that enalapril was undetected in surface water or < LOQ. 33,34The maximum concentration of enalpril was 14 ng L −1 .Hydrochlorothiazide is a diuretic drug and is the most detected pharmaceutical (13:15) in this group.The maximum and lowest concentrations were 54 and 4 ng L −1 , respectively.Hydrochlorothiazide was detected in Spain, at varying concentrations 164 and 8 ng L −1 . 33,35Simvastatin is the methylated form of lovastatin, is used to treat primary hypercholesterolemia, and is effective in reducing total and low-density lipoprotein (LDL)-cholesterol.Simvastatin was detected in surface water at low concentrations of 6 ng L −1 , as a maximum concentration with a poor frequency of detection 5:15.This result was in line with previous study reported by Al-Odaini, 29 who detected simvastatin in Malaysia at concentration less than method detection limit (MDL).Carbamazepine, diclofenac, mefenamic acid, and gliclazide, that belong to different therapeutic classes, have been detected at different concentrations of 53, 54, 40 and 36 ng L −1 , respectively (Table 7).The anti-diabetic drug, gliclazide, was detected at a high frequency number of 12:13.This observation might be attributed to 8% of the population in Malaysia take this drug. 20The impact of pharmaceuticals detected in surface water varied from sewage treatment plants to hospital discharge.However, most pharmaceuticals detected in SW HSP2 have concentration higher than those detected in SW STP1.This result could be attributed to the efficiency of treatment process in sewage treatment plants and hospitals.

Chemometric analysis
Chemometric analysis results showed that the water sampling stations were classified into three clusters, as shown in the dendogram (Figure 4).
The first cluster included the influents STP 1, STP 2, STP 3, STP 4 and the HSP 3, whereas the HSP 1 forms the second cluster.The effluents STP 1, STP 2, STP 3, and STP 4, HSP 2, downstream river of STP 1, and downstream river hospital HSP 2 represent the third cluster.The effluents from the two hospitals (i.e., HSP 2 and HSP 3) formed different clusters because of the various consumed pharmaceuticals and treatment process efficiency in each hospital.The similarity between sampling stations in the first cluster may be attributed to the similarity of the type of sample in terms of a highly polluted sample (e.g., sewage treatment plants).However, the effluent of the hospital HSP 3 was included in the first cluster because of the frequency of detection of most compounds in this location.The influent of the hospital HSP 1 was clustered alone (second cluster) because of the Vol.26, No. 6, 2015 design of the sampling point, which is as pipe discharge direct to the sewage treatment plant STP 4, it was not sink like other stations as common.The other sampling points, including the effluent of the STPs, effluent of the hospital HSP 2, and their related downstream river, was considered as cluster three.These locations were grouped together because of the high similarity in terms of concentration and frequency of detection in these locations.
Figure 5 shows the dendogram cluster of the pharmaceuticals, in which three clusters were formed.All studied compounds were clustered in different groups because they do not form one cluster.These differences were attributed to the differences in the physicochemical properties, environmental fate, and ability of the pharmaceuticals to resist degradation.
Caffeine, gliclazide, and diclofenac-Na comprised the first cluster.The second cluster included prazosin, mefenamic acid, and hydrochlorothiazide, whereas enalapril, simvastatin, carbamazepine, and levonorgestrel constituted the third cluster.The studied pharmaceuticals were clustered based on their environmental behaviour or properties.The grouping of caffeine, gliclazide, and diclofenac-Na may be due to their high consumption and non-prescribed pharmaceuticals (caffeine).
The principal component analysis was used to select the number of components extracted.According to the eigenvalue-one criterion, only principal components (PCs) with eigenvalues greater than 1 are considered as important values.In the scree plot test (Figure 6), three PCs were selected, in which the eigenvalues for PC1, PC2, and PC3 were 5.756, 1.5877, and 1.091, respectively.The scores of the samples corresponding to PC1 and PC2 are presented in Figure 7. Table 8 shows the loading of varimax-rotated factor matrix for three-factor models.Evidently, the first factor was generally more correlated with the variables than the second and the third factors.The terms 'strong', 'moderate', and 'weak' as applied to factor loadings, refer to absolute loading values of > 0.75, 0.75-0.5, and 0.5-0.3,respectively.

Figure 1 .
Figure 1.Chemical structure of the studied pharmaceuticals and internal and/or surrogate standards.
Each sample was identified by the name of the corresponding sampling station.The sample sites were classified in three different groups.The first group corresponds to the zone with high concentration influence, which included the influent of the four STPs (i.e., INF STP 1, INF STP 2, INF STP 3, and INF STP 4) and the effluent of the HSP 2. The second group corresponds to the influent of the hospital (INF HSP 1).Meanwhile, the third group includes the effluent of the STPs, effluent of the HSP 3, and their downstream river (i.e., SW STP 1 and SW HSP 2), which considered less pollution compared with zones 1 and 2.

Figure 5 .
Figure 5. Dendogram showing different clusters of pharmaceuticals detected at different water samples.

Figure 6 .
Figure 6.Scree plot of the eigenvalues of principal components.

Figure 7 .
Figure 7. Scores of all 13 water samples on the plane defined by the first two principal components.

Table 2 .
Information about the studied STPs and HSPs

Table 3 .
Mean, minimum, maximum and median concentrations (ng L −1 ) for all studied pharmaceuticals in influent sewage treatment plants during eight months

Table 4 .
Mean, minimum, maximum and median concentrations (ng L −1 ) for all studied pharmaceuticals in effluent sewage treatment plants during eight months DIC: diclofenac-Na; MEF: mefenamic acid; ND: not detected; < LOQ: the lowest limit at which analyte could be quantified.

Table 6 .
Mean, minimum, maximum and median concentrations (ng L −1 ) for all studied pharmaceuticals in effluent hospitals during nine months