Click for next page ( 191


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 190
5 Conclusions and Recommendations Dl Sl N FECTION M ETHODS AN D EFFICACY Waterborne disease outbreaks continue to occur, not only in developing nations, but also in the United States, where almost 70,000 cases were reported in 235 disease outbreaks during the period 1978-1983. The etiol- ogy of waterborne disease in the United States has changed dramatically since the early 1900s. Recent outbreaks have generally been dominated by gastrointestinal illness associated with viruses and protozoan cysts. These pathogens are generally more resistant to disinfection than the kinds of pathogenic bacteria that formerly caused most outbreaks. Problems also continue to occur in cases of consumption of untreated water, errors of insufficient or interrupted disinfection, failures to maintain adequate levels of residual disinfectant in potable water distribution systems, and breaches in the systems. Although chlorination continues to be the predominant method of drink- ing water disinfection in the United States, the use of alternative methods is increasing. Treatment facilities in several states have recently increased and/or switched to chloramination for primary disinfection, largely in response to the maximum contaminant level of 100 log of total trihalo- methanes per liter set by the Environmental Protection Agency under the Safe Drinking Water Act. Kansas, which formerly prohibited chlorami- nation, now requires the use of ammonia to convert all free chlorine residual to chloramines following 30 minutes of chlorination. The Met- ropolitan Water District of Southern California now uses chloramination for distribution system residual maintenance. Other alternatives, such as 190

OCR for page 190
Conclusions and Recommendations 19 ~ ozone disinfection, are being used increasingly in Europe and Canada. Potentially substantial increases in ozone use in the United States also appear possible in view of recent improvements in the reliability and efficiency of ozone disinfection technology, together with the higher ef- ficacy of ozone (compared with chlorine) against resistant protozoan cysts and viruses. Research is needed to improve understanding of the comparative effi- cacy of major drinking water disinfection practices (especially chlorina- tion, chloramination, ozone, and chlorine dioxide) against the currently most important, resistant protozoan cysts and viruses. Studies of the major factors affecting such efficacy under treatment plant operating conditions are also important. Without such studies, it is possible that many future drinking water treatment operations, decisions about alternative methods, and trade-offs regarding toxic by-products of chlorination may be inor- dinately influenced by the preponderance of existing knowledge about types of bacterial pathogens that pose less of a public health problem in the United States today than several decades ago. "Life-cycle" studies of disinfectants are also needed for comprehensive examination of the direct and indirect implications of potential widespread local conversions to alternative disinfection practices. For example, the reliability of ozone disinfection technology and the tropospheric impact of potentially large increases in ozone generation that might result from widespread application should be investigated more thoroughly, before national and local decisions regarding conversion become a matter of fact. CH EM ISTRY AN D TOXICITY OF DIS I N FECTION Reactions and By-Products Studies of chlorination of model compounds and isolated humic and fulvic acid precursors during the past few years have improved under- standing of the reaction mechanisms and types of by-products formed during chlorine disinfection of drinking water. Although many of the specific by-products are not yet well characterized, they appear to vary according to the structures of the humic and fulvic acid molecules being chlorinated, the chlorine-to-carbon (CI/C) ratio, the pH, the time of re- action, and other factors. The principal by-products, especially at high Cl/C molar ratios of 3:1 or 4:1, are volatile, hydrophobic compounds (mainly chloroform). However, a large variety of nonvolatile, hydrophilic compounds are also produced. These nonvolatile products include both chlorinated and unchlorinated aromatic and aliphatic compounds. The production of these smaller, hydrophilic molecules appears to increase at lower pH and Cl/C molar ratios (less than 1:1), while higher ratios favor the formation of chloroform and other volatile by-products. At the lower

OCR for page 190
192 DRINKING WATER AND HEALTH CT/C ratios, which more closely approximate typical drinking water dis- infection conditions, the humic acid precursors appear to support the formation of unchlorinated by-products (e.g., monobasic and dibasic al- iphatic acids) to a greater extent than do fulvic acid precursors. Increasing the Cl/C ratio appears to drive both types of precursors toward chloroform production and a larger fraction of identifiable products, which neverthe- less represent only a small fraction of the initial organic material. It is clear from the studies described in Chapter 4 that the importance of trihalomethanes may be overestimated from experiments involving Cl/ C ratios and chlorination times that greatly exceed typical drinking water disinfection conditions. Nonvolatile by-products of humic and fulvic acid chlorination may be more important than previously believed. Further studies of reaction mechanisms, controlling factors, and by-product iden- tification are needed. Improved methods for characterizing the nonvolatile products are also needed to support such studies. Toxicity Both chlorinated and untreated drinking water contain genotoxic com- pounds identified in concentrated residues by short-term assays. Chlori- nation tends to destroy some of these compounds, as well as produce new ones. Short-term animal skin tests, although not conclusive, provide in- dications that organic concentrates from chlorinated water are tumorigenic under some experimental conditions. Studies by routes other than dermal application have not shown such an effect. Based on the available data, the recommended SNARLs and risk assessments developed by the com- mittee are shown in Table 5-1. One finding common to most studies performed throughout the world is that chlorination as a means of disinfection introduces mutagens that are not present, or that are present in lower amounts, in raw, untreated water. Recent studies indicate that most of the mutagenic activity in treated water may be due to nonvolatile (rather than volatile) compounds that are produced from chlorination of humic and fulvic acids. The upper 95% lifetime cancer risk for humans based on drinking water studies in laboratory animals show a risk of 8.9 x 10-8, or approximately 1 chance in 10 million of cancer for the consumption of 1 ~g/liter of chloroform in water. This and other risk assessment calculations are shown in the section on trihalomethanes. The committee in reviewing the results of these calculations recommends that the current level of total trihalo- methanes (THMs), regulated at 100 ~g/liter in finished drinking water, be reduced. This level is unsupportable on the basis of the risk values for chloroform developed in this review, noting that chloroform is the principal THM. Nonetheless, the committee is concerned about the toxicity of the

OCR for page 190
Conclusions and Recommendations 193 TABLE 5-1 Summation of Suggested No-Adverse-Response Levels (SNARLs) and Risk Estimates for Chemicals Reviewed in This Volume Estimated SNARLs Upper 95% Confidence Estimate of Lifetime CompoundAdultChild Cancer Risk Disinfectants Chlorineaa Ozoneaa Chlorine dioxide210 ~g/liter60 ~g/liter Chloramine581 ~g/liter166 ~g/liter Disinfectant by-products Chlorate24 ~g/liter7 ~g/liter Chlorite24 ~g/liter7 ~g/liter Trihalomethanes Chloroform 8.9 x 10-8e 1.9 x 10-6 Chlorodibromomethane 8.3 x 10 7 Haloacids Dichloroacetic acid420 ~g/liter175 ~g/liter Trichloroacetic acid120 ~g/liter50 ~g/liter Haloaldehydes Chloroacetaldehydebb Dichloroacetaldehydebb Trichloroacetaldehydebb Haloketones 1, 1,1-Trichloroacetonebb 1,1,3,3-Tetrachloroacetonebb Hexachloroacetonebb Haloacetonitriles Dichloroacetonitrile56 ~g/literC Dibromoacetonitrile161 ~g/liter23 ~g/liter Bromochloroacetonitrilebb Trichloroacetonitrilebb Chloropicrin40 ~g/liter12 ~g/liter Chlorophenols 2,4-Dichlorophenol7,000 ~g/liter2,000 ~g/liter 2,4,6-Trichlorophenolbb 2-Hydroxylchlorophenolbb aNot calculated. bInsufficient data for calculation. CNot calculated; the adult value was calculated for comparison purposes; it is not recommended by the committee. ~Tumor data for risk assessment calculation from drinking water animal study. eTumor data for risk assessment calculation from corn oil gavage animal study.

OCR for page 190
194 DRINKING WATER AND HEALTH other individual by-products produced by the reactions of alternative chem- ical disinfectants in common use in water supplies because their toxicity is essentially unknown and their potential health impacts cannot be ade- quately assessed. There is a larger risk associated with the unidentifiable by-products of water disinfection, especially with chlorine. The magnitude of this risk is not quantifiable by the studies done to date, but it is high enough to warrant additional effort to determine its qualitative source and quantitative magnitude. Associated with this effort, methods should be sought to follow the risk associated with chlorination by-products even in the absence of individually quantifiable compound risks. Correlation of health risks with surrogate parameters is a classic method, but unfortunately the parameters measured to date, those of THMs and total organic halogen, do not appear to correlate well. Also many of the risk studies with humic material, the principal source of precursor, have been done with commercial humic acid. It is recommended that future studies focus on humic material con- centrated from aquatic sources. A large variety of such materials has been collected by the USCS laboratory in Denver, Colorado, under the auspices of the International Humic Substances Society, and methods have been developed for the concentration of large quantities of aquatic humic ma- terials from waters rich in such materials. Chloramination is becoming widely practiced as a method of disinfec- tion because of the regulation of THM levels and the low capacity of chloramine to form THM. Because it is a much weaker disinfectant than chlorine, chloramine must be used at higher concentrations and for longer periods of contact to achieve sufficient disinfection. Even with extended contact time and higher concentrations, however, chloramination is not recommended as a primary disinfectant, especially where virus or parasitic cyst contamination is potentially present. Preformed monochloramine is undesirable as a primary disinfectant. The use of marginal chlorination as a method of introducing chloramines into a water supply system is spe- cifically not recommended because, along with the depletion of chlorine to produce inorganic monochloramine, organic chloramines that have even lower efficacy as disinfectants are formed. Organic chloramines have also been implicated as major contributors to the mutagenicity of chlorinated drinking and natural waters. For these reasons, chloramine treatment is used to minimize by-product formation. When free chlorine is used as the primary disinfectant, an amount should be used that is sufficient to produce a slight residual of free chlorine above that required to oxidize nitrogen, followed by the addition of ammonia to form monochloramine and limit THM formation. Chloramine, as inorganic monochloramine, is only weakly mutagenic. Organic concentrates from chloraminated water exhibited half the muta

OCR for page 190
Conclusions and Recommendations 195 genie response of those from chlorinated water. Even though it does not form significant levels of THM, chloramine is nevertheless capable of producing halogen substitution to form organic compounds, and thus may produce important levels of total organic halogen (TOX). Little is known about the oxidant residuals formed in drinking water because of the general lack of accuracy in measuring them. There is also a great need to define more accurately the nature of the oxidant residuals referred to as total chlorine, by which chloramines are measured by dif- ference. The determination of adequate water disinfection has traditionally relied on accurate measurement of the concentration of residual disinfec- tant. There are currently no suitable methods for fully quantifying the organic chloramine fraction in the presence of inorganic monochloramine. Until such methods are developed, utilities that handle water supplies containing high concentrations of organic nitrogen run the risk of over- estimating the ability of their systems to maintain adequate disinfection. Additional work is required on organic base precursor fractions, pri- marily on the organic nitrogen precursors. Neither the toxicity nor products formed from these precursors have been well documented. This is espe- cially true for the organic chloramine portion of the chlorine residual. These compounds appear to be candidates for significant health concern because of their potential to interfere with judgments made on the basis of chlorine residual as to the degree of disinfection achieved, as well as their potential for increasing mutagenicity. When possible, organic precursors should be removed prior to the dis- infection process. This can be achieved by changing the order of the procedures of conventional treatment. A better approach, however, is to improve specific conventional water treatment processes to remove organic compounds and to add processes such as carbon absorption and preoxi- dation. Initial removal of organic by-product precursors precludes the need for reducing contact time, thus improving the efficacy of the disinfection processes and minimizing formation of organic chlorine by-products. The use of alternative oxidants, especially ozone and chlorine dioxide, will increase in the United States in the coming decades. Little is known about the types of by-products produced by ozonation of natural organics. Well-conceived studies need to be conducted that will focus on the stable compounds expected from ozone reactions with humic material. But these studies must recognize that the analytical methods used for nonpolar chloroorganics (extraction, GC/MS) may not be successful for the more polar, more labile compounds expected from ozonation processes. Par- ticular attention should be given to the search for unsaturated aldehydes and the hydroxy-hydroperoxides. Following these studies, further health effect studies are needed to determine whether ozone by-products are mutagenic or carcinogenic or

OCR for page 190
196 DRINKING WATER AND HEALTH produce other adverse effects. These studies should take into account variations that are likely to occur when the oxidation process is carried out in different matrices (pH, O3/TOC ratio, alkalinity). Notwithstanding the fact that these studies need to be carried out, drink- ing water suppliers should not dismiss the possibility of using ozone as an alternative to chlorine and chloramines in water treatment. Ozone is an excellent disinfectant (although it must be used in combination with a secondary disinfectant to maintain a residual in the distribution system); ozone is an excellent oxidant for the various needs of water treatment; it does not form chlorinated by-products; and the admittedly inadequate studies now available point to lower toxicities of ozonated water than of chlorinated water. EPIDEMIOLOGY The studies reviewed in this volume present progressive improvement in design and in suggestive evidence that by-products induced by chlo- rination, or some other water parameter, may be causally related to some internal cancers of the epithelial tissue of the digestive organs and the lower urinary tract. Confidence in the demonstration of causal relationships would be increased if well-designed studies could be replicated in other populations. There have been many differences in research design among studies by various investigators. These differences can greatly influence observed risk ratios. No epidemiological study has measured actual levels of THMs or other potentially carcinogenic materials over periods of time in drinking water. Many have relied on dichotomous coding of chlorination as a yes- or-no variable. Few have considered population migration. It is possible that true geographic differences are involved in exposure to chlorination by-products, with rural areas having generally higher lev- els. The National Bladder Cancer Study (see Chapter 3) demonstrated distinct geographic variations in risk with the use of a common method- ology. Geographic differences in available substrate may lead to varying amounts of chlorination by-products, or there may be other carcinogens, such as pesticides, in rural drinking water. This latter factor may be important for two reasons: (1) the presence of other causal factors in the drinking water that are randomly distributed with respect to chlorination but unaccounted for in the analysis would produce a lower observed risk ratio than the true risk ratio for chlorination; and (2) other unknown causal variables whose distribution parallels that of chlorination could act as classical confounders. Surface waters are the most frequently chlorinated source of drinking water, followed by shallow alluvial wells; the least frequently chlorinated sources are deep wells. The same distributional

OCR for page 190
Conclusions and Recommendations 197 pattern would be expected for many runoff contaminants. A major fault of virtually all previous studies has been the failure to obtain exposure data on carcinogens unrelated to chlorination in finished water supplies. In addition, by-products of chlorination other than the trihalomethanes may be of importance in the genesis of human cancer. Even in the studies showing statistically significant risk ratios, the mag- nitude of the ratios has been relatively small but of major public health importance. For example, if the risk ratio for rectal cancer associated with drinking chlorinated water is 1.5 and 50% of the persons in the United States drink chlorinated water, about 6,400 new cases of rectal cancer might be caused each year by chlorinated drinking water. Given the fre- quency of water consumption, even a small excess risk (less than 10%) may account for a lot of disease. The entire issue of drinking water and cancer deserves continued investigation. The following are recommendations of approaches to the assessment of human health risks to water disinfectants. One role of epidemiological studies is to provide qualitative descriptions of the relation of disease to the various methods of disinfection. Current evidence suggests that the relative risk to human health from exposure to disinfectants is probably small in settings with effective control of treatment levels in distribution systems. Routine monitoring of disease occurrence in populations with different treatment methods may provide crude qualitative descriptions of the patterns of disease in populations using different treatment strategies. The expense of such studies can be kept relatively low, as they can use existing data bases maintained by state health departments, cancer regis- tries, and lists of residents. A major limitation to this approach is that it is difficult to control for important confounders, such as smoking history. The recommended design for descriptive studies, aimed toward gen- erating hypotheses about the health effects of drinking water disinfectants, is a retrospective cohort study. The health outcomes typically would be prevalent cases of disease (mortality is an exception), and exposure clas- sification would be the type of disinfectant in drinking water in the resi- dence at the time of ascertainment of health status (for example, if cause of death is the outcome of interest, then residence at death would determine the exposure classification). Public health department statistics routinely include information on some important covariates of disease that are po- tential confounders, such as socioeconomic status, last occupation, race, age, and gender. These qualitative descriptions would be based on a range of point estimates (usually rate ratios of standardized mortality ratios) that offer evidence about the nature of an association. There is often little quantitative meaning to these estimates. To suggest that rate ratios ranging from 1.13 to 1.39, which represent low to high doses of chlorine, re- spectively, are evidence of a dose-response relationship of chlorine to a

OCR for page 190
i98 DRINKING WATER AND HEALTH particular cancer is an inappropriate interpretation of results based on crude and indirect measures of human exposure. The implication of such crude analyses is that one must report results that are consistent with the accuracy of the exposure measures. For descriptive studies such as these, estimates of effect (e.g., rate ratios) should not be reported beyond two significant digits. Point estimates would best be described within a range of confidence intervals, rather than a p-value, since it is easy to achieve "statistical significance" levels when there is a lot of information. Statistical differ- ences may not have bearing on biological differences. After studies have been undertaken to generate hypotheses about the health effects of water disinfectants, the role for epidemiological studies is to quantify associations when data are available. For these studies, data on confounding factors and accurate historical information on the duration and type of disinfectant exposures should be required. Information on the timing of the onset of disease should also be confirmed if there truly were a health hazard associated with a particular disinfectant. Design options for these analytical studies depend on the frequency of occurrence of the health effects of interest and the true relationship of a particular disinfectant to the occurrence of these health effects. One ap- proach is a prospective cohort study, in which members of a healthy population are chosen as samples according to the type of disinfectant in their drinking water supply. Over time, the cohort is monitored for disease occurrence, and at the termination of follow-up, the rates of disease are compared. The advantages of this design are (1) direct measures of in- dividual water consumption can be obtained; (2) by-products can be mea- sured from periodic tap samples; (3) incident cases of disease are measured (as opposed to prevalent cases); and (4) a variety of outcomes can be investigated. The major disadvantages of this design are the expense and administrative tedium in carrying out these studies. However, if large cohorts have already been assembled for other investigations, then it may be feasible to add a component to the study objectives to investigate potential adverse human health effects of water disinfectants. Usually a prospective cohort design is not feasible, particularly if the etiologic period is many years or decades, as it may well be for the chemicals under discussion, and if the disease occurrence due to disinfection by-products is rare. The occasion of a change in water disinfection treatment is an ideal time to begin such a study. A more suitable approach is to sample people according to their health status and historically assess their usual source of drinking water. This design is a case-control sampling strategy. The advantages to this design are (1) a large series can be identified; (2) the cost, usually, is lower than a prospective sample; and (3) if data are available on lifetime drinking water history, it may be possible to identify the etiologic period during

OCR for page 190
Conclusions and Recommendations 199 which these by-products have an effect on human health. To this end, both the age of exposure and duration, as well as the lag time from exposure to disease onset, could be investigated. The disadvantages are (1) infor- mation on the duration and type of disinfectant is difficult to obtain, and direct measures of chemical levels are not possible unless routine testing of water has occurred in the past; (2) the study cost can increase dra- matically if data collection includes personal interviews to learn about exposure to potential confounders. Despite these advantages, case-control designs are the most effective in meeting the objective of quantifying the risk of a particular disinfectant strategy on human health. Incident cases of disease can be identified from existing registries (e.g., cancer or birth defects records) or pathology logs. A third role for epidemiological studies of the health effects of drinking water disinfectant methods is that these studies can contribute to experi- mental models of chemical carcinogenesis, atherogenesis, and teratoge- nesis. As noted above in the list of advantages of case-control studies, it is possible, although difficult, to identify the relevant etiologic period when exposure has the greatest impact on disease occurrence. Usual life- time exposure will dilute the measure of effect, if there truly is an adverse effect of a particular treatment strategy. In addition, exposure measured only at the time of disease onset or death may not accurately reflect the relevant exposures that occurred or began to occur 10, 20, or 30 years earlier, during childhood or early adulthood. Finally, a fourth contribution is to describe the effect of disinfectant by-products in conjunction with other risk factors for disease. For example, epidemiological studies can evaluate interaction between smoking and drinking water or concurrent chemical contamination and disinfectant con- tent on the risk to human health. Epidemiological studies undertaken to learn about effect modification require stratified analysis or a sampling approach that selects individuals according to their exposure to a particular covariate so that the distribution of exposure is balanced for efficient contrasts of the disease rates according to categories of the covariates. For example, if the interest were to study the interaction of smoking and chlorination on bladder cancer risk, then the study population should be selected expressly in terms of their smoking experience. A 50% split between current smokers and nonsmokers would provide an efficient ex- perience to examine the additional risk of chlorination in the presence of smoking relative to chlorination exposure in the absence of smoking and no exposure to either smoking or chlorination. Clearly there are limitations to epidemiological research in clarifying the risks of environmental exposures to human health. Human studies are unlikely to contribute knowledge about dose effects because the levels of exposure to hazardous by-products of chlorination hardly vary. In addition,

OCR for page 190
200 DRINKING WATER AND HEALTH the weaker the effect of a particular chemical on disease occurrence, the more difficult (in terms of measurement accuracy, cost, and administration of the study) it is to detect this association in an epidemiological study. In light of current knowledge, it seems probable that the by-products of concern incur a relatively low risk on human health, regardless of the treatment strategy. One would therefore want to be particularly conscious of locating a study base that is potentially informative about the relation under study. In principle, the population experience should be quite ho- mogenous with respect to covariates of the diseases of interest and, in addition, at low risk for the same diseases so that a small excess risk could be detected. Four potential contributions from epidemiological observational studies have been discussed: (1) descriptive research, primarily aimed at gener- ating hypotheses; (2) analytic research, aimed at quantifying risk with proper control for confounding; (3) evidence for improving models of chemical carcinogenesis, teratogenesis, and atherogenesis; and (4) effect modification by covariates of disease, such as smoking and occupation. Results from epidemiological studies that are properly conceptualized and employ valid methodologies can offer important information for policy- makers on disinfectant treatment strategies.