Model Compound Studies

Mechanisms of chlorination by-product formation have been investigated through the use of isolated humic and fulvic acids, as well as simple compounds that are viewed as models of the complex molecules found in natural humic materials. Many humic molecules (and study models) contain electron-rich phenolic structures and/or aliphatic side chains that are vulnerable to attack by chlorine (Liao et al., 1982). It has generally been found during the studies discussed below that the specific by-products depend on the molecular structures of the humic and fulvic acids undergoing chlorination, the chlorine-to-carbon ratio, pH, and several other factors. The by-products fall into two general categories: volatile hydrophobic and nonvolatile hydrophilic compounds.

Christman et al. (1978) and Norwood et al. (1980) used resorcinol, orcinol, 3, 5-dihydroxybenzoic acid, 3-methoxy-4-hydroxycinnamic acid, and 3, 5-dimethoxybenzoic acid as models of humic molecules, based on copper oxide degradation products of humic materials (Figure 3-1). Their model compounds all consumed significant amounts of chlorine and produced measurable levels of chloroform. Resorcinol, as suggested earlier by Rook (1977), consumed a large quantity of chlorine (7 moles per mole of resorcinol) and rapidly produced 1 mole of chloroform. Similar results were produced with their other model compounds, suggesting that chloroform is a primary reaction product of chlorination of aquatic humic materials that contain substructures similar to these model compounds. Other by-products produced by their model compounds are shown in Table 3-1. High chlorine-to-carbon ratios favored the production of nonvolatile hydrophilic by-products.

Figure 3-1

Aquatic humic model compounds. From Norwood et al. (1980) with permission.

TABLE 3-1

Reaction Products from Model Compounds and Hypochlorous Acid (HOCl).

Boyce and Hornig (1983) studied chloroform production from chlorination of 1, 3-dihydroxyaromatic compounds and simple methyl ketones, which they confirmed to be efficient at producing chloroform. With isotope labeling, they unambiguously demonstrated that the C₂ position of resorcinol is responsible for chloroform generation, as previously hypothesized by Rook (1977) and Norwood et al. (1980). Boyce and Hornig (1983) further demonstrated that the specific types of chlorinated products depend on both pH and the relative concentrations of chlorine and substrate in solution. The by-products that they obtained from resorcinol at various chlorine concentrations and pH values are shown in Table 3-2 and confirm the previous observations of Norwood et al. (1980) regarding by-products formed at neutral pH.

TABLE 3-2

Reaction Products Identified from the Chlorination of 5 x 10^-4 M Resorcinol Using 5 x 10-4 M to 5 x 10-3 M Chlorine Dioxide at 10°C.

Based on these results and previous hypotheses of Moye (1967) and Rook (1980), Boyce and Hornig proposed a comprehensive mechanism for the conversion of 1, 3-dihydroxyaromatic structures to chloroform by aqueous chlorination. A portion of this proposed mechanism, modified and reproduced in Figure 3-2, involves successive electrophilic attack of chlorine to produce substituted resorcinols (I) with the eventual loss of aromatic character to produce the intermediate pentachlororesorcinol (II). This is followed by hydrolytic ring cleavage and a number of other substitution and hydrolysis reactions to produce chloroform and short-chain chlorinated acids, in this case chloromatic acid (VI).

Figure 3-2

Abbreviation of mechanism proposed by Boyce and Hornig (1983) for the aqueous chlorination of resorcinol (adapted from Norwood, 1985).

De Leer and Erkelens (1985) attempted to support the mechanism proposed by Boyce and Hornig (1983) by synthesizing the proposed intermediate pentachlororesorcinol according to the method of Zincke (1890) and subjecting it to aqueous chlorination at neutral pH. Although the chlorination of resorcinol and pentachlororesorcinol produced several identical products, large discrepancies were seen in apparent reaction rate, chloroform production, and products, indicating that pentachlororesorcinol is not a major intermediate. De Leer and Erkelens (1985) further concluded that the principal reaction and most important side reaction are

but that many side reactions producing other chloroform precursors and highly oxidized products occur.

Thus, it appears from the above studies of model compounds that nonselective aqueous chlorination of activated aromatic ring systems produces not only chloroform (a volatile hydrophobic by-product) but many nonvolatile hydrophilic chlorinated aromatic by-products as well.

Isolated Acids

Working with isolated aquatic humic and fulvic acids, Christman and co-workers (Christman et al. 1980, 1983; Johnson et al., 1982; Norwood et al., 1983) identified more than 100 different chlorination products by gas chromatographic/mass spectroscopic methods at a 4:1 chlorine-to-carbon mole ratio. Some of these products are shown in Tables 3-1 and 3-2. Chlorination of several humic and fulvic acid samples from the same source produced significant differences in product mixtures. A notable difference was that most products of fulvic acid chlorination contained chlorine, whereas most humic acid samples produced at high pH did not. In both cases, however, the dominant chlorinated products were chloroform and chlorinated aliphatic acids, especially dichloroacetic acid (DCA), trichloroacetic acid (TCA), chloroform, dichlorosuccinic acid, and dichloromalonic acid.

A variety of short-chain, nonvolatile aliphatic halogenated products (listed by Norwood, 1985) result from the exposure of aquatic humic and fulvic acids to chlorine:

The apparent dominance of C₂-chlorinated acids is in agreement with the findings of Quimby et al. (1980), who reported the tentative identification of TCA and halogenated phenols after soil extract chlorination, and Rook (1980), who found that DCA and TCA were the principal constituents in methylene chloride extracts of Rotterdam drinking water after breakpoint chlorination. However, no halogenated aromatic products were detected after chlorination of actual aquatic humic and fulvic acids under high pH conditions.

A large number of monobasic and dibasic unchlorinated aliphatic acids, from oxalic up to the C₂₇ monobasic fatty acid, were identified from the humic acid fraction (Table 3-3). Only a few of the dibasic acids were associated with the fulvic acid fraction, and almost none of the monobasic acids were detected. The dibasic aliphatic acids are generally of low molecular weight, containing 2 to 10 carbons. Most of these were detected in relatively low yield. Aromatic acids were also detected, including monobenzoic to hexabenzoic acid in all isomers, as well as small quantities of methyl-substituted aromatic acids (tentatively identified) and isomers of (carboxyphenyl) glyoxylic acids (tentatively identified). These non-chlorine-containing products of each acid are similar to the polybasic aromatic and aliphatic acids reported from potassium permanganate (KMnO₄) oxidation (Christman et al., 1981; Liao et al., 1982).

TABLE 3-3

Non-Chlorine-Containing Products of Aquatic Humic and Fulvic Acids.

Recently de Leer et al. (1985) subjected humic acid extracted from a peat soil to aqueous chlorination under degradation-scale conditions (0.38 g humic acid per liter of solution, pH 7.2, 24-hour reaction time, ambient temperature, chlorine-to-carbon ratios of 0.39:1 and 3.35:1). The lower chlorine-to-carbon mole ratio was chosen to represent typical drinking water disinfection practice, while the higher ratio was chosen to maximize product yields. Utilizing gas chromatography/mass spectrometry (GC/MS) methods, structures were assigned to more than 100 products. The product distribution was different for the two reaction mixtures.

The products detected in ether and ethyl acetate extracts of the acidified high chlorine-to-carbon ratio aqueous reaction mixture were a series of unchlorinated aliphatic monobasic and dibasic acids, aromatic carboxylic acids, and chlorinated aliphatic monobasic and dibasic acids, both saturated and unsaturated, that correspond well to those reported in the experiments on isolated aquatic humic and fulvic acids (Tables 3-1 and 3-2). The predominant chlorinated compounds were DCA, TCA, and 2, 2-dichlorobutanedioic acid (α, α-dichlorosuccinic acid), also in agreement with the earlier findings.

Aqueous chlorination of humic acid derived from soil at a high chlorine-to-carbon ratio (3.35:1) produced two new classes of compounds (Figure 3-3) (de Leer et al., 1985). These were the cyano-substituted alphatic monobasic acids, 3-cyanopropanoic acid and 4-cyanobutanoic acid, and the chlorinated aromatic carboxylic acids, 4-chlorobenzoic acid, 2-chlorobenzoic acid, 2-chlorophenylacetic acid, 4-chlorophenylacetic acid, 2, 6-dichlorophenylacetic acid, and 2, 4-dichlorophenylacetic acid. This constituted the first definitive report of the production of chlorinated aromatic compounds from the aqueous chlorination of humic material.

Figure 3-3

Chloroform precursors detected from the aqueous chlorination of a soil-derived humic acid at chlorine-to-carbon ratios of 0.39 and 3.35 by de Leer and colleagues (1985).

De Leer and coworkers (1985) found that a greater number of compounds with higher boiling points were formed at the lower chlorine-to-carbon ratio than at the higher ratio, although the classes of compounds formed were similar. Also produced at the lower ratio was a group of compounds termed ''chloroform precursors'' because they contained a trichloromethyl group adjacent to a group susceptible to further oxidation. These structures, described above, may be divided into two groups: one with the trichloromethyl group next to a hydroxyl group and the other with the trichloromethyl group next to a carbonyl group conjugated with a carbon-to-carbon double bond (Figure 3-3).

Holmbom et al. (1981, 1984) discovered a series of acids, the furanones, in chlorinated kraft pulp waste. Recently, Hemming and colleagues (1986) showed that low concentrations (µg/liter) of these compounds were formed when aqueous humic and drinking water samples were chlorinated at 1:1 chlorine-to-carbon weight ratios at pH 7. After chlorination, these nonvolatile compounds were concentrated and separated by high-pressure liquid chromatography (HPLC). Almost all of the mutagenic activity injected by chlorination was found to be in a relatively narrow HPLC fraction. After methylation by CI and EI mass spectrometry, the major contributor was tentatively identified as 3-chloro-4-(dichloromethyl)-5-hydroxy-2(5 H)-furanone. This same compound was also found by Meier et al. (1986).

A number of studies have been conducted with commercial materials of unknown origin sold as humic acid (Bull et al., 1982; Coleman et al., 1984; Meier et al., 1983; Seeger et al., 1985). These materials appear to be European lignitic coal extract rather than soil or aquatic humic acid (Malcolm and MacCarthy, 1986). Chlorination products included chloroacetonitriles, chloroketones, and chlorobenzenes (Coleman et al., 1984).

Small quantities of dichloroacetonitrile (0.2 µg/ml), 3, 3-dichloro-2-butanone (0.4 µg/liter), and 1, 1-dichloro-2-propanone (0.6 µg/liter) and relatively large quantities of pentachloro-2-propanone (1.1 µg/liter) and 1, 1, 1-trichloro-2-propanone (11 µg/liter) were also identified from a 1-g/liter Ohio River humic fraction chlorinated at pH 7 with a 1:1 chlorine-to-carbon mole ratio for 90 hours. The major products were similar to the 14 µg/ml of DCA, 35 µg/ml of TCA, and 66 µg/ml of chloroform previously found. Thus, large quantities of DCA and TCA were recovered even though extraction into organic solvents from water was carried out at pH 3.1, where recoveries of the salts of these very strong acids (pK _a <1) are poor. Even larger quantities of the chloroacetic acids (77-122 µg/ml) and significant quantities of the chloroacetonitriles (4.3-4.4 µg/ml) were found in the commercial humic material. The identifiable products for both the Ohio River sample and the commercial humic material, however, were only 23% to 28% of the measured total organic halogen (TOX) produced even though long reaction times (90 hours) and very high concentrations of starting material (1 g total organic carbon [TOC]/liter) and chlorine (35.5 g/liter) were used. Such low yields of identifiable products, even under conditions expected to produce highly degraded humic material and short-chain cleavage products, are typical of the identifiable yields of products found by others. Even under conditions where chloroform would be expected to predominate, it represents only a small quantity of the TOX produced, and DCA and TCA are produced in nearly identical amounts.

The yield of identifiable products from chlorination of fulvic and humic acids isolated from natural surface water is a small fraction of the starting organic material. In the work of Coleman et al. (1984), the 28% of the TOX identified represents less than 10% of the starting organic material identified in the Ohio River humic fraction and a much lower percentage of the TOC in the river. Some of the highest recoveries reported (Christman et al., 1983) are 14% of starting aquatic fulvic material and 53% of TOX. Both of these studies were conducted under conditions of high initial carbon concentration (0.5 to 1 g/liter) and high chlorine-to-carbon (4:1 to 1:2) mole ratios designed to maximize identifiable product yield. Christman et al. (1983) showed that the yields of chloroform and the C,—C₄ chlorinated aliphatic acids make up 17% and 36% of the TOX, respectively. The high yield of TCA in these samples is confirmed by the use of isotope-dilution MS (Norwood et al., 1986) with ¹³C-labeled TCA added to the aqueous chlorination mixture before any separation or analysis is performed (Christman et al., 1983). The confirmed dominance of TCA is in agreement with Quimby et al. (1980), who used a GC microwave-plasma emission method (Miller et al., 1982).

Miller and Uden (1983) have shown the relative concentration of chloroform, DCA, TCA, and chloral hydrate in chlorinated reaction mixtures of soil fulvic acid to be a function of pH, chlorine-to-carbon mole ratio, and time. The quantity of chloroform produced generally increased with increasing pH, chlorine-to-carbon ratio, and time. Based on the mechanism of de Leer et al. (1985), this is to be expected because of the greater number of intermediates formed from the chlorination of humic materials.

Using chlorine-to-carbon mole ratios and TOC values more typical of natural water, Reckhow and Singer (1985) showed that when aquatic fulvic acid (4.1 mg of TOC/liter at pH 7) was treated with 20 mg of HOCl/liter of water, TOX, chloroform, TCA (Figure 3-4), and DCA all increased with time. TCA and dichloroacetonitrile reached their maximum concentrations within the first few hours, then decreased rapidly with time. Their results with pH were similar to those of Fleischacker and Randtke (1983), an increase in chloroform but a decrease in TOX with increasing pH (Figure 3-5). It is clear from these studies that the importance of trihalomethanes is overstated in quantitative studies done at high concentration for long time periods and at a high chlorine-to-carbon mole ratio. Trihalomethanes are also easily quantified by GC procedures (EPA, 1979, 1980). No simple and accurate method exists, however, for the identification and quantification of even the major individual nonvolatile chlorinated compounds formed in chlorinated surface water. The principal method has been solvent extraction, followed by derivatization and GC/MS (Norwood et al., 1983) and, more recently, by isotope dilution, GC/MS (Norwood et al., 1986), and GC microwave-plasma emission (Miller and Uden, 1983). Unfortunately, even using these sophisticated methods we are unable to identify the majority of products formed in water chlorination.

Figure 3-4

Formation of organic halides from Black Lake fulvic acid as a function of chlorine contact time. Conditions: 4.1 mg TOC/liter, pH 7.0, 20 mg applied hypochlorous acid/liter. From Reckhow and Singer (1985) with permission.

Figure 3-5

Effect of pH on the formation of organic chlorine by free and combined chlorine. Conditions: 3.0 mg TOC from peat fulvic acid/liter, 20 mg chlorine dosage/liter, 100 hours contact time, combined chlorine formed by addition of free chlorine to samples (more...)

Isolated Bases

Numerous organic nitrogen compounds are present in natural surface waters (C. Le Cloirec et al., 1983a, b, c; P. Le Cloirec et al., 1983; Mallevialle et al., 1984; Ram and Morris, 1980; Thurman, 1985). These include a number of man-made nitrogen-containing pesticides and industrial compounds in trace quantities. However, the most abundant nitrogen-containing compounds are the naturally occurring amino acids, nucleic acids, amino sugars, natural porphyrin-based pigments (such as chlorophyll), and proteins. The higher molecular weight members of this group of compounds have been difficult to isolate and characterize, and little is known about the products of their reactions with aqueous chlorine other than the low-molecular-weight by-products such as chloroform.

Within the past several years, research using model compounds has elucidated the reactions of aqueous chlorine with some of the more nucleophilic examples of these compounds, which explain the origins of some of the chlorine demand of natural waters and by-products of water disinfection.

Amino acids react rapidly with one equivalent of aqueous chlorine to form N-chloramino acids (Morris, 1967):

After isolating colloidal particles from river water Helz et al. (1983) showed that amino acids associated with the particles in either a free or proteinaceous form are depleted by chlorination. The amino acids containing reactive side groups were the most reactive. A number of the N-chloroamino acids have comparatively short lifetimes and decompose losing carbon dioxide to produce aldehydes (see Scheme Ia) (Dakin, 1916; Friedman and Morgulis, 1936; Golschmidt et al., 1927; Langheld, 1909a, b; C. Le Cloirec et al., 1983a, b, c; P. Le Cloirec et al., 1983; Stanbro and Smith, 1979).

Scheme I

The intermediacy of an imine (Schiff base) that hydrolyzes to an aldehyde is suggested. Stanbro and Smith (1979) have studied the effect of pH on the decomposition of N-chloroalanine and found that the decomposition is independent of pH in the range between pH 3 and pH 9. At lower pH values the rate of decomposition accelerates. At higher chlorine-to-amino acid mole ratios, the amino acid becomes dichlorinated (Scheme Ib), and the dichloroamino group, which is even less stable than the monochlorinated derivative, decomposes to a nitrile group. The amount of nitrile increases relative to the aldehyde as the pH of the solution increases and likely involves chlorination of the imine intermediate followed by dehydrohalogenation.

Recently, C. Le Cloirec and Martin (1985) demonstrated that inorganic monochloramine can react with acetaldehyde to produce acetonitrile (Scheme II).

Scheme II

In 1976 McKinney et al. reported the presence of dichloroacetonitrile in tap water in Raleigh, North Carolina. Trehy and Bieber (1981) identified both dichloroacetonitrile and bromochloroacetonitrile in chlorinated lake and well waters in south Florida. Using model solutions of a number of naturally occurring nitrogen-containing compounds, Trehy and Bieber (1981) and Bieber and Trehy (1983) have shown that the amino acids, aspartic acid, tyrosine, and tryptophan, as well as the catabolites of tryptophan, kynurenine, and kynurenic acid, react with hypochlorite at pH 7-8 to produce significant quantities of dichloroacetonitrile. A model protein was also shown to produce considerable quantities of dihaloacetonitriles.

Trehy and Bieber (1981) have proposed that dichloroacetaldehyde and its reaction product with chlorine, trichloroacetaldehyde (chloral), would be formed by chlorination of aspartic acid, but suggest that because these aldehydes are extremely water soluble, their presence has not yet been reported.

Proteins (Scully et al., 1985) and amino acids (Bieber and Trehy, 1983; Trehy and Bieber, 1981) also react with hypochlorite to produce trihalomethanes. Although the yields of chloroform from the individual amino acids are generally low, the overall yield of chloroform from proteins is comparable to that of humic acid for solutions containing equivalent amounts of organic carbon.

Chloropicrin has been identified in chlorinated surface waters (Duguet et al., 1985; Mallevialle et al., 1983; Sayato et al., 1982). Sayato et al. (1982) showed that it can be formed by the reaction of chlorine with humic acid, amino acids, and nitro-or nitrosophenols (Sayato et al., 1982). However, the yields are not appreciable at pH values normally present during water treatment. At extremely basic pH, the yield of chloropicrin is enhanced. Duguet et al. (1985) showed that the presence of nitrite greatly enhanced the formation of chloropicrin both in chlorinated model solutions and in chlorinated natural waters. They suggest that chemists might overlook the presence of chloropicrin in water samples, if they dechlorinate those samples before analysis with thiosulfate, sulfite, or ascorbic acid. They concluded that "for customary TOC levels, low nitrite concentrations are sufficient to explain the levels of chloropicrin actually found in full-scale water treatment plants." Becke et al. (1984) demonstrated that preozonation of a natural lake water enhanced the formation of chloropicrin over nonozonation water. They confirmed the significance of nitrite in the generation chloropicrin and further noted that N₂O₅, which is present in ozone generated from air by silent electric discharge, also reacts with natural organic components of water to produce this by-product.

Uracil has been identified in river water and, since the identification of the mutagen, 5-chlorouracil, in chlorinated wastewater effluent (Jolley, 1975), the products of the reaction of nucleic bases (purines and pyrimidines) with chlorine have been of concern. Gould and Hay (1982), Gould et al. (1984a, b), and Dennis et al. (1978, 1979) have studied the reaction of several biologically important purines and pyrimidines with aqueous chlorine. Uracil reacts with hypochlorous acid to produce 5-chlorouracil as well as several other products including an N-chlorinated product (Gould et al., 1984b). Cytosine reacts to produce an unusually stable chloramine product (Gould et al., 1984a; Patton et al., 1972).

Toxicity of Concentrated Drinking Water

Little if any genetic toxicity has been found in unconcentrated drinking water (Forster et al., 1983; Harrington et al., 1983), so a number of studies have addressed concentrated drinking water samples and their subfractions. Numerous methods have been used to concentrate drinking water prior to its evaluation in mutagenicity and carcinogenicity bioassays. The methods employed most frequently utilize macroreticular resin chromatography (commercial XAD-2) and subsequent testing with the Salmonella /mammalian-microsome mutagenicity test (EPA, 1985; Nestmann et al., 1979).

The Ames Salmonella assay has been the primary source of toxicity information on drinking water samples (Cumming et al., 1983; Kool et al., 1985b; Meier and Bull, 1985). It requires only minimal amounts of material and is compatible with the broad range of solvents used to reconstitute concentrated solids or elute resin columns.

Concentrated residues from both chlorinated and untreated drinking water samples have been evaluated in the Ames test; most show some mutagenic activity, a subject that has been reviewed in depth (Loper, 1980a; Nestmann, 1983). The mutagens appear mixed between frameshift and base-pair substitution types and are, by and large, direct-acting mutagens. Some mutagenic species are quite stable. The levels found in tap water have been considered difficult to eliminate by such purifying methods as distillation, reverse osmosis, or carbon filtration (Cheh et al., 1983), although activated carbon systems have recently been used successfully (Loper et al., 1985). The mutagenicity of drinking water also appears to fluctuate in direct proportion with the organic content of the water. Water disinfection, particularly chlorination, has been shown to affect the mutagenicity of concentrated samples (Douglas et al., 1986; Loper, 1980a).

Using a model system in which aqueous solutions of organics were chlorinated, Bull et al. (1982) showed that by-products formed by chlorination of either humic or fulvic acids were mutagenic in salmonellae. This report was followed by a more detailed study of the reaction conditions required to produce the mutagens and to maintain mutagenic activity (Meier et al., 1983). Certain parallels were noted between mutagenic activity of drinking water samples and the model reaction involving chlorinated humic acids. For example, unchlorinated samples were nonmutagenic in salmonellae; and the mutagenic activity observed in chlorinated samples was higher without an extract of mammalian enzymes (S9) for metabolic activation (Meier et al., 1983).

The toxicological significance to humans of bacterial mutagens found in drinking water concentrates is not clear, and the Ames assay may best be used as a biological monitor for drinking water sources over time or to assess the consequences of various treatment procedures. While there may be epidemiological evidence supporting an association between chronic toxicity and drinking water contaminants (organics, for example), one cannot assume that biological activity in the Ames test is a reflection of the cause of the increased risk.

Lang et al. (1980) reported that organic residues from drinking water samples were able to transform BALB/C3T3 cells in culture and that the transformed cells were capable of producing tumors when transplanted to athymic mice (Kurzepa et al., 1984). This assay responds to many of the same classes of chemicals that are active in the Ames test. These results, however, add some significance to the biological activity of the organic residues in that in vitro transformation is performed with animal cells and the tumorigenic properties of the transformed cells can be verified in vivo.

Table 3-4 compares bacterial mutation with sister chromatid exchange (SCE) activity in unchlorinated and chlorinated humic acid. These data provide further evidence of the genetic toxicity of drinking water organics but also show that humic acid alone may have some biological activity. In vivo, however, chlorinated humic acid samples were not active in assays designed to detect alterations in chromosome structure or spermhead morphology (Meier and Bull, 1985).

TABLE 3-4

Dependence (pH) of Mutagenic Activity in the Ames Test and SCE Induction in CHO Cells on Treatment of Humic Acid with Chlorine (HOCl/OCl).

Rodent skin tumorigenesis studies have been used to evaluate drinking water concentrates (Kool et al., 1985b). Responses in these tests were variable but did seem to be associated with the concentrations of total organics applied. Mouse skin initiation/promotion studies also suggested that some drinking water concentrations contain tumor initiators but do not act as promoters or complete carcinogens. Subcutaneous injection and skin painting studies are considered to be reasonably reliable models for some chemical classes; but like the Ames test and in vitro cell transformation assays, this group of tests may be responding to a class of chemicals that are not particularly relevant to carcinogenic risk in humans, whose primary route of exposure is by ingestion (OSTP, 1985, p. 10414).

Another explanation is that some chemicals interfere with the expression of mutagenic properties of other agents. Numerous examples of chemical interference resulting in inhibition or elimination of mutagenicity have been reported for the Ames test.

A much more relevant in vivo approach would be lifetime exposures to drinking water. Two studies of this type have been reviewed (Kool et al., 1985b). One addressed chronic carcinogenicity in rats, employing a synthetic residue containing 11 chlorinated hydrocarbons most commonly detected in drinking water samples. The results of this study, in which the high-dose animals received 22 mg/day for 27 months, were negative. In the second study (Kool et al., 1985a), rats were exposed to drinking water concentrates obtained from commercial XAD-4/8 resins. The duration of the study was 26 months, and no increases in tumor incidence were observed. The samples used in this study were mutagenic in the Ames assay.

Toxicity of Fractionated Drinking Water Concentrates

Some drinking water concentrates produced by resin columns exhibited little or no bacterial mutagenicity until the complex residues were fractionated by HPLC. After separation, several subfractions showed activity (Cumming et al., 1983). These observations may indicate toxicity associated with unfractionated concentrates, or they may indicate chemical interference. The expression of weakly mutagenic components requiring large doses may be prevented by premature target cell cytotoxicity from other nonmutagenic components. Once the components are separated from each other by HPLC, the weak mutagens can easily be detected.

On the other hand, other research efforts have identified methods either to prevent the formation of mutagens during disinfection processes or to reduce their levels subsequent to formation. For example, using ozone instead of chlorine as a disinfectant, with fulvic acids as a model mixture, ozonated fulvic acids were found to produce only weak (Kowbel et al., 1982) or no mutagenic activity (Kowbel et al., 1984) compared with the results of Bull et al. (1982) with chlorinated fulvic acids. Depending on the dose of ozone and the pH of the reaction mixtures, preozonation of soil or water fulvic acids could partially or even totally prevent the formation of mutagens during subsequent chlorination treatment (Kowbel et al., 1984, 1986). In addition, Meier et al. (1983) showed that mutagenic activity of chlorinated humic acid can be prevented or reduced either by chlorination at alkaline pH or by raising the pH of samples following chlorination. This change in activity is probably due to the lability of the direct-acting, chlorine-substituted mutagens at alkaline pH, as also observed in drinking water samples by Loper (1980b) and in an experimental system by Nazar and Rapson (1982).

One observation derived from analysis of HPLC fractions was that disinfection processes alter the total mutagenicity of pooled HPLC fractions. This suggests that even though water concentrate samples may be mutagenic both before and after disinfection, the mutagenic components of the residue change. Some mutagenic species seem to disappear, while new ones are formed.

It is not clear whether comparisons of pooled HPLC fractions are relevant to an assessment of biological activity. If nonmutagenic compounds are capable of reducing or suppressing mutagenicity of other agents in concentrates, the combined mutagenicity obtained from pooling fractions may give misleading indications of activity. Conversely, concentrated residues are not comparable with normal water in chemical/chemical interactions. Mutagens in dilute samples may act more like the HPLC subfractions.

Erie County, New York

Carlo and Mettlin (1980) studied 4,255 cases of esophageal, stomach, colon, rectal, bladder, and pancreatic cancers reported through the New York State Tumor Registry for Erie County, New York (Buffalo and environs), between 1973 and 1976. Age-adjusted incidence rates were calculated by census tract and related to water source, level of THMs from a single survey in July 1978, and a variety of socioeconomic parameters of the tracts. Statistically positive associations were found between surface water and esophageal and pancreatic cancer and between pancreatic cancer in white males and THM levels. The authors themselves placed little credence on these findings, noting that the pancreas-THM relationship was found only in one sex-race subgroup and that only 10% of the census tracts were served by groundwater. Finally, the range of THM measurements was narrow (the largest variation was 71 ppb), and no trend data were obtained. Given only a single measurement per source, the opportunity to form meaningful associations was limited.

Massachusetts

Tuthill and Moore (1980) related cancer mortality rates for the 1969 to 1976 period in Massachusetts communities supplied by surface water to chlorination exposure data as measured by average past chlorine dose, recent total THM levels, and recent chlorine dose. Stomach and rectal cancers significantly correlated with recent THM levels and chlorine dose but not with estimates of past chlorine dose. In addition, when stepwise regression models with migration patterns and ethnic data were used, the significance of the associations between cancer rates and recent total THM and chlorine dose disappeared. The authors believed that failure to control first for social variables and then for changing patterns of chlorination over time may have led in previous studies to spurious associations of chlorination of drinking water with cancer.

Iowa

Bean et al. (1982a, b) examined age-adjusted cancer incidence rates in Iowa communities supplied by a single major source of drinking water for the period 1969-1978 and related these to the source and characteristics of the water supply after stratification for population density. In each population group, rates of lung and rectal cancers were higher in communities supplied by surface water than in communities supplied by groundwater; the risk ratio for colon cancer (1.38) was higher only in the 1975-1978 period. When communities supplied only by groundwater were included, risk ratios of male (1.32) and female (1.29) lung and female rectal (1.39) cancers were found to be higher in communities with chlorinated water, while for male rectal cancer, rates were higher in communities with nonchlorinated water.

Isacson et al. (1985) examined cancer incidence in communities of 1,000 to 10,000 inhabitants supplied by groundwater with nonchlorination-induced contamination as indicated by levels of 1, 2-dichloroethane or nickel in the finished supplies. Significantly elevated rates of colon and rectal cancers were found in residents of communities with detectable levels of 1, 2-dichloroethane and of bladder and lung cancer in residents of communities with detectable levels of nickel. The associations were independent of chlorination status. Results did not necessarily indicate a specific relationship between nickel or 1, 2-dichloroethane, but rather that these variables served as indicators of likely contamination from external sources. These results suggest that water-quality variables other than THMs may be associated with cancer.

Illinois

Brenniman et al. (1980) conducted a case-control study of gastrointestinal and genitourinary cancers in Illinois residents, excluding Cook County (Chicago). The cases were cancer deaths from 1973 to 1976; controls were noncancer deaths over the same time period. Noting that the composition of surface waters differed from groundwaters in many respects other than chlorination and THM production, the authors limited their analysis to communities served by groundwater supplies. Although elevated relative risks of chlorination were found for colon and rectal cancer, particularly in females, the authors believed that the results showed no clear associations, since little consistency appeared in the analysis by subgroups, especially in degree of urbanization. In a detailed review, Crump and Guess (1982) observed that the numbers in the Illinois study limited the power to detect significant but relatively small associations because only a dichotomous chlorination variable was used, no control for population migration was employed, and the restriction of analysis to groundwater limited the range of THM values that could be used in analysis.

Wisconsin

Young et al. (1981) and Kanarek and Young (1982) examined associations among gastrointestinal, genitourinary, brain, lung, and breast cancers in white females in Wisconsin from 1972 to 1977 by a death-certificate-based case-control design. Detailed information on past source and treatment characteristics of the community water supplies was obtained by interviewing plant operators to elicit those factors presumed to influence the organic content of the raw water. Based on these factors, estimates of by-products were constructed. Other variables included in the analysis were occupation, urbanicity, and marital status. Only colon cancer was significantly associated with the estimated chlorine dose for the past 20 years. No relative risk gradient was found according to high, medium, or low chlorine dose, but an approximate doubling of the risk (1.5 to 3.0) occurred when the analysis was restricted to chlorinated sources affected by rural runoff. This was presumed to be related to the increased THM formation that occurred when added substrate was present. Rural runoff was not evaluated as an independent risk factor, nor was population mobility assessed.

Louisiana

Gottlieb et al. (1981, 1982) compared cancer and noncancer deaths from 1960 to 1975 in Louisiana parishes selected for similarities in industrialization and approximately equal exposure of the population to surface water and groundwater. The length of time of water source exposure was estimated by relating place of birth to place of death. The study also compared cardiovascular death of controls to death of controls from all other causes. No evidence was found of bias resulting from the use of cardiovascular deaths among controls. Three types of cancer (rectum, breast, and lung) showed significant association with drinking surface water. The risk considered to be most suggestive of a causal relationship was found for rectal cancer. Elevated odds ratios were seen in both sexes, and a dose-response gradient was noted, with odds ratios of 2.50, 1.57, and 1.00 for lifetime surface, some surface, and lifetime groundwater use, respectively. Odds ratios for males were highest, increasing to over 3.0 when the lifetime water use variable was used. The association of lung cancer and surface water was statistically significant only among nonwhite males and only for water source at death. Breast cancer also showed a gradient effect, but significance was found only for white females. For all cancers, the effect of chlorination, as expected, paralleled the relationship to surface water, but for breast cancer the odds ratio increased when chlorination was considered independently. It was suggested that confounding by population density may have occurred. Kidney and liver cancers also showed elevated odds ratios, but to a lesser degree and without statistical significance.

In one of the early studies of cancer mortality in Louisiana, Page et al. (1976) used multivariate regression analysis to show an association of cancer mortality rates with drinking water obtained from the Mississippi River. (Some parishes in Louisiana, mostly in the southern part of the state, receive all or most of their drinking water from the river, while other parishes do not.) They found an apparent association between the use of water from the Mississippi River and mortality rates from all cancers, from cancers of the urinary organs, and from cancers of the gastrointestinal tract. The possible role of water disinfection was not postulated, but the investigators pointed to the high incidence of bladder cancer in New Orleans and the finding of carcinogens in water from that river.

New York

Lawrence et al. (1984) used the New York State retirement system to identify public school teachers and recorded deaths among them between 1962 and 1978. A total of 395 colon and rectal cancers in white females in the central geographic corridor of the state were identified and matched by age and year of death (2 years) with noncancer deaths from the same pool. Water source and treatment were recorded for each study subject for a 20-year period at home or work prior to death. Cumulative chloroform exposure was modeled from previous THM surveys, with significant predictor variables being prechlorine and postchlorine dose, chlorine residual, and type of water source. Calculation of odds ratios showed no associations among colon or rectal cancers and surface water or cumulative distribution of chloroform exposure after control by logistic analysis for average source type, population density, marital status, age, or year of death.

Massachusetts

Zierler et al. (1986) examined the patterns of mortality of residents of Massachusetts who died from 1969 to 1983 and lived in communities using drinking water that was disinfected by either chlorine or chloramine. There were 51,645 deaths due to selected cancers and 214,988 controls who died from cardiovascular, cerebrovascular, or pulmonary disease or from lymphatic cancer. Data were analyzed by calculating standardized mortality ratios for cancer and other diseases in residents of communities with chloraminated drinking water. Expected rates were derived from cause-specific deaths in Massachusetts and also by examining mortality ratios of selected cancer sites in comparison with mortality ratios of controls in communities with chlorinated versus chloraminated drinking water; these were termed the mortality odds ratio. Bladder cancer mortality was elevated (the mortality odds ratio was 1.7 with a 95% confidence interval of 1.3-2.2) in residents of communities with chlorinated water relative to mortality in residents of communities with chloraminated drinking water, a factor possibly related to the higher levels of THMs produced by chlorination. Also of interest was a small increase in deaths due to pneumonia and influenza among residents of communities using chloramine as their drinking water disinfectant.

North Carolina

Cragle et al. (1985) performed an incidence-based case-control study of colon cancer and water chlorination in North Carolina in which detailed personal interviews were used to collect information on pertinent risk variables, including exposure to chlorinated water through a 25-year residence history. Cases were hospitalized males and females with primary colon cancer; controls were patients with the closest admission date who matched on age, race, sex, vital status, and hospital and who had no previous history of cancer of any type, mental disorder, or major chronic intestinal disorder. The sources of water exposure were initially classified as unchlorinated groundwater, chlorinated groundwater, or chlorinated surface water, with no consideration of levels of pollutants. When it was found that chlorinated groundwater represented only a small fraction (7%), this group was lumped with chlorinated surface water. Thus, the results represent essentially a dichotomous comparison of chlorinated surface water versus nonchlorinated groundwater sources. Nonwater variables positively associated with colon cancer were genetic risk and a factor that is alcohol consumption times a high-fat diet, while smoking and number of pregnancies were negatively associated.

For reasons that are not clear, an interaction between age and chlorination was found even after adjustment for length of exposure. Above the age of 60, there was a statistically significant relationship between chlorination and colon cancer, using a logistic regression model with control of confounders. Although this effect was not seen in younger individuals, for all age groups the odds ratio was higher for persons who drank chlorinated water at their home for 16 years or more than it was in those who drank chlorinated water for 15 years or less. For persons 80 years of age or older, the odds ratio reached 3.36 in those exposed for more than 15 years. In this study no attempt was made to model the levels of THMs in the drinking water in past years, nor were current levels reported.

Wisconsin

Young et al. (in press) conducted a case-control study of colon cancer and drinking water trihalomethanes in white males and females between the ages of 35 and 90 in Wisconsin. There were 400 living colon cancer cases selected from the Wisconsin Cancer Reporting System; 600 controls came from two sources: a random selection from a statewide listing of motor vehicle operators and cancer cases other than gastrointestinal or genitourinary from the Wisconsin reporting system. Lifetime residential and drinking water source histories, diet, medical history, social class, and other life-style factors were obtained by questionnaire. Highly detailed historical data on community water source and treatment were collected, as well as data from individuals on the amount of water consumed per day. These data, together with current levels of THMs from a recent survey, allowed the construction of a model for estimating period-specific THM concentrations for the length of the study period. Logistic regression was used to estimate risks associated with THMs at 10-year periods up to 50 years before cancer diagnosis. Small risks of marginal significance were found for exposure to chlorinated water at the time of diagnosis, but no significant risk ratios were found for any other time period, for any specific relation to THMs, or for any specific age groupings. This held true when each control group was analyzed separately.

These results were of particular interest in view of the earlier results, discussed above, of a case-control mortality study conducted by the same investigators in the same general population of Wisconsin, in which a significant positive association was found between colon cancer and sources affected by rural runoff. The reasons for the difference in outcome are not definitively known, but the second Wisconsin study comes as close to meeting ideal methodological criteria for cancer-THM associations as any study yet presented. This suggests that design differences—specifically, the inclusion of residential mobility in the latter study—could have been responsible. As noted by the authors themselves, however, another possible explanation could be the generally low levels of THMs found in Wisconsin surface waters, which would limit the ability to detect significant differences.

National Bladder Cancer Study

In 1979 the National Cancer Institute launched a nationwide collaborative study of the relationship between bladder cancer and the use of artificial sweeteners. Cantor et al. (1985) were able additionally to analyze the effects of the chlorination of drinking water on bladder cancer. The drinking water regions were metropolitan Atlanta, Detroit, New Orleans, San Francisco, and Seattle and the states of Connecticut, Iowa, New Jersey, New Mexico, and Utah. All had population-based cancer incidence registries from which live cases were selected. Controls were randomly picked from the general population and frequency matched to cases by sex, 5-year age group, and geographic area. Detailed information on geographic mobility and water source was collected, as well as information on other pertinent variables. In a separate data collection, water utilities serving more than 1,000 persons were surveyed, and information on source, chlorination, and protection of watershed was noted.

Risk of bladder cancer among white respondents was examined in logistic regression models that included age, smoking of cigarettes, sex, study area, and usual employment as a farmer. These initial analyses were restricted to the 1,244 cases and 2,550 controls who were never employed in a high-risk occupation for bladder cancer and whose residential water was supplied from either a nonchlorinated ground source or a chlorinated surface source for at least 50% of their lifetime. When risk of bladder cancer was evaluated by the usual use of chlorinated surface source, as compared with the usual use of nonchlorinated groundwater, there was no overall association. However, among nonsmokers, a group generally at low risk for bladder cancer, those whose usual source was of chlorinated surface origin had an odds ratio of 1.4, relative to usual users of nonchlorinated groundwater, and there was a relationship between risk and duration of chlorinated surface water use. Among nonsmokers, the relative risk increased from 1.3 among users of chlorinated surface water for less than 20 years to 2.3 among those who used chlorinated surface water for 60 or more years.

When relative risks by duration of chlorinated surface water use were examined by reporting area, risk appeared to be significantly higher in the relatively rural areas (Iowa, Utah, and New Mexico) than in the metropolitan areas. No reasons for this difference could be positively identified, but it was noted that high levels of chlorination by-products are present in many community water supplies serving towns in farming areas. An elevated risk could also potentially be related to the use of agricultural chemicals or, conversely, to unknown, and therefore uncontrolled, independent causal variables in the urban areas that masked a chlorination effect.

Further analyses of the NCI data set (Kenneth Cantor, National Cancer Institute, Bethesda, Maryland, personal communication, 1986) have revealed positive associations of bladder cancer risk with level of tap water ingestion and duration of exposure, predominantly among study subjects with long-term residence in communities served by chlorinated surface waters.

Chloramine Analysis

The analysis of chloramines in natural water samples has been of two types. The most widely used methods are oxidant or chlorine residual measurements. Chloramines are strong oxidants that, like chlorine, can oxidize iodide to iodine. The measurement of iodine, or iodometry, is a classical method of analysis, although total oxidant methods such as iodometry are notoriously susceptible to interferences. The oxidation of iodide to iodine is relatively easy; the standard oxidation potential of the couple is -0.54 V. Thus, most oxidizing agents such as manganese (IV) (Strupler and Rouault, 1979), hydroperoxides, and at least some N-chloroorganic compounds (Gray and Workman, 1983) are capable of making iodine under the conditions used to measure monochloramine (NH₂Cl). The compounds measured as chloramines, therefore, include a wide variety of oxidants that may contain no chlorine.

Thus, all the common chlorine residual measurements are relatively nonspecific or nonselective for the compound that it is desirable to measure. The most selective methods include the free-chlorine procedures, such as FACTS and amperometric titration without the addition of iodide. The least selective methods are the total chlorine residual techniques. The latter methods include nearly all of the oxidants because they use either high concentrations of iodide (e.g., the N, N-diethyl-p-phenylenediamine ferrous ammonium sulfate [DPD-FAS] total chlorine method) or low pH (e.g., amperometric titration for dichloramine at pH 4) (APHA, 1985). The second and more selective type of method measures the chloramine compounds after a separation process. These methods include the amperometric membrane electrode (Stanley and Nossel, 1983) and chromatographic methods (Kearney and Sansone, 1985; Scully et al., 1984). Although less precise, these methods are less subject to interferences than the iodometric methods.

Organic Nitrogen Compounds

In all methods of chloramination, the generation of the disinfectant relies on the fact that ammonia reacts rapidly with hypochlorite to produce monochloramine (Morris, 1967):

Most organic amines and amino acids, however, react even more rapidly with hypochlorite to form organic N-chloramines (Morris, 1967; Weil and Morris, 1949). In water containing both ammonia and organic amino-nitrogen compounds, the relative amounts of organic (versus inorganic) chloramines formed when the water is chlorinated depend on the concentration ratios of ammonia to organic amino-nitrogen, the temperature, the pH, and the relative reaction rates (Isaac and Morris, 1980). However, Isaac and Morris (1980) have explained that chlorination of water containing 20 mg ammonia nitrogen per liter and 2 mg organic amino-nitrogen per liter will form 54% inorganic chloramine and 46% organic chloramine within 0.3 seconds at pH 7 and 20°C if the relative specific rates of reaction are 1:8.5 (NH₃ to organic amino-nitrogen).

In Volume 2 of Drinking Water and Health (NRC, 1980), some of the problems with the analysis of free and combined residual chlorine were discussed briefly. However, since that time, considerable uncertainties surrounding the interpretation of conventional measurements of free and combined residual chlorine have been pointed out (Cooper et al., 1982; Gould, 1986; Johnson, 1978; Jolley and Carpenter, 1983; Ram and Malley, 1984; Scully, 1986; Wajon and Morris, 1980; Wolfe and Olson, 1985; Wolfe et al., 1984, 1985). Both organic and inorganic N-chloramines respond in an identical manner to conventional methods of analysis (APHA, 1985) because both oxidize iodide to iodine in the determination of ''combined residual'' chlorine.

As a result, the breakpoint curve of water containing both ammonia and organic amines is a composite of the individual breakpoint curves of ammonia and every organic amino-nitrogen compound in the water that can react with hypochlorite. Figure 3-6 illustrates this using a recently reported method for the derivatization and analysis of organic and inorganic N-chloramines in dilute aqueous solution (Scully et al., 1984). Equimolar solutions of glycine and ammonia (4 mg of total nitrogen/liter) were chlorinated to different levels along the breakpoint curve. Figure 3-6 plots the relative amounts of chloramine derivatives recovered along with the total residual chlorine measured by the DPD-ferrous ammonium sulfate (DPD-FAS) method (APHA, 1985). The plot demonstrates how N-chloroglycine is formed to a greater extent than NH₂Cl at low chlorine dosages.

Figure 3-6

N-Chloroglycine () and NH₂Cl () recovered after derivatization of dilute aqueous solutions of glycine and ammonium sulfate in 0.01 M phosphate buffer (pH 7.2) that had been chlorinated to different levels. After chlorination, each solution was incubated (more...)

From a water treatment standpoint, organic N-chloramines are undesirable because they are not effective disinfectants (Feng, 1966; Johnson, 1978; Marks and Strandskov, 1950; Wolfe et al., 1984, 1985). Consequently, a water treatment facility that practices marginal chlorination of water containing high concentrations of organic amino-nitrogen compounds runs the risk of overestimating the ability of its systems to maintain adequate disinfection.

The implications of this have been demonstrated by Wolfe et al. (1984, 1985). Using water from the San Joaquin Reservoir, they examined the effect of added glycine on the reduction of total count bacteria after chlorination or chloramination. Total count bacteria were reduced by 2 log units within 60 minutes when preammoniated water was chlorinated to a chlorine-to-nitrogen ratio of 3:1 by weight. However, when increasing amounts of glycine (0.1, 0.25, and 0.55 mg/liter) were added to the ammoniated samples before they were chlorinated to the same residual as in the initial experiment, inactivation of the bacteria was significantly inhibited to an extent proportional to the concentration of the glycine added (see Figure 3-7). Nevertheless, both amperometric titration and DPD-FAS determination of the "combined residual" chlorine suggested that all solutions had equivalent bactericidal capabilities. These results could only be explained by the competition between glycine and ammonia for reaction with chlorine and formation of the less-bactericidal N-chloroglycine. By contrast, preformed inorganic monochloramine was an effective disinfectant when added to water whether or not it contained glycine.

Figure 3-7

Inactivation of total count bacteria in a San Joaquin Reservoir sample using preammoniation and prereacted application techniques. Nitrogen (as glycine) was added to the samples at levels of 0.1, 0.25, and 0.55 mg/ liter prior to preammoniation treatment. (more...)

Organic N-chloramines can also form slowly by the reaction of inorganic chloramine with organic amines (Isaac and Morris, 1983, 1985; Snyder and Margerum, 1982):

However, because the chlorine transfer reaction is slow, its significance may be limited to water distribution systems that use inorganic chloramine as the disinfectant when detention time in the system is considerable.

Several studies (Cooper et al., 1982; Wajon and Morris, 1980; White et al., 1983) have shown that a "false" free residual can be obtained by conventional methods of analysis in the presence of a number of organic chloramine compounds. White, for instance, failed to obtain adequate disinfection of wastewater that contained low concentrations of ammonia and significant concentrations of organic nitrogen. The effluent showed an apparent free residual chlorine level that should have been sufficient.

Ram and Malley (1984), on the other hand, examined the disinfecting ability of a number of model organic chloramino-nitrogen compounds that produce a free-chlorine residual. All appeared to exhibit bactericidal effectiveness toward E. coli when the bacterial cultures were inoculated so that an apparent free residual of 0.2 mg/liter of chlorine was maintained after 15 minutes.

Although interference of disinfection by organic nitrogen compounds can be demonstrated in laboratory experiments and these used to implicate interferences in treatment plants, there is still a poor understanding of the specific compounds responsible for interferences and their true significance in actual treatment practice. The studies discussed here suggest that conventional methods of chemical analysis of residual chlorine tend to overestimate the effectiveness of disinfection. Until these processes are better understood, an awareness of such potential interferences is needed in the water treatment industry.

Use Patterns of Ozone and Chlorine Dioxide

Concern over by-products of chlorine has caused municipal water authorities to consider alternatives for disinfection and oxidation of drinking water. As a result, use of chlorine dioxide and ozone is on the increase in the United States, and many more utilities are considering these alternatives to chlorine.

New research has shown that ozone has the property of improving coagulation (flocculation is a more effective disinfectant for resistant pathogens) and can control taste and odor compounds and manganese at least as well as chlorine. The city of Los Angeles (Department of Water and Power) is currently building a 600-million-gallon per day (26 m³/sec) direct-filtration plant with 1 mg of ozone/liter of water for pretreatment, making it the site of largest ozone use in water treatment in the world. This plant has also given the industry a new standard for the cost of ozone in a large-scale plant, one that is at least 20% lower than projected costs only 5 years ago. In summary, all indicators point to the increased use of ozone in water treatment.

How extensive this adoption of ozone and chlorine dioxide technology will be is not yet clear. However, it is clear that we know very little about the potential impact of these disinfectants (oxidants) if they are used in place of chlorine. This section does not focus on problems of engineering, thought to be particularly challenging for ozone, or the problems of disinfection efficacy. Rather, we emphasize the lack of information on by-product formation.