Cytotoxic drugs in drinking water: A prediction and risk assessment exercise for the thames catchment in the United Kingdom


  • Nicole C. Rowney,

    Corresponding author
    1. School of Geography and Environment, University of Oxford, South Parks Road, Oxford, Oxfordshire OX1 3QY, United Kingdom
    Current affiliation:
    1. The current address of N.C. Rowney is 303 1603 26 Avenue SW, Calgary, Alberta T2T 1C7, Canada
    • School of Geography and Environment, University of Oxford, South Parks Road, Oxford, Oxfordshire OX1 3QY, United Kingdom
    Search for more papers by this author
  • Andrew C. Johnson,

    1. Centre for Ecology and Hydrology, Benson Lane, Crowmarsh Gifford, Wallingford, Oxfordshire OX10 8BB, United Kingdom
    Search for more papers by this author
  • Richard J. Williams

    1. Centre for Ecology and Hydrology, Benson Lane, Crowmarsh Gifford, Wallingford, Oxfordshire OX10 8BB, United Kingdom
    Search for more papers by this author


This article is corrected by:

  1. Errata: Erratum: Cytotoxic drugs in drinking water: a prediction and risk assessment exercise for the Thames catchment in the United Kingdom Volume 30, Issue 7, 1729, Article first published online: 13 May 2011

  • Published on the Web 8/19/2009.


Cytotoxic, also known as antineoplastic, drugs remain an important weapon in the fight against cancer. This study considers the water quality implications for the Thames catchment (United Kingdom) arising from the routine discharge of these drugs after use, down the drain and into the river. The review focuses on 13 different cytotoxic drugs from the alkylating agent, antimetabolite, and anthracycline antibiotic families. A geographic-information-system-based water quality model was used in the present study. The model was informed by literature values on consumption, excretion, and fate data to predict raw drinking water concentrations at the River Thames abstraction points at Farmoor, near Oxford, and Walton, in West London. To discover the highest plausible values, upper boundary values for consumption and excretion together with lower removal values for sewage treatment were used. The raw drinking water cytotoxic drug maximum concentrations at Walton (the higher of the two) representative of mean and low flow conditions were predicted to be 11 and 20 ng/L for the five combined alkylating agents, 2 and 4 ng/L for the three combined antimetabolites, and 0.05 and 0.10 ng/L the for two combined anthracycline antibiotics, respectively. If they were to escape into tap water, then the highest predicted concentrations would still be a factor of between 25 and 40 below the current recommended daily doses of concern. Although the risks may be negligible for healthy adults, more concern may be associated with special subgroup populations, such as pregnant women, their fetuses, and breast-feeding infants, due to their developmental vulnerability.


Pharmaceuticals and water

A proportion of the pharmaceuticals that are consumed everyday by the public are excreted unmetabolized, or as conjugates that may subsequently be reactivated, into the sewage treatment system. Sewage treatment plants (STPs) often cannot completely eliminate these compounds entirely [1], and as a result, trace levels are discharged into surface waters. Increasingly, researchers have identified over 100 such compounds contaminating surface and groundwater environments [2–4]. At present, there are no regulatory requirements to monitor pharmaceuticals in the environment or statutory maximum emission levels [5].

Many researchers have focused on unintended effects that discharged pharmaceuticals may have on wildlife due to their biological mode of action. Previous examples of an impact on wildlife that have been demonstrated include the significant role of ethinylestradiol in the sexual disruption of fish [6] and the role of diclofenac in kidney failure and death in vultures in the Asian subcontinent [7]. Surface water is an important source of drinking water. In parts of England, drinking water supplies are extracted from rivers, stored in reservoirs, treated, and distributed to the public. Now that pharmaceutical contaminants have been reported in some American [8,9] and European drinking water supplies [3,10] the possibility that human health risk exists must be considered. A number of risk assessments have predicted that low levels of pharmaceuticals in drinking water are unlikely to be of concern to the general adult population and that thousands of millions of liters would need to be consumed to reach a therapeutic dose [11,12]. However, the issue of drugs that act on DNA synthesis together with vulnerable sections of the population have not yet been specifically addressed together and require closer examination. Collier [13] identified special subgroup populations, such as pregnant women, their fetuses, and breastfeeding infants, as particularly vulnerable to highly potent pharmaceuticals. Cytotoxic drugs, used in chemotherapy, are designed to disrupt or prevent cellular proliferation, usually by interfering in some way with DNA synthesis. Due to their high pharmacological potencies and fetotoxic, genotoxic (mutagenic), and teratogenic properties [14], they are potentially the most dangerous contaminants of our water system.

Because few references exist on the presence [14] or absence of cytotoxic drugs in river water, the present study undertook to use the best available information to predict what such concentrations might be at drinking water abstraction points in a real river environment and compare these with effect concentrations for these drugs. Literature was collected for 13 cytotoxic drugs on the consumption per capita, typical human excretion, and estimated proportion of residue escaping from a typical STP. These values were used to give an effluent emission per capita per day and used in the Low Flows 2000™ Water Quality Extension (LF2000-WQX) model for the Thames catchment. Subsequently, the predicted concentrations at drinking water abstraction points were calculated. These predicted raw water concentrations were then compared with effects or assumed safe levels for human consumption.

River Thames catchment as a source of drinking water

The Thames River Basin in England spans an area of 12,935 km2 and is home to 14 million people ( Mean annual rainfall for the Thames is 720 mm, runoff is 249 mm [15], and annual maximum temperature is 14.7°C and annual minimum temperature is 6.7°C ( The River Thames flows 330 km from Thames Head Bridge in the Cotswolds to Shoeburyness where it joins the North Sea near Essex, passing through towns such as Oxford, Maidenhead, Reading, and the London metropolis [15]. River water is abstracted as a source of drinking water throughout the year at Farmoor, West of Oxford, and at a number of locations around London. At Walton, the furthest downstream abstraction point on the Thames, between 20 and 40 m3/s is abstracted throughout the year. For example, in July 2003 the mean naturalized flow upstream of Walton was only 34 m3/s, but the mean abstraction was 22 m3/s; thus two-thirds of the flow was diverted for drinking water, a not untypical summer occurrence [16].

Cytotoxic drugs and chemotherapy treatment

Cytotoxic drugs may be administered intravenously, orally, or topically and dosed according to the patients' weight (g/kg) or body surface area (g/m2) [13,17,18]. To enhance cytotoxicity during cancer treatment, combination therapy is often employed [19]. Combination therapy is the use of two or more drugs, often simultaneously, to treat a medical condition. A number of scientists have suggested that chemicals or drugs with similar modes of action [3,20–22] could also act additively in the environment. Therefore, in assessments of the risk that cytotoxic drugs pose to the environment and to human health, it is important to examine families of similarly acting compounds that may be present together rather than the individual compounds acting alone.

In 2005, 239,000 new cases of malignant cancer were registered in England ( Due to an aging population, the Thames Cancer Registry predicts cancer rates to increase by one-third in England by 2020 [23]. The popularity of chemotherapy as a treatment is also increasing [24,25]. Thus, the use of cytotoxic drugs is destined to increase. The ability to carry out a study of such drugs in water requires good information on per capita drug consumption. Therefore, the choice of drugs examined in the present study was largely based on their inclusion in the most recent Department of Health (DoH) study [25]. Specifically, we focused on three important cytotoxic drug groups: alkylating agents, antimetabolites, and anthracycline antibiotics. It is important to acknowledge that the information provided in the DoH study [25] is 4-year-old data and is likely to be an underestimate of today's use because of the increasing popularity of chemotherapy in the United Kingdom [23].

Important cytotoxic groups

Physical and chemical properties. In general, cytotoxic drugs are highly water soluble with low log Kow values. (Table 1). This property is helpful from the pharmacological point of view, because it increases bioavailability and rapid clearance from the body [26]. From the point of view of a drinking water risk assessment, this is also an important property, because oral ingestion of water is the main route of exposure considered here. All cytotoxic drugs reviewed in the present study have been reported to be water soluble and orally bioavailable (Table 1). The Kow values of cisplatin and gemcitabine have not been publicly reported. Where data was available, the vapor pressures for these drugs ranged from 8.99 × 10-25 to 4.45 × 10-5 mm Hg. A low vapor pressure indicates it is unlikely that a compound will volatilize under normal conditions (25°C).

Alkylating agents. Alkylating cytotoxic drugs are nonspecific chemotherapy drugs used to stop tumor growth. They function by attaching an alkyl group onto the DNA helix [17,27]. By doing so, alkylating agents inhibit or alter DNA replication, resulting in mutation or cell death [17,28]. A potential outcome of alkylating agents' mutagenic capability is the possibility of teratogenic effects. Unwanted side effects of alkylating agents include bone marrow suppression, fertility impairment, development of acute myeloid leukemia, and urinary disorders [17]. In this class, five alkylating agents oxaliplatin, temozolomide, cisplatin, carboplatin, and cyclophosphamide were examined. Information on their excretion suggests that 5 to 68% [29–34] of the alkylating agent dosed is expelled from the body unchanged (Table 2).

In a sorption study using 3 to 10 g/L suspended solids activated sludge, the removal rates recorded for oxaliplatin, cisplatin, and carboplatin were 74 ± 6%, 96 ± 8%, and 70 ± 6% [35] at pH 7. In a further study [36], a pilot membrane bioreactor (MBR) system associated with an oncology ward in Vienna, Austria, removed 51 to 64% of these platinum-containing drugs. Degradation of temozolomide has only been reported with base hydrolysis and oxidation experiments at <5% [37]. No reports to date have indicated that temozolomide is removed from wastewater via activated sludge treatment.

The fate of cyclophosphamide has been studied in a range of laboratory activated sludge systems by Steger-Hartmann et al. [38,39] without appreciable removal being detected. Similarly, little or no removal was observed in a comparison of real sewage influent and effluent at a Swiss STP [40]. This apparent persistence and mobility in sewage systems together with its resistance to ultraviolet photolysis [41] suggests that cyclophosphamide removal will remain in the water column and persist in freshwater systems [42].

To summarize, reports so far suggest that 70 to 96% of the small, platinum-containing compounds could be removed in sewage treatment primarily through sorption mechanisms onto the sludge. Little information is available on temozolomide. Laboratory and field observations indicate little or no cyclophosphamide can be anticipated in sewage treatment.

Antimetabolites. Antimetabolite cytotoxic drugs are a cell-specific class of compounds that hinder cellular metabolism and the production of DNA. They target cells typically in the G1-S phase of mitosis [17]. Antimetabolite agents mimic essential DNA precursors and therefore interfere with cell division metabolic pathways [18]. Because antimetabolite agents are mistaken by the cell as a normal metabolite, they either inhibit critical enzymes involved in nucleic acid synthesis or become incorporated into the nucleic acid [17]. As a result, DNA replication is inhibited or incorrect codes are synthesized, thus causing apoptosis. In this class, gemcitabine, fludarabine, and fluorouracil (5-FU, a metabolite to the prodrug capecitabine) were examined. Information on their excretion suggests that 5 to 65% of the antimetabolite dosed is expelled from the body unchanged [29,30,41,43] (Table 3).

Fluorouracil has been demonstrated to be very susceptible to biodegradation in an activated sludge microcosm study, with approximately 90% transformation in 10 h at a concentration of 5 mg/L [44]. Using a laboratory-scale sewage treatment plant, Kiffmeyer et al. [41] conducted a test at milligram per liter concentrations and reported 92% removal of 5-FU after 10 d.

Table Table 1.. Properties of cytotoxic drugs used in England
FamilyDrugStructureaBioavailabilitybWater solubility (g/L) 19−25°Cb,clog KowaVapor pressure, (mm Hg) at 25°Cb
  1. a Chemfinder (

  2. b United States National Library of Medicine (

  3. c Drug Bank (

  4. d Where experimental water solubility was unavailable, predicted solubility was substituted.

inline image

A test conducted with activated sludge, hospital wastewater, and 1,660 mg/L gemcitabine reported 50% biodegradation [45]. No degradation or absorption studies were found for fludarabine. However, because it is an analog of gemcitabine, the same STP elimination was assumed in the present study.

To summarize, it would seem that biodegradation in activated sludge is important for capecitabine/5-FU, with up to 99% degradation, whereas gemcitabine and fludarabine may be biodegraded by 50%. However, it should be noted that the hydraulic retention time in activated sludge tanks is typically 10 to 14 h (a shorter duration than most of the experimental studies reported here) and the environmental concentrations would be approximately five orders of magnitude lower than those used in these microcosm studies; therefore, these values may be a guide only.

Anthracycline antibiotics. Anthracycline antibiotics originate from a microorganism belonging to the genus Streptomyces. Although the exact mechanism of action is not yet understood, it is known that anthracycline antibiotics bind tightly to and intercalate with double-stranded DNA, thus preventing replication [17,46]. In this class, epirubicin and doxorubicin were examined. Information on their excretion suggests that 6 to 45% of the anthracycline antibiotic dosed is expelled from the body unchanged [29,32] (Table 4).

Using a pilot MBR system connected to an oncology ward, Mahnik et al. [44] reported 90% removal of epirubicin and doxorubicin spiked at a concentration of 2.5 mg/L. It was assumed that activated sludge flocs would behave the same as those in a MBR; thus, this removal rate was used in our exposure assessment. Analysis of the sewage sludge indicates that the elimination of anthracycline antibiotics seems to be mainly by adsorption [14,44]. It is not clear if these compounds were also biodegraded [44,47].


Predicting environmental concentrations

A modeling exercise was performed to assess the additive effluent concentrations of major cytotoxic drugs used in England. Consumption data were obtained from a 2005 DoH report [25] assessing the usage of cancer drugs approved by the National Institute for Health and Clinical Excellence (NICE). The DoH [25] report undertaken by the National Cancer Director reviewed the variations in usage of cancer drugs across England. The report detailed consumption statistics for 2003 and 2005 for 16 recently licensed cytotoxic drugs and four standard cytotoxic drugs. In the present study, 12 DoH-reviewed cytotoxic drugs and four standards were selected based on the highest usage in England and then categorized into “mechanism of action” families. Cyclophosphamide was an additional cytotoxic drug selected for the present study because there is a large body of literature on its use and behavior [40,42]. For the purposes of the present study, it was assumed that England and Switzerland used cyclophosphamide in chemotherapy to a similar extent. Other reports have shown similar use in other European countries [41,48].

The approach used to predict drug sewage loadings at STP discharge sites followed that described by Johnson et al. [11]. Drug consumption and excretion rates were converted into per capita loading values that would arrive at an STP influent [11]. To predict the effluent values, the influent values were modified by the sewage treatment removal rates found in the literature. The 2001 England population ( was used to predict the microgram per head per day influent load for each family of cytotoxic drugs (Tables 2, 3, and 4). The 2001 census data were selected to remain consistent with the population used in the DoH report [25]. For the purpose of the present study, the upper boundary level for drug discharge from an STP was modeled. Thus, input values were derived by selecting the highest drug usage values in England from the DoH review [25] and the highest drug excretion value given in the literature. Sewage treatment removal rates were selected on the basis of the lowest reported removal rate. Similar to Johnson et al. [11], it was assumed that drug excretion would be related to the national population distribution and their associated STPs. This is based on the assumption that much of cancer treatment is handled in outpatient departments and that excretion will occur over a 24-h period (i.e., at home) [14]. However, it is still possible that higher effluent cytotoxic drug loads could occur in locations where large specialty hospitals and cancer treatment centers exist [42]. The calculations in the present study only included drugs that were administered nontopically, and this may lead to an underestimation of effluent load. Moreover, of the cytotoxic drugs selected in the present study, the DoH report [25] suggests that 5.8 metric tons of four selected antimetabolites, 43.5 kg of the two anthracycline antibiotics, and 251 kg of the five selected alkylating agents are used in England each year. Because the cytotoxic drugs selected in this study are only a subsample of each cytotoxic drug family used in the United Kingdom (, the total annual use for each group will actually be greater.

Predicting concentrations throughout the Thames catchment

Predictions were made using the LF2000-WQX model (Wallingford HydroSolutions), which was developed from the Low Flows 2000 geographic information systems (GIS) hydrological model to predict concentrations of chemicals in real catchments. Low Flows 2000 [49] was developed by the Centre for Ecology and Hydrology and Wallingford Hydro-Solutions Limited and is used for estimating flows at ungauged sites. It has been widely used by the Environment Agency of England and Wales and by the Scottish Environment Protection Agency. The LF2000-WQX can predict statistical distributions of concentrations of down-the-drain chemicals in river stretches downstream of all major STPs in England and Wales. The LF2000-WQX model is essentially based on GREAT-ER (Geography-Referenced Regional Exposure Assessment for European Rivers). The GREAT-ER model has been applied to a number of rivers across Europe and has been shown to give reasonable evidence of measured concentrations of down-the-drain chemicals [50,51]. Because the output data from the model are geo-referenced, it is easy to associate features of interest (in this case water abstraction points) with river reaches and hence predicted environmental concentrations (PECs).

Generally, estimates of per capita loads from the population served by the STP are combined with estimates of chemical removal efficiencies in STPs to give effluent loads to the river. This information combined with the population served and the dry weather flow from the each STP allows calculation of the concentrations in the STP effluents. Effluent discharges are then combined with reach-specific flow statistics (derived from an annual daily flow duration curve) at the appropriate location in the river network to calculate in-river concentrations after mixing at the point of discharge and correcting for upstream concentrations, using a simple mass balance equation in Monte Carlo simulations. The concentration downstream of a discharge point is a function of dilution and dissipation, but in this case the dissipation rate was set to zero. The calculated distribution of concentrations of each drug within a river stretch primarily represents the temporal variability in river discharge within individual river stretches. The average concentration can therefore be taken to represent the concentration of a chemical under average flow conditions, and the 90th percentile concentration values represent concentrations in rivers under low flow conditions (i.e., when minimal dilution of effluent). Williams et al. [52] provide further information on LF2000-WQX and demonstrate that modeled concentrations of biological oxygen demand, ortho-phosphate, and chloride match well with observed values. In addition, they describe, in the context of steroid estrogens (another set of chemicals discharged down the drain), the confidence and the limitations of PECs derived from GIS models in general and LF2000-WQX in particular.

Table Table 2.. Predicted alkylating agents likely to be present in sewage effluent in England based on 2005 maximum drug consumption values and Switzerland cyclophosphamide use. Predictions assume maximum excretion of unchanged drug and minimum sewage treatment plant removal rates
DrugFamilyaMaximum use in EnglandbConsumption (μcapita/d)Consumption (kg/UKc/pa)Therapeutic doseExcretion of original drugPredicted influent load (ng/capita/d)Sewage treatment plant removalPredicted load (ng/capita/d)Predicted effluent concentration (ng/L)d
  1. a Brunton et al. [73].

  2. b All figures are estimated mg used per 1,000 population [25].

  3. c The 2001 United Kingdom, England census population: 49,138,831 (

  4. d Johnson et al. [11].

  5. e Moffat et al. [32].

  6. f Lévi et al. [31].

  7. g Lenz et al. [35].

  8. h Lenz et al. [36].

  9. i Shen et al. [34].

  10. j Saravanan et al. [37].

  11. k Drugs that have not been approved by the National Institute of Health and Clinical Excellence (London, UK) [25].

  12. l Dollery [29].

  13. m Harland et al. [30].

  14. n Ren et al. [33].

  15. o Anderson et al. [74].

  16. p Bagley et al. [75].

  17. q Kiffmeyer et al. [41].

  18. r Kummerer et al. [76].

  19. s Buerge et al. [40].

OxaliplatinAlkylating agent78.900.437.78130 mg/m2e5–50% unchanged in urinee,f216.7654% activated sludge, 60% ultravioletg,h9.97E+014.99E–01
TemozolomideAlkylating agent152.200.8415.00150–200 mg/m2e5–10% unchanged in urinee,i83.63No removalj8.36E+014.18E–01
CisplatinkAlkylating agent199.001.0919.6175–120 mg/m2e,l27–65% unchanged in urinee,l546.7088% activated sludgeg1.20E+026.01E–01
CarboplatinkAlkylating agent1,458.608.01143.74200–300 mg/m2l32–65% unchanged in urinel,m5,209.2926% activated sludgeg4.38E+032.19E+01
Cyclophosphamide (Switzerland)Alkylating agentNot applicable20.6455.00300 mg/m2l<20% unchanged in urinen,o, 30–68% unchangedp14,036.40No removalq,r,s1.40E+047.02E+01
Additive concentration        1.86E+049.36E+01
Table Table 3.. Predicted antimetabolites likely to be present in sewage effluent in England based on 2005 maximum drug consumption values. Predictions assume maximum excretion of unchanged drug and minimum sewage treatment plant removal rates
DrugFamilyaMaximum use in EnglandbConsumption (μcapita/d)Consumption (kg/UKc/pa)Therapeutic doseExcretion of original drugPredicted influent load (ng/capita/d)Sewage treatment plant removalPredicted load (ng/capita/d)Predicted effluent concentration (ng/L)d
  1. a Brunton et al. [73].

  2. b All figures are estimated milligrams used per 1,000 population [25].

  3. c The 2001 United Kingdom, England census population: 49,138,831 (

  4. d Johnson et al. [11].

  5. e Dollery [29].

  6. f Moffat et al. [32].

  7. g Kiffmeyer et al. [41].

  8. h Al-Ahmad et al. [77].

  9. i Gandhi and Plunkett [78].

  10. j Kummerer et al. [45].

  11. k Mahnik et al. [44].

GemcitabineAntimetabolite2,591.5014.24255.391,000–1,500 mg/m2e<5% unchanged in fecesf, <10% unchanged in urinef,g1,423.9050%h7.12E+023.56E+00
FludarabineAntimetabolite37.600.213.7125 mg/m2e,fOral: 40% unchanged in urinea IV: 60e,i–65%f unchanged in urine134.29No sewage treatment plant removal research available; however, analogs to gemcitabine, assumed 50%6.71E+013.36E–01
Capecitabine/5-FUAntimetabolite56,758.00311.865,593.372,000 mg/m2e,f7–11% unchanged in urinej34,304.2992–99%g,k2.74E+031.37E+01
Additive concentration        3.52E+031.76E+01
Table Table 4.. Predicted anthracycline antibiotics likely to be present in sewage effluent in England based on 2005 maximum drug consumption values. Predictions assume maximum excretion of unchanged drug and minimum sewage treatment plant removal rates
DrugFamilyaMaximum use in EnglandbConsumption (μcapita/d)Consumption (kg/UKc/pa)Therapeutic doseExcretion of original drugPredicted influent load (ng/capita/d)scharge tewassetto plant removalPredicted load (ng/capita/d)Predicted effluent concentration (ng/L)d
  1. a Brunton et al. [73].

  2. b All figures are estimated mg used per 1,000 population [25].

  3. c The 2001 United Kingdom, England census population: 49,138,831 (

  4. d Johnson et al. [11].

  5. e Drugs that have not been approved by National Institute of Health and Clinical Excellence (London, UK) [25].

  6. f Dollery [29].

  7. g Moffat et al. [32].

  8. h Mahnik et al. [44].

  9. i Camaggi et al. [79].

EpirubicineAnthracycline antibiotic164.000.9016.1660–90 mg/m2f,g6–20% excreted ischargewith in urinef,g180.2290%h1.80E+019.01E–02
DoxorubicineAnthracycline antibiotic277.501.5227.3530–80 mg/m2f12i–45%f excreted unchanged in urine686.1390%h6.86E+013.43E–01
Additive concentration        8.66E+014.33E–01

Although this risk assessment study is based only on raw drinking water concentrations, it is worth reviewing what is known about the efficacy of current water purification technology as used in the Thames region. This involves holding the water for several days in a reservoir followed by sand filtration, ozonation, a second filtration through granular activated charcoal (GAC), and finally a chlorination/disinfection treatment (; [53]). Although these techniques have been demonstrated to successfully eliminate many pharmaceuticals in laboratory tests [12,54,55], the extent does vary from compound to compound [56]. Unfortunately, little information exists on their performance with cytotoxic drugs.

Granular activated charcoal has been termed the best available technology for removing natural and synthetic organic contaminants from drinking water supplies [57]. However, the efficacy of GAC to remove synthetic compounds can be hindered by the magnitude of natural organics in the water [58]. Natural organics can have a greater affinity to GAC than many synthetic organic compounds and therefore compete for adsorption sites [56]. Granular activated charcoal is best suited to hydrophobic molecules rather than more polar compounds such as cytotoxic drugs [57,59].

Ozonation (O3) often partners with GAC (usually before and after the GAC filter) to breakdown recalcitrant organic contaminants [53,60,61]. Kim et al. [62] assessed the decomposition of pharmaceuticals by O3, ultraviolet processes, and advanced oxidation processes in a laboratory-scale batch reactor. Although 90% of the pharmaceuticals were degraded by 13.2 mg/L O3, the alkylating agent cyclophosphamide (16 ng/L) was found to be less degradable (46%).

Rey et al. [63] explored ozone inactivation of cytostatic nitrogen base antimetabolites (5-FU, cytarabarine, azathio-prine, and methotrexate). Results indicate that antimetabolite concentrations of approximately 300 to 500 mg/L can be inactivated by ozonation (45 min at 16–18 mg/L). Successive Ames tests confirmed the absence of mutagenic residuals. Gemcitabine and fludarabine are also cytostatic nitrogen base antimetabolites that are likely to respond to ozonation in a similar manner.

Figure Fig. 1..

Predicted alkylating agents in the Thames (United Kingdom) catchment showing mean and 90th percentile concentrations.

Thames Water (local utility company) uses a maximum of 5 mg/L ozone to produce finished water [53]. It is unclear if an ozone treatment three times less than that used in these laboratory studies would be equally effective at eliminating all cytotoxic drugs or how effective GAC would be against these very water soluble compounds. In summary, it cannot yet be confirmed that cytotoxic drugs would all be eliminated by current water purification techniques in the Thames region.


Predicting concentrations throughout the Thames catchment

It should be recalled that in carrying out these predictions the higher consumption and excretion percentage and lower STP removal factors were selected. Therefore, our PECs represent the upper end loading of cytotoxic drugs into British rivers but occasionally are an underestimation if local hotspots exist [42,48]. With the additive calculated influent loads (Tables 2, 3, and 4), mean and 90th percentile concentration values were predicted for the combined cytotoxic drug families in the Thames catchment. For the alkylating agent group, concentrations in the catchment ranged from 0 to 145 ng/L (Fig. 1), whereas for the antimetabolites the values ranged from 0 to 27.4 ng/L and for anthracycline antibiotics from 0 to 0.7 ng/L (data not shown). It is interesting to note that predicted concentrations for the combined alkylating and antimetabolites are similar to, or exceed, those predicted for estrogens for this catchment [52].

Predicting drinking water exposure

The drinking water abstraction points examined in the present study are at Farmoor, which serves the population of much of southern Oxfordshire and Wiltshire, and Walton, which serves the citizens of London. For this risk assessment, it was assumed that no cytotoxic drugs would be removed in drinking water purification. Values were selected from predicted mean and 90th percentile raw water intake concentrations, and it was assumed that an individual consumes 2 L of water per day [13]. Exposure calculations are a product of multiplying drinking water consumption by the PEC at the intake point (Table 5).


Risk assessment of cytotoxic drugs

This risk assessment has attempted to calculate plausible upper boundary concentrations for the different selected cytotoxic drug families in raw intake water taken from the River Thames. To attempt to predict the meaning of the exposure values, they were compared with the threshold of toxicological concern (TTC) and no-significant-risk-level (NSRL) standards found in the literature and used by regulatory agencies. Threshold of toxicological concern is the level, based on lifetime exposure, for “which continual exposure can be sustained and for which the lifetime excess cancer risk is limited to the upper bound of 1 in 105 to 106” [64]. At present, the European Drinking Water Directive (98/83/EC) does not contain water quality goal or thresholds for cytotoxins [65]. The European Medicines Agency (EMEA) set a TTC intake value of 1.5 μd [66] for any individual genotoxic impurity. Kroes et al. [67] has suggested a 10-fold lower level of 0.15 μd TTC for high-potency carcinogens. Schulman et al. [68] employed a NSRL of 1 μd (based on a cancer risk recommended by the State of California, USA) for cyclophosphamide as this is the level of risk recommended by the State of California U.S. Environmental Protection Agency for drinking water. It should be noted that some have argued that there is no safe level of exposure for genotoxic contaminants [12].

Table Table 5.. Drinking water risk assessment at the Farmoor and Walton intakes on the Thames River (UK) during mean flow and 90th percentile
   Ratio of exposureVulnerable groups
FamilyPredicted environmental concentrations at intake (ng/L)Drinking water exposure (μcapita/d)Effect threshold (1.5 μd)aEffect threshold (0.15 μd)bEffect threshold (1.0 μd)cDose ingested during pregnancy (36 weeks) (μg)
  1. a European Medical Agency [66].

  2. b Kroes et al. [67].

  3. c Schulman et al. [68].

Drinking water risk assessment at the Farmoor and Walton intakes on the Thames River during mean flow conditions
 Alkylating agents
 Anthracycline antibiotics
Drinking water risk assessment at the Farmoor and Walton intakes on the Thames River during 90th percentile flow conditions
 Alkylating agent
 Anthracycline antibiotics

Given that water is abstracted from the Thames throughout the year [16], it is not unreasonable to include the 90th percentile exposure assessment, which can be considered to represent concentrations likely under low flow conditions. The predicted anthracycline concentrations even under low flow conditions are extremely low, at less than 0.1 ng/L combined, and thus represent negligible risk. However, the antimetabolite and alkylating agent combinations could reach 4 to 20 ng/L, respectively (Table 5). Accordingly, if a person drinks 2 L of water per day under these predicted cytotoxic worst case conditions, then he or she would be exposed to 8 to 40 ng/d of these combined cytotoxic drugs during the low flow period (assuming no removal in drinking water purification). The predicted concentrations for the selected combined alkylating agents are approximately a factor of 25 below the Schulman et al. [68] no risk factor, a factor of 40 below the EMEA [66] level, and a factor of 4 below the Kroes et al. [67] limit for high-potency carcinogens. It should be noted that there are over 50 cytotoxic drugs in daily use in the United Kingdom ( and the present study only looked at the combined concentration of 13.

It is not clear how the different regulatory authorities would assess the risk of inadvertent exposure to combinations of cytotoxic drugs. Webb et al. [12] argue that risk assessments based on therapeutic dose benchmarks are not suitable for genotoxic drugs. This is primarily because there is no threshold dose below which no carcinogenic effects may occur [69], suggesting that any level of exposure to genotoxic drugs is capable of causing cancer. Moreover, threshold benchmarks are typically not established for special subgroup populations, let alone for combinations of cytotoxic drugs [65]; therefore pregnant women, their fetuses, and breast-feeding infants remain an important group for risk evaluation.

Collier [13] took a pharmacological exposure approach to assess the cumulative risk of individual pharmaceutical contaminants in potable water to pregnant and pediatric patients. Of the selected list of contaminants, two were alkylating agents (ifosfamide and cyclophosphamide), and one was an antimetabolite (methotrexate). All cytotoxic drugs were categorized as contraindicated in pregnant or breastfeeding mothers unless the benefit outweighed the risk. Although the time calculated to reach a minimum clinical dose did not pose an immediate risk, it is known that subclinical doses can result in cellular physiological and morphological effects [22]. For example, cytotoxic drug exposure may induce subtle changes that manifest later in life [13]. Such manifestations include short stature and cardiovascular anomalies such as those in prenatal and pediatric cancer survivors in chemotherapy-treated patients [70,71]. Therefore, subclinical chronic exposure of cytotoxic drugs to special subgroup populations could also potentially cause long-term physiological changes [13] and therefore pose a risk.


Future challenges and research

This study has identified a number of issues why inadvertent exposure to cytotoxic drugs through drinking water in the Thames area requires further research. With cancer rates predicted to increase, the prevalence and increasing use of cytotoxic drugs will likely correlate. The ability of a proportion of cytotoxic drugs to pass through sewage treatment unchanged, the limited dilution in the Thames, and the abstraction of this water for drinking purposes even during low flows are also grounds for further research. Lastly, there is inadequate information on their removal in drinking water purification.

The modeling exercise suggested that at Walton during low flows and using high, but plausible loadings, the inlet concentrations of one group of cytotoxic drugs were close to some of the proposed safety margins. Of course, this risk is still based on informed speculation as no monitoring of these drugs at abstraction points, or tap water, has yet occurred. However, given the apparent increasing consumption of these drugs, the projected increase in population in the southeast of England, and the possibly hotter and drier summers [72] of the future, the issue warrants further investigation. The risk to healthy adults from this exposure is low. However, special subgroup populations such as newborn babies may be at an elevated risk due to their developmental vulnerability.


The Centre for Ecology and Hydrology (CEH) authors are grateful for CEH science budget support and Huber Technology (Wiltshire, UK) for contributing funding for this project. Monika Juergens at CEH is thanked for her help in manuscript translation.