Human Health Risk Assessment of Pharmaceuticals in Water: Issues and Challenges Ahead

This study identified existing issues related to quantitative pharmaceutical risk assessment (QPhRA, hereafter) for pharmaceuticals in water and proposed possible solutions by analyzing methodologies and findings of different published QPhRA studies. Retrospective site-specific QPhRA studies from different parts of the world (U.S.A., United Kingdom, Europe, India, etc.) were reviewed in a structured manner to understand different assumptions, outcomes obtained and issues, identified/addressed/raised by the different QPhRA studies. Till date, most of the published studies have concluded that there is no appreciable risk to human health during environmental exposures of pharmaceuticals; however, attention is still required to following identified issues: (1) Use of measured versus predicted pharmaceutical concentration, (2) Identification of pharmaceuticals-of-concern and compounds needing special considerations, (3) Use of source water versus finished drinking water-related exposure scenarios, (4) Selection of representative exposure routes, (5) Valuation of uncertainty factors, and (6) Risk assessment for mixture of chemicals. To close the existing data and methodology gaps, this study proposed possible ways to address and/or incorporation these considerations within the QPhRA framework; however, more research work is still required to address issues, such as incorporation of short-term to long-term extrapolation and mixture effects in the QPhRA framework. Specifically, this study proposed a development of a new “mixture effects-related uncertainty factor” for mixture of chemicals (i.e., mixUFcomposite), similar to an uncertainty factor of a single chemical, within the QPhRA framework. In addition to all five traditionally used uncertainty factors, this uncertainty factor is also proposed to include concentration effects due to presence of different range of concentration levels of pharmaceuticals in a mixture. However, further work is required to determine values of all six uncertainty factors and incorporate them to use during estimation of point-of-departure values within the QPhRA framework.


Introduction
In recent years, pharmaceuticals in water have received growing attention from environmental and health agencies all over the world and have become one of the emerging pollutants due to their frequent detection in the water environment [1][2][3][4][5]. The fact that pharmaceuticals are manufactured with the intention to cause biological effects has raised concerns about the impacts of unintentional pharmaceutical exposure on the health of human and ecological communities. Despite the relatively fast growing numbers of studies on ecological/environmental risk associated with pharmaceuticals in water, the number of publications related to studies on human health risks remains small (Figure 1), however, the trend is increasing with time. Even though risk from exposure to pharmaceuticals in drinking water is minimal [3,[6][7][8][9][10][11][12], information about characterization of pharmaceuticals is still lacking.
In addition because of increasing public concern regarding potential health effects due to presence of pharmaceuticals in environment [13][14][15], it becomes important to understand and analyze different aspects of pharmaceutical exposures to humans, the associated health risks, and existing knowledge and data gaps.
The objective of this study is to identify existing issues within the quantitative pharmaceutical risk assessment (QPhRA, hereafter) framework by analyzing published risk assessment methodologies and frameworks and propose possible suggestions and research needs. Findings of this study are expected to highlight existing issues within the QPhRA framework and help in shaping future research directions towards filling the data and methodology gaps. Search results using keywords shown in legends from the -ScienceDirect‖ database on December 31, 2009. Note: For -Pharmaceuticals + risk + water‖ keyword, total of 39,039 articles were found. For -Personal care products+risk+water‖ keyword, total of 20,438 articles were found. For -Endocrine disrupting chemicals + risk + water‖, total of 3,601 articles were found.

Identification of Existing Issues
This study (1) reviewed retrospective site-specific quantitative pharmaceuticals risk assessment (QPhRA) studies from different parts of the world (U.S.A., United Kingdom, Europe, India, etc.), (2) Analyzed information about four steps of the QPhRA methodology, and (3) Analyzed different identified/addressed/raised issues by different studies. The QPhRA process helps in estimating the nature and probability of adverse health effects in humans who may be exposed to pharmaceuticals from contaminated environmental media [16]. It primarily involves four major steps: (1) Hazard identification, (2) Exposure assessment, (3) Dose-response relationship, and (4) Risk characterization [3,[6][7][8][9][10][11][12][15][16][17][18]. The reviewed retrospective site-specific QPhRA are summarized in Table 1. Following sections briefly discuss the stages of the QPhRA and related existing issues, needing more attention. It is important to note here that this list presents a brief summary of QPhRA studies highlighting different QPhRA steps and methodologies used by these studies and it does not necessarily represent the complete list of all QPhRA studies published so far.  Generally, both detected and modeled pharmaceutical concentrations are used in estimating risk for humans due to pharmaceuticals in water. The different pharmaceuticals studied can be seen in Table 1.
With the advancement of detection techniques for pharmaceuticals in the environment, there are extensive published data on the occurrence of pharmaceuticals in water [4,18], some of which have been used in risk assessment studies. For example, the results of the national reconnaissance for pharmaceuticals and other contaminants in U.S. streams conducted by the U.S. Geological Survey during 1999 and 2000 [23] was used by Schwab et al. [3] to conduct the risk assessment. Also, Benotti et al. [1] analyzed 20 pharmaceuticals in source water, finished drinking water, and distribution system water from 19 U.S. water utilities between 2006 and 2007. These occurrence data are useful for conducting QPhRA, especially for the assessment of risks associated with drinking water consumption.
Modeled pharmaceuticals concentrations in water have also been used in different QPhRA studies. For example, two commonly used models are: (1) Pharmaceutical Assessment and Transport Evaluation (PhATE) [24] and (2) Geography-referenced Regional Exposure Assessment Tool for European Rivers [GREAT-ER, 17,25]. The PhATE model is used to obtain predicted environmental concentrations (PECs) of active pharmaceutical ingredients (APIs) that results from patient's use of medicines in 11 watersheds selected to be representative of most hydrologic regions of the U.S. [24]. It estimates values of PECs at drinking water locations and at stream segments based on flow summary statistics. The GREAT-ER model estimates concentrations of pharmaceuticals in stream segments of ten watersheds in Belgium, France, Germany, the U.K. and Netherlands by using Monte Carlo simulation to generate distributions of concentrations in segments which reflect the variability of various model parameters [8,17].
In general, modeled pharmaceutical concentrations are used in the case of no availability of appropriate concentration data and are useful for those substances which might be present at very low levels in environment. However, very few studies have validated these models before using for risk assessment purposes, suggesting the need for extensive model validation with field observed pharmaceutical concentration data. Further, proper considerations are also important for pharmaceuticals with low environmental concentrations due to problems with analytical detection methods.

Pharmaceuticals-of-concern
The review of pharmaceuticals studied during different QPhRA studies, summarized in Table 1, indicates that studies have conducted risk assessment on a diverse range of pharmaceuticals in environmental waters depending on multiple criteria, such as occurrence, analytical capability, chemical properties, public perception, and possible health effects [3,[6][7][8][9][10][11][12][15][16][17][18]. The selection of important pharmaceuticals in water depending on multiple criteria and subsequent risk assessment is a complex task. Although, different prioritization approaches are available for identifying pharmaceuticals-of-concern in both stream water and finished drinking water [10,22], they have not yet been integrated with the QPhRA framework.
The integration of pharmaceutical prioritization frameworks with the QPhRA frameworks has the potential of providing a holistic tool to different stakeholders for conducting pharmaceuticals risk assessment on a priority-basis. With regards to types of pharmaceuticals, a constant update of the pharmaceuticals list is required to include all newly detected pharmaceuticals, such as different metabolites and transformed products from parent organic compounds [26][27][28], depending on advancement of analytical methods.
It is worth noting here that the list of emerging organic compounds detected in environmental waters is increasing everyday due to improved detection of metabolites and transformed products from parent compounds [28], and this, frequent update of prioritized list is required to reflect occurrence of newly detected compounds in environmental waters. The developed lists should be region-specific reflecting concentration profiles of pharmaceuticals on a water body-level. Findings of the Larsson et al. [14] study illustrated the need for taking this kind of prioritization approach, where they reported very high levels of pharmaceuticals in wastewater effluent in Pantanchery (INDIA), with the most abundant pharmaceutical ciprofloxacin reaching up to 31,000 μg/L concentration level. Further, Schwab et al. [3] also proposed to consider regional effect during prediction of pharmaceuticals consumption and thus excreted pharmaceutical concentration levels in domestic wastewater. These reported findings illustrate the need for conducting water body-level QPhRA and avoid the practice of generalization the QPhRA risk estimates for different exposure scenarios.

Exposure routes
Most of the previous QPhRA studies have used scenario evaluation-based approach for estimating risk using different assumptions for developing exposure [3,[6][7][8][9][10][11][12][17][18][19][20][21] (Table 1). This approach is one of the three U.S. EPA recommended approaches for exposure assessment [29] and it involves consideration of exposures through oral ingestion route, which generally happens during consumption of finished drinking water and/or fish (Table 1). This approach is consistent with the criteria used for estimating predicted-no-effect-concentration (PNEC) for pharmaceuticals in water [30]; however, other indirect exposure scenarios are also possible. For example, exposure to pharmaceuticals could also occur in following scenarios: (1) Indirect ingestion of food crops and/or vegetables irrigated with reclaimed wastewater or grown on sewage-sludge-amended soil [2], (2) Inhalation of pharmaceuticals during application of reclaimed wastewater for irrigation purposes, and (3) Dermal exposure. Most of the previous QPhRA studies have generally ignored these exposure scenarios, assuming a relatively smaller risk estimate from these scenarios compared to direct exposure scenario [3,10], which could be valid for pharmaceuticals with low vapor pressure but not for different class of pharmaceuticals. Review of most of the QPhRA studies indicates that comprehensive risk assessment studies including different exposure routes are required for different classes of pharmaceuticals.

Exposure-related parameters
Exposure-related parameters (such as human ingestion rate and exposure duration) have been generally obtained using scenario-specific information during the hazard identification step and also using information given in the U.S. EPA Exposure Factors Handbook [31]. In addition, assumptions are also made in the absence of exposure-related data for pharmaceutical concentration, exposure frequency, and exposure duration. For example, studies have used surface water pharmaceutical concentration for estimating risk due to pharmaceuticals in finished drinking water, assuming that the drinking water treatment plant (DWTP) does not remove any pharmaceuticals from surface water [3,8,11,12,32]. Although, this approach is a conservative way of checking if there is any risk due to surface water, it does not represent the effect of water treatment plant on fate of pharmaceuticals and possible production of any other metabolites, which might be more dangerous than the parent pharmaceutical compound [12,26]. For example, a brief review of removal of different pharmaceuticals from full-scale DWTPs (Table 2) shows that some of the pharmaceutical compounds are removed completely from water whereas some of other pharmaceutical compounds are persistent in water in a conventional DWTP. Also, most of the pharmaceutical compounds investigated were found to be removed more than 90% from water in advanced DWTP. These findings indicate that the effect of removal effectiveness of different plant types should be included in QPhRA. For the case of low pharmaceutical-based exposure risks from stream water, related pharmaceutical-based exposure risks from finished drinking water would also be smaller due to effect of DWTP in removing pharmaceuticals from water. Considering this aspect, use of a source-specific pharmaceutical concentration is recommended for risk estimation purposes.

Uncertainty factors
This step involves estimation of (1) response values (i.e., chronic daily intake (CDI) of pharmaceutical compounds) using exposed pharmaceutical dose and (2) benchmark values to compare calculated exposed pharmaceutical dose. For non-cancer effects, the exposed dose is usually compared with a health-based limit (HBL) (such as reference dose (RfD) or acceptable daily intake (ADI)); for carcinogenic effects, the dose is usually compared with a risk-specific dose (RSD) (i.e., dose associated with a target risk, for example 10 −6 , 10 −5 , etc.) and estimations of these benchmark values generally involve utilization of results from previous toxicity assessment studies [32] and characterization of safety/uncertainty factors (UFs). Table 2. Removals of emerging organic chemicals in drinking water treatment plans: (a) Conventional treatment: Combination of filtration (sand), clarification, GAC adsorption, and chlorination unit processes and (b) Advanced treatment: Combination of conventional treatment unit processes with ozonation, ultra-violet irradiation, membrane filtration unit processes) [1,4].

Less than 90% removal
More than 90% removal Conventional treatment Name of chemicals Generally, uncertainty factors have been classified into five categories: (1) Intraspecies variability (i.e., human to human) (UF 1 ), (2) Interspecies variability (i.e., extrapolation from animal to human) (UF 2 ), (3) Extrapolation from subchronic to chronic exposure (UF 3 ), (4) Extrapolation from low-observable-adverse-effect-level (LOAEL) estimate to no-observable-adverse-effect-level (NOAEL) estimate (UF 4 ), (5) Database quality and extrapolation (UF 5 ), and a modifying factor based on professional assessment (MF). A review of different QPhRA studies indicated that the degree to which UFs were investigated and utilized varied significantly. Unlike NOEL estimates determined from toxicity studies, selection of UF values involve more subjective judgment [3] and presents a state of difficulty. Some studies have used default values of different UFs, depending on types of uncertainties they represent. For example, Schulman et al. [11] considered only UF 1 and UF 4 types of uncertainties and used a default value of 30 for the combination of these two factors. Schwab et al. [3], Cunningham et al. [8], and Kumar and Xagoraraki [10] discussed all five types of UFs and incorporated them into the development of ADIs.
Recently, some researchers have advocated the derivation and use of non-default values for UFs [3,6]. For example, Bercu et al. [6] essentially used UF 1 , UF 2 and UF 4 types of uncertainty factors in their risk assessment studies and used non-default values for some of these UFs. The use of non-default values of UFs for deriving estimates of ADIs appears to be a more representative approach as it does not include any extrapolation-based assumption and it determines values of UFs depending on uncertainty type and other considerations. Depending on availability of pharmaceutical-based data, non-default values representing toxicodynamics and toxicokinetics of different pharmaceuticals (i.e., chemical-specific adjustment factors) should be used [6,26].
In addition, the uncertainty of long-term/chronic effects associated with exposure to pharmaceuticals in water has also been mentioned in most of the previous QPhRA studies ( Table 1). Considerations of interactions of exposure duration and environmental pharmaceuticals concentrations become important due to the fact that some pharmaceuticals are designed to achieve acute effects and some pharmaceuticals are designed to achieve chronic effects. Although, the current QPhRA methodology uses uncertainty factor (UF 3 ) to account for sub-chronic to chronic exposure extrapolation long-term effects might occur at relatively lower concentrations than those tested in toxicity experiments and might follow different toxicodynamic mechanisms than those extrapolated from short-term studies [20]. Thus, more long-term toxicity studies or experimental-simulation based hybrid approach are required to predict long-term toxicity effects.

Endpoints
A chemical may elicit more than one toxic effect (i.e., endpoint), even in one test animal, resulting in different NOEL values corresponding to different effects [32]. Generally, the identification of toxicological properties of a given pharmaceutical during QPhRA may include analyses of all possible health endpoints. However, due to constraints of time and resources, an in-depth analysis is rarely carried out for each health endpoint. For certain pharmaceuticals, endpoints might be defined from different types of experiments to further calculate values of ADIs. Uncertainties exist with the choice of endpoint and thus with the estimation of ADI values [3]. For example, Webb et al. [12] used toxicologically-based ADI, microbiologically-based ADI, pharmacologically-based ADI, and also therapeutic dosage as estimate of ADI similarly to other studies [6,8,11].
A direct consequence of identification of different endpoints is the generation of different ADIs for chemicals, which can be a source of considerable confusion when the ADIs are used exclusively in risk management decision making [32]. The use of different approximations for calculating ADI estimates poses an uncertainty in risk estimates, needing proper consideration. Theoretically, the critical endpoint used in the dose-response assessment should be the effect exhibiting the lowest NOEL [32]. However, in the previous practice of QPhRA, significant differences often exist between the ADIs of the same pharmaceutical calculated by different studies (Table 3). For example, for antibiotics such as doxycycline, tetracycline, and oxytetracycline, Webb et al. [12] used 3µg/kg/d whereas Schwab et al. [3] used 30 µg/kg/d as estimates of ADI values for each of these three antibiotics ( Table 3). The primary reason of discrepancy of ADI estimates between these two studies was that Webb et al. [12] used therapeutic doses for estimating ADI values whereas Schwab et al. [3] used ADI value, developed on the basis of antimicrobial resistance of human intestinal microflora using the WHO guidelines. Different estimates of ADI values for other pharmaceutical compounds such as cyclophosphamide, acetylsalicylic acid are also shown in Table 3. These observations illustrate the importance of proper selection of endpoint for a given receptor.
At present, since most of the previous QPhRA studies have reported no appreciable human health risks associated with pharmaceuticals in water, the diversified choices of health endpoints do not essentially make a significant difference in risk characterization. However, if new circumstances emerge, different choices of endpoint might lead to different results of risk characterization, and even different decisions by risk management groups. From this point of view, it might be necessary to invest efforts to standardize or give authoritative reference on the general choice of endpoints of pharmaceuticals in water regarding QPhRA studies, or even provide reference values of ADIs.
During determination of endpoints, proper consideration of receptor's susceptibility to the particular pharmaceutical is also required. For pharmaceuticals, while the therapeutic effect is considered beneficial for patient population, no benefit is presumed to be received by the individuals incidentally exposed to pharmaceuticals via ingesting drinking water or consuming fish, and hence it is often treated as an adverse effect in many QPhRA studies. For some pharmaceuticals that are developed for just one gender or age class, the therapeutic dose for the target population may not be the appropriate point-of-departure (POD) for calculating values of ADI for the non-targeted population, and consequently they may need individual evaluation [3,8]. Although the attention is generally given to the most sensitive adverse effect and sometimes, the lowest therapeutic dose has been used as the most -sensitive‖ POD for estimating values of ADI [3,8,10,12], this approach does not represent the effect of a pharmaceutical on a specific subpopulation type. For example, Kumar and Xagoraraki [10] used two types of ADI values (i.e., toxicity-and therapeutic-based ADI values) for both adults and children subpopulations for estimating risks due to exposures of carbamazepine, meprobamate, and phenytoin from stream water or finished drinking water. For proper characterization of risk estimates, this approach appears to be preferable as it provides a better understanding about characterization of risk estimates. Table 3. Summary of previous studies using different acceptable daily intake (ADI) values for the same pharmaceutical compound (Information about endpoints considered during estimation of ADI values are presented in parentheses).

Pharmaceutical name
Webb et al. [12] (assumed body weight = 60 kg) Schwab et al. [3] Schulman et al. [ In addition, special considerations are also required for some classes of pharmaceuticals, such as antibiotics (i.e., with non-human target effects; for example: trimethoprim, tetracycline, oxytetracycline, doxycycline), chemicals with therapeutic dose above a toxic dose (i.e., cytotoxic effect; for example: cyclophosphamide), chemicals which have high allergenic responses (for example: benzyl penicillin, phenoxymethyl-penicillin), or chemicals with high bioaccumulation potentials (for example: 17α-estradiol) ( [3,7,12,17], Table 3). For example, cancer risk exists at any concentration levels of cyclophosphamide thus the therapeutic-based benchmark cannot be used for this pharmaceutical compound [17,33]. Antibiotics present a cause of concern due to their reported occurrence in environmental media and due to their potential for inducing antibiotic resistance. Although sufficient margin-of-safety has been observed during exposures of these pharmaceuticals from the aquatic environment [8,17], proper consideration and risk estimation are required for the case of occurrence of very high levels of antibiotics in wastewater effluents as reported recently by the Larsson et al. [14] and Phillips et al. [28] studies (concentration: >1,000 µg/L). Proper considerations are required during estimation of POD for pharmaceuticals with regards to their pharmacological or allergenic effects once their therapeutic effects subside. Overall, values of PODs should be estimated based on interaction of pharmaceuticals with endpoint-under consideration for a given subpopulation. Further, PODs should not be used interchangeably for different subpopulations, unless assumptions and conditions are documented adequately.

Sensitive subpopulation
Proper considerations of gender or age class are also required during estimation of representative ADI for QPhRA for different sensitive subpopulations (i.e., pregnant women, elderly, and children). For some pharmaceuticals that are developed for just one gender or age class, the therapeutic dose for the target population may not be the appropriate point-of-departure (POD) for calculating estimates of ADI for non-targeted population. Although an uncertainty factor of 10 is usually used to account for the variability among humans, its strength in protecting the special subpopulation remains difficult to verify for different pharmaceuticals, found in drinking water sources, thus these subpopulations need individual evaluations [3,6,9,10,11].

Mixture effects
As occurrence of multiple pharmaceuticals in water at low concentrations have been reported [1,4,13,23,33], consideration of their interactions in QPhRA becomes important as it constitutes an important uncertainty [3,8,11]. Due to lack of understanding about (1) actual composition of pharmaceutical mixtures and (2) toxicity of pharmaceuticals at low concentration levels in mixture of other pharmaceuticals, it becomes difficult to predict bodily responses to mixture of pharmaceuticals. A review of QPhRA studies presented in Table 1 indicates that so far, most of the QPhRA studies have considered risk assessment due to individual pharmaceuticals and none of them have considered effect of mixtures of different pharmaceuticals. Recently Kumar and Xagoraraki [10] used information about carbamazepine, meprobamate, and phenytoin provided by the RxList to understand their interaction with each other using a pair of two active pharmaceutical ingredients (APIs) ( Table 1) and qualitatively discussed the potential effect of simultaneous presence of different APIs. Although this approach appears to serve the purpose of understanding the interactive effect of APIs, it does not help in getting quantitative risk estimates.
To circumvent the issue of QPhRA of mixture of pharmaceuticals in water, studies have generally discussed different assumptions following the U.S.EPA [33] guideline for health risk assessment of chemical mixtures. Further, due to the present use of consideration of different UFs for estimation of HBLs and its subjectivity, the current QPhRA methodology overestimates risk estimates and is expected to compensate the effect of simplified assumption of consideration of no mixture effect on risk estimates.
Due to the potential additive, antagonistic, or synergistic nature of pharmaceuticals, any comprehensive risk assessment method addressing the issue of mixture effects is expected to be complicated [11,18]. Generally, the additive effect due to different pharmaceuticals is expected if pharmaceuticals act through the same mechanism [34]. It is worth noting here that Cleuvers [33] reported that even at concentrations at which the single substance showed no or only very slight effects, toxicity of the mixture was considerable [33]. Further these effects could be concentrationdependent as Pomati et al. [35] observed during their toxicity study using 13 drugs. A summary of these toxicity studies using mixtures of chemicals is presented in Table 4. Although most of these studies have assessed toxicity using aquatic indicator species or non-specific tests [21,33,34], findings of these studies provide perspectives about affects due to presence of different pharmaceuticals at different levels. For example, findings of Cleuvers [33] or Pomati et al. [35] are useful in conducting ecological risk assessment for aquatic species due to mixture of these pharmaceuticals within the range of concentration levels studied. Recently, Watts et al. [19] considered mixture toxicity quantitatively in QPhRA and estimated exposure ratio (i.e., ratio of minimum therapeutic dose (MTD) to environmental dose) for total of 19 non-steroidal-anti-inflammatory-drugs (NSAIDs) by combining their exposure dose values and using lowest value of MTD (i.e.,7.5 mg for meloxicam), illustrating the approach for addressing mixture effects in QPhRA quantitatively.
In general, more toxicological work is required to study interactive effects of different pharmaceuticals present in water on different end points. To use the observed mixture effects data (Table 4) for conducting QPhRA for humans, we propose to use a composite uncertainty factor representing effect of mixture of pharmaceuticals on endpoint of a particular pharmaceutical, i.e., -mixture effects-related uncertainty factor‖ (

Conclusions
This study reviewed different QPhRA studies to identify existing issues and proposed possible suggestions to address these issues, as summarized in Table 5. In general, for low concentrations of APIs, none of the QPhRA studies has identified any human health risks via exposure to drinking water, but uncertainties related to the QPhRA still exist and warrant consideration. The existing findings do not rule out the possibility of any human health. As the present risk values are estimated based on very limited knowledge about chronic effects and mixture effects of pharmaceuticals, this study proposes a development of a new -mixture effects-related uncertainty factor‖ for mixture of pharmaceuticals, similar to an uncertainty factor used for a single chemical within the QPhRA framework. In addition to all five traditionally used uncertainty factors, this factor is also proposed to include concentration effects due to presence of different concentration levels of pharmaceuticals in a mixture. However, further work is required to determine these factors and incorporate them within the QPhRA framework. Table 5. Summary of identified issues related to QPhRA and possible suggestions.

Issue
Issue description Research needs/Suggestions Measured versus predicted pharmaceutical concentration Very few predictive models for pharmaceutical concentrations have been validated [8,9,24,25]; It is difficult to model low-detected pharmaceuticals.
Validate models using measured concentrations ; Conduct uncertainty analysis of risk estimates to address issue of low detection.

Pharmaceuticals-of-concern
The list of both parent compounds and metabolites is consistently increasing [12,25,28] and it becomes difficult to conduct QPhRA for all detected compounds.
Update pharmaceuticals list and integrate prioritization approach with the QPhRA framework [36].
Consider final effects of these pharmaceuticals on different receptors during estimation of POD and conduct group-specific QPhRA for these pharmaceuticals. Source water versus finished drinking water Use of source water pharmaceutical concentration for risk estimation as a conservative approach for exposures to pharmaceutical from finished drinking water [6,8,25,12].
Conduct water source-specific QPhRA; Use source water pharmaceutical concentration as finished drinking water pharmaceutical if data on pharmaceutical concentration in finished drinking water is missing. Exposure route Assumed dominance of oral ingestion route compared to other indirect ingestion-or inhalation-related exposure routes [2,3,12].
Conduct pharmaceutical class-specific comprehensive QPhRA studies using all exposure routes for a given receptor.
Use chemical-specific adjustment factors (CSAFs) [6,26]; Use default UF values only if CSAFs are not available. Conduct long-term toxicity studies or combination of experiment-simulation based studies to predict long-term toxicity using short-term toxicity data to address the issue of uncertainty related to short-term/long-term extrapolation. Sensitive subpopulation For some pharmaceuticals that are developed for just one gender or age class, the therapeutic dose for the target population may not be the appropriate point-of-departure (POD) for calculating estimates of ADI for non-targeted population (i.e., pregnant women, elderly, children) Use subpopulation-specific POD values [3,6,8,9,11,12]; Use uncertainty factor equal to 10 only in the absence of subpopulation-related endpoints information.
Mixture effects Co-occurrence of different pharmaceuticals in water may affect risk estimates.
Discuss all assumptions involved during QPhRA for mixture of pharmaceuticals [33]. Conduct more toxicity studies to develop mixture effects-related uncertainty factors.