Chlorination of drinking water that contains organic constituents results in the formation of halogenated chemical by-products, some of which are trihalomethanes (THMs). 1 Types of THMs are chloroform, bromodichloromethane, dibromochloromethane and bromoform. 2 Since the 1970s, there have been concerns about potential health effects of THMs and other chlorination by-products in municipal water supplies. 3 These concerns have prompted a number of investigations to determine whether infants whose mothers were periconceptionally exposed to drinking water chlorination by-products are at greater risk of developing congenital malformations. 4–12
Investigations have been conducted on specific congenital malformations, such as central nervous system anomalies, neural tube defects (NTDs), oral clefts, respiratory defects, urinary tract defects and some defects of heart development. Twofold or greater risks of these congenital malformations associated with THMs (or surrogate measures of THMs) have been observed in some, 5,6,8–10,12 but not all, studies. 4,7,11 This inconsistency in results may reflect differences in study population and methods. Such differences could include variations in completeness of case ascertainment, specificity of exposure assessment and completeness of information on covariates, or case phenotypes. To overcome some of the limitations of previous studies, we investigated the potential risks of specific phenotypes in three groups of congenital malformations that have been associated with drinking water containing THMs: NTDs, orofacial clefts and selected heart defects known as conotruncal defects. Our investigation uses data from two large case-control maternal interview studies. We also investigated a suggested gene-environment interaction mechanism involving a folate-related gene in combination with THM exposure. 13
The two case-control studies analyzed here have been described elsewhere. 14–16 The study population for study 1 16 consisted of the cohort of livebirths and fetal deaths (≥20 weeks gestation) among women who were residents of 55 of 58 California counties between June 1989 and May 1991 (N = 708,129). Cases were singleton liveborn infants, fetal deaths, or fetuses that were electively terminated after a prenatal diagnosis of NTDs (anencephaly, spina bifida cystica, craniorachischisis, or iniencephaly). Cases were ascertained from medical records, including ultrasonography, at all hospitals and clinics. Infants and fetuses with NTDs were further classified into the two most common subphenotypes, anencephaly and spina bifida. Infants used as controls were randomly selected from each area hospital in proportion to the hospital’s estimated contribution to the total population of singleton infants born alive in a given month from June 1989 to May 1991. This process identified 644 liveborn singleton infants who had no congenital anomaly. Women who did not speak English or Spanish or who had a previous NTD-affected pregnancy were not eligible.
The study population for study 2 14,15 consisted of all deliveries (infants or fetal deaths) that occurred among California residents between January 1987 and December 1988 (N = 344,214). Cases included infants and fetuses with orofacial clefts, conotruncal heart defects, or NTDs (not included in study 1) who were diagnosed within 1 year after birth among infants and fetal deaths delivered to women residing in 46 of 58 California counties. Eligible as orofacial cleft cases were those infants and fetuses with cleft palate alone or cleft lip with or without cleft palate, as confirmed by surgical or autopsy report. Based on the presence of accompanying congenital anomalies, these cleft cases were further phenotypically classified as either isolated or multiple. Infants and fetuses who had no other anomaly (or whose other anomalies were considered minor) were classified as isolated. Infants and fetuses who had at least one other major anomaly were classified as multiple. Infants and fetuses who had conditions associated with a single gene defect were excluded.
All infants and fetuses with anomalies affecting aorticopulmonary septation (confirmed by echocardiography, cardiac catheterization, surgery, or autopsy) were classified as having conotruncal heart defects. Infants or fetuses (including elective terminations) with diagnoses of NTDs confirmed by autopsy, surgery report, ultrasound, or x-ray scan were classified as having NTDs. As in study 1, NTD cases were subclassified as anencephaly or spina bifida. Control infants (infants with no congenital anomaly; N = 652) were randomly selected from all infants born alive in the same geographic area and time period as cases.
In study 1, in-person interviews were completed with mothers of 538 cases (88% of those eligible) and of 539 controls (88%). These interviews were conducted an average of 4.9 months (cases) and 4.6 months (controls) after the actual or projected date of term delivery. In study 2, telephone interviews were completed with 265 NTD case mothers (84%), 409 orofacial cleft case mothers (85%), 207 conotruncal defect case mothers (87%) and 481 control mothers (78%). Interviews were completed an average of 3.7 years after the date of delivery for cases and 3.8 years for controls.
Interviews in both studies elicited information from women on their residences and dates of residence for all locations at which they had resided for 2 weeks or more during the periconceptional period (defined as the 6-month period from 3 months before to 3 months after conception in study 1 and the 4-month period from 1 month before to 3 months after conception in study 2). Information on maternal medical history, reproductive history and activities associated with various lifestyles was also obtained. In study 1, a beverage consumption history was also elicited, allowing for an estimation of daily tapwater intake.
Estimation of Trihalomethane Exposure
The objective of our exposure assessment was to assign THM levels on the basis of mothers’ residence addresses during their periconceptional periods. More than 95% of all cases and controls had maternal periconceptional addresses that could be accurately matched to 1990 census tracts and to latitude and longitude coordinates. Each address was then linked to a water source and municipal water company by city or geographic coordinates. Only women whose address could be matched to a municipal water company were eligible for further analysis, accounting for approximately 87% of the total number of women who provided accurate address information. Unmatched addresses were predominantly associated with private wells. We also determined which water companies potentially chlorinated their water supplies during the 1986–1991 time period by querying a database at the California Department of Health Services Division of Drinking Water and Environmental Management. This information was also linked to addresses.
We contacted 147 municipal water companies in California that potentially chlorinated water. We asked each one whether specific addresses were in their service area and whether their water system was chlorinated during the time period of interest. We also collected information on total THM and levels of each of the four main THMs (chloroform, bromodichloromethane, dibromochloromethane and bromoform) corresponding to each street address. We asked knowledgeable water company personnel to use quarterly monitoring data on THMs from their water system to estimate, as accurately as possible, total THM and each of the four main individual THM levels for a particular street residence for a given time window corresponding to the actual dates of residence by a woman during the periconceptional period. Certain circumstances may have led personnel to make estimations beyond quarterly sampling data. These circumstances included knowledge about mixing of water sources, changes in pumping practices, or unavailability of monitoring data for an address/time location. For the latter circumstance, data from a comparable water source were provided if available. In circumstances involving water source mixing, personnel provided THM levels averaged across sources. We specifically requested the following: “For each address listed below, please note whether drinking water was chlorinated during the indicated time period. If so, please note if data on total THM or individual chemical species are available. When possible, include the levels of each in parts per billion. We ask that you give your best estimates of values that would represent that address and time period.” Water company staff were further directed that “It is important that you make clear any uncertainties about the data and your ability to match available water quality data with the specified time period and with specific locations in your system.” All estimations using quarterly monitoring data were made by persons unaware of whether an address corresponded to a case or a control.
We obtained information from more than 98% of the 147 water companies. Each case or control mother was assigned a mean total THM level in parts per billion, calculated by averaging THM concentration estimates for each periconceptional address. Exposure levels based on maximum concentration by address (for women who had more than one periconceptional address) were not substantially different. THM exposure levels were considered 0 if they were below detectable measurement (typically ≤ 0.5 ppb) or if the mother’s address was in an area where the water was not chlorinated. The study base for analyses included the numbers of cases and controls displayed in Table 1.
In study 1, women’s THM exposures were further estimated by computing a daily average of cold water consumed at home (milliliters per day). The estimate was produced by multiplying the frequency of tap water intake from all cold fluids (eg, juices made from concentrate) by the serving size. This reflected an average use during the 3 months before conception. Average total THM levels were combined with these questionnaire responses. Information on home tap water intake was not collected in study 2.
The binary outcome of malformation risk was described by using logistic regression models. Odds ratios (ORs) and their 95% confidence intervals (CIs) were computed to estimate effects. Models estimating effects associated with THM exposures were constructed for comparisons that involved 10 or more “exposed” cases and controls combined. Exposure definitions included continuous THM levels as well as selected THM ppb cutpoints. The three cutpoints selected for total THMs were 25, 50 and 75 ppb, approximating cutpoints used in previous studies. 5,6,17 These cutpoints were contrasted to 0 ppb for most risk estimations. We also estimated effects associated with the main chemicals of total THMs. Effect estimates for total THM exposures were further computed for potential covariate strata of maternal race and ethnicity (Latina, foreign-born; Latina, US-born; white; non-Hispanic; or other); education (< high school graduate, high school graduate, some college attended, college graduate), body mass index (weight/height2), and use of multivitamins containing folic acid (4-month periconceptional period in study 2; the 3 months before conception in study 1).
A mechanism has recently been suggested for folate-responsive congenital malformations, which may include the malformations in this study. 13 According to this line of reasoning, mutations in folate-related genes, combined with exposures to certain chemicals such as carbon tetrachloride or THMs, could lead to elevated risks of NTDs and other congenital anomalies for which risk reduction may be associated with maternal folate intake. One particular folate gene is methylenetetrahydrofolate reductase (MTHFR). Because previous investigations had obtained infant MTHFR genotype information on some of the cases and controls in this study, 18–20 we were able to incorporate this genotype information into our THM analyses. These analyses targeted NTDs, isolated cleft lip palate and isolated cleft palate case groups, using a random sample of control infants from studies 1 and 2. All infant DNA that was genotyped came from newborn screening blood specimens; genotyping was done by laboratory staff unaware of the case-control status or exposure status of the infant.
Case and control infants were categorized as TT if they were homozygous for the C677T MTHFR allele, CT if they were heterozygous for the C677T allele, and CC if they were homozygous for the C677 (wild-type) allele. Comparisons were restricted to infants homozygous (mutant or wildtype) for the MTHFR genotype. Our working hypothesis was that infants whose mothers were “chemically exposed” and who were TT genotype would be at higher risk for NTDs, isolated cleft lip palate with or without cleft palate, or isolated, compared with infants who were CC genotype and whose mothers were not “chemically exposed.”
In study 1, the risk of NTDs was inversely associated with total THM exposures. We observed reduced risks of NTDs, including anencephaly and spina bifida, irrespective of whether the THM measurement was modeled as a continuous exposure or as a categorical exposure at any of the selected cutpoints, compared with 0 ppb total THM (Table 2). The computed effect estimates were not substantially influenced by controlling for maternal race/ethnicity, education level, periconceptional vitamin use, and body mass index (Table 2). Analyses adjusted for intake of nitrates from water (an exposure previously studied in these data 21) also did not substantially alter observed effects (data not shown). We explored the effect of total THM exposures in combination with amount of cold tap water consumed. Modeling these as continuous measures (ie, ml per day × total THM level) revealed the following ORs: 1.00 (95% CI = 0.99–1.00) for overall NTDs, 1.00 (0.99–1.00) for anencephaly, and 1.00 (0.99–1.00) for spina bifida. We also modeled the amount of cold tap water consumed, using a cutoff of ≥5 glasses per day compared with <5 glasses per day. This approach was used previously. 17Table 2 shows the results of analyses for total THM exposures ≥50 ppb compared with <50 ppb. Effects associated with THM exposures were not statistically heterogeneous for different levels of cold tap water intake.
The inverse associations between total THM exposures were not explained by geographic area (ie, those case and control women with the highest total THM exposures were not from one geographic area). The inverse association did not appear to be targeted to a specific NTD phenotype, such as anencephaly and spina bifida, nor to a finer classification of spinal defects, such as high spinal defects, low spinal defects, skin-covered defects and open spinal defects (data not shown). All phenotypes were represented in the infants and fetuses whose mothers had the highest exposures.
Effect estimates associated with the selected defects from study 2 (Table 2) did not show the same inverse relationship between NTDs and THM exposure. In fact, elevated effects appeared to be associated with exposure levels of 1 to 24 ppb, particularly for spina bifida. Restricting the geographic area of study 1 to reflect that of study 2 did not substantially alter the observed reduced risks from study 1 (data not shown).
For the other anomaly groups (study 2), effect estimates from continuous models did not reveal associations between total THM exposures and risk. Effect estimates based on categorical total THM cutoffs were elevated for the exposure category 1–24 compared with 0 ppb, but were either not substantially elevated or imprecise owing to sparse data for higher exposure categories.
We also investigated potential effects associated with the main chemicals of total THMs: chloroform, bromodichloromethane, dibromochloromethane and bromoform. The data supporting these analyses were sparse. Fewer than 10% of women had measurable levels of bromoform, so no effects could be estimated for this chemical constituent. To estimate risks, we used the upper 20th percentile level established in controls as the cutpoint. ORs were computed for cases and controls at or above this level to those below this cutpoint. As displayed in Table 3, exposures to each of these chemicals at ≥80th percentile ppb cutpoint levels did not show elevated effects; in fact, most exposures suggested reduced risks.
We examined polymorphisms of the MTHFR gene in conjunction with total THM exposures. It did not appear that the MTHFR TT genotype modified the effects of THM exposure on NTDs, isolated cleft lip with or without cleft palate, or isolated cleft palate (Table 4). Analyses were restricted to a total THM cutpoint of ≥25 ppb owing to small sample sizes of cases and controls whose mothers had exposures at higher ppb cutpoints. Moreover, data were too sparse to consider the additional impact that maternal vitamin use may have had on the potential relationships between maternal THM exposure and infant gene combinations.
Our results do not provide a clear pattern of association between THM exposures and risks of specific congenital malformations. In one study data set (study 1), we observed a fairly strong inverse association between THM exposure and NTD risk. This pattern, however, was not observed in the smaller data set (study 2). We were unable to identify an explanation for these discrepant results. Furthermore, the elevated risks associated with the malformation phenotypes in study 2 were limited to the exposure category of 1–24 ppb, with observed risks either not substantially elevated or imprecise for higher levels of exposure. It is possible that higher risks in the lower exposure category may be indicative of too few controls rather than an excess of cases; for example, 22% of controls were in the 1–24 ppb group in study 1, whereas 9% of controls from study 2 were in this exposure group. It is noteworthy that neither study showed evidence of a dose-response relation.
One possible interpretation of the reduced NTD risk observed in study 1 is that higher total THM exposures may have resulted in selective losses of NTD-affected fetuses. There is some support for such a speculation based on the findings by Waller et al., 17 who found a higher frequency of spontaneous abortions in women with higher levels of THM exposures. NTDs are also more frequent in spontaneously aborted fetuses than in liveborn infants, and spontaneous abortions are 10 to 100 times more prevalent than NTDs. 22 Despite this speculative evidence, we could not determine whether this finding reflected a true association or an association biased by the use of prevalent rather than incident NTD cases.
Reconciling our results with previous findings is not straightforward because there are substantial differences in the measurements of exposures, the completeness of case ascertainment, and the malformation phenotypes that were included. Two studies in New Jersey observed two- to threefold increased risks for NTDs in offspring of women who consumed drinking water containing THMs early in pregnancy. 5,6 The first study showed that infants whose mothers consumed water with more than 80 ppb THM were three times more likely to have infants with NTDs or cleft defects, and approximately two times more likely to have infants with heart defects. 5 The second New Jersey study showed that infants whose mothers consumed water with THM levels of 40 ppb or higher were two times more likely to have NTDs. 6 However, this increased risk was not observed in infants whose mothers were Hispanic, nor in those whose mothers used vitamins containing folic acid periconceptionally. We did not observe this heterogeneity of effect in our study (data not shown). In a study from Norway, Magnus et al.10 observed a modest elevated risk of NTDs (OR = 1.3), but not for orofacial clefts and heart defects, associated with higher levels of organic material (color) and chlorine in water supplies. A subsequent study from Norway observed elevated risks for selected heart defects, but not for NTDs or oral clefts associated with chlorinated water. 12 Our results do not directly coincide with those studies, but they do appear to coincide with some others. 4,7,8,11 In a small California study, women’s consumption of chlorinated tap water at home did not explain the previously identified risk of severe heart defects in tap water drinkers. 4 In a study in Nova Scotia, Dodds et al.8 observed reduced risks for NTDs (eg, OR = 0.41; 95% CI = 0.17–1.00), with no associations of orofacial clefts or severe heart defects with total THM exposures. Although few details are available, an abstract by Lin et al.11 also did not indicate strong associations between THM exposures and risks of NTDs, orofacial clefts, or selected heart defects. Kallen and Robert, 7 in a registry-based study in Sweden, did not observe elevated effects of NTDs, orofacial clefts, or specific heart defect phenotypes in infants whose mothers resided in areas with chlorinated water supplies.
Each of these studies, including ours, has its strengths and limitations. The strengths of our study include the population-based ascertainment of cases and controls, the large sample sizes, the inclusion of several finely classified malformation phenotypes, and the use of water experts using quarterly monitored THM data (blinded to case status) to establish THM levels for individual residences and periconceptional time periods. The latter, however, is of uncertain value because we do not know how often water experts were able to provide better estimates than would have been available from routine monitoring data alone.
The complexities of estimating human exposure to constituents in community water supplies have been well described. 23 The effect estimates that we observed may be biased owing to nondifferential misclassification of exposures. We could not directly measure exposure at the tap for the relevant embryological time period; instead, we assessed exposures by asking water “experts” to estimate the level of THM corresponding to a particular residence address and for a specific time period. However, the concentration of THMs at the tap can vary substantially. THM levels at the tap depend upon the presence and concentration of naturally occurring aquatic humic substances, chlorine dose and contact time, water temperature, and pH level. 2
In addition, water intake is only one potential exposure route of volatile contaminants in drinking water. Bathing and showering have also been identified as important routes of human exposure to THMs. 24 Indeed, indoor air modeling has demonstrated that inhalation of such contaminants has the potential for much higher levels of exposure than drinking does. 25 Other activities, such as washing dishes and clothing, may also be important sources of exposure not measured in this study. Moreover, the current study, like previous ones, was limited to THM exposure levels measured only at women’s periconceptional residences. Women who spent considerable time at other locations where they also consumed tap water (such as at work), may be misclassified in their THM exposure if their THM intake levels varied between home and other locations.
Imprecise exposure measures, coupled with a lack of information about other possible sources of exposure (eg, bathing, using swimming pools, or using water for food preparation at home), may also have led to underestimations of true associations.
Limitations extend beyond the exposure assessment used. In particular, sample sizes were not large for some comparisons. For example, study 1 included only nine cases and 17 control infants whose mothers had a periconceptional residence categorized as receiving ≥75 ppb total THM. Furthermore, our ability to look at the main individual components of total THMs was limited, eg, potential effects associated with bromoform could not be estimated. We were also unable to investigate potential effects of other chemical compounds, such as haloacetic acids, that occur as disinfection by-products. Exposure to such chemicals in experimental systems has demonstrated a range of reproductive effects. 26 Monitoring data were not available for these types of compounds during the study time period. We were unable to investigate constituents other than nitrate level that may have existed in finished water supplies.
THM exposure has been associated with human carcinogenesis, microbial mutagenesis and rodent teratogenesis. 27 However, teratological data for THMs and other chemicals related to water disinfection are limited. Experimental studies have observed NTDs, craniofacial defects and heart anomalies in rats that were given varying doses of haloacetic acids. 28,29 These organic acids have not been studied for their potential association with human malformations with the exception of the investigation by Klotz and colleagues, 6 who did not observe an association with NTDs.
A biologically plausible teratogenic mechanism has been proposed that suggests THMs may increase the risk of congenital malformations that are also caused by folate deficiency. 13 This further suggests that mutations in folate-related genes, combined with exposures to certain chemicals such as carbon tetrachloride or THMs, could lead to elevated risks of NTDs and possibly other congenital anomalies whose risk reduction may be associated with maternal folate intake. 13 Our data did not support such an association.
We thank the field engineers and their staff at the many water companies throughout California who provided us with the water monitoring data.
1. Rook JJ. Formation of haloforms during chlorination of natural waters. Water Treat Exam 1974; 23: 90–92.
2. Symons JM, Stevens AA, Clark RM, et al.
Treatment techniques for controlling trihalomethanes in drinking water. Cincinnati, OH: Environmental Protection Agency; 1981.
3. Stevens AA, Slocum CJ, Seeger DR, et al.
Chlorination of organics in drinking water. J Am Water Works Assoc 1976; 68: 615–620.
4. Shaw GM, Malcoe LH, Milea A, et al.
Chlorinated water exposures and congenital cardiac anomalies. Epidemiology
1991; 2: 459–460.
5. Bove FJ, Fulcomer MC, Klotz JB, et al.
Public drinking water contamination and birth outcomes. Am J Epidemiol 1995; 141: 850–862.
6. Klotz JB, Pyrch LA. Neural tube defects and drinking water disinfection by-products
1999; 10: 383–390.
7. Kallen BAJ, Robert E. Drinking water chlorination and delivery outcome—a registry-based study in Sweden. Reprod Toxicol 2000; 14: 303–309.
8. Dodds L, King W, Woolcott C, et al.
Trihalomethanes in public water supplies and adverse birth outcomes. Epidemiology
1999; 10: 233–237.
9. Aschengrau A, Zierler S, Cohen A. Quality of community drinking water and the occurrence of late adverse pregnancy outcomes. Arch Environ Health 1993; 48: 105–113.
10. Magnus P, Jaakkola JJ, Skrondal A, et al.
Water chlorination and birth defects
1999; 10: 513–517.
11. Lin BK, Lynberg MC, Wong L-Y, et al.
Trihalomethane exposure and risk of selected birth defects
. Paediatr Perinatol Epidemiol 2000; 14: A19.
12. Hwang B-F, Magnus P, Jaakkola JJK. Risk of specific birth defects
in relation to chlorination and the amount of natural organic matter in the water supply. Am J Epidemiol 2002; 156: 374–382.
13. Chen ATL, Sever LE. Re: Public drinking water contamination and birth outcomes (letter). Am J Epidemiol 1996; 143: 1179–1180.
14. Shaw GM, Wasserman CR, Lammer EJ, et al.
Orofacial clefts, parental cigarette smoking, and transforming growth factor-alpha gene variants. Am J Hum Genet 1996; 58: 551–561.
15. Wasserman CR, Shaw GM, O’Malley CD, et al.
Parental cigarette smoking and risk for congenital anomalies of the heart, neural tube, or limb. Teratology 1996; 53: 261–267.
16. Shaw GM, Schaffer D, Velie EM, et al.
Periconceptional vitamin use, dietary folate, and the occurrence of neural tube defects. Epidemiology
1995; 6: 219–226.
17. Waller K, Swan SH, DeLorenze G, et al.
Trihalomethanes in drinking water and spontaneous abortion. Epidemiology
1998; 9: 134–140.
18. Shaw GM, Rozen R, Finnell RH, et al.
Maternal vitamin use, genetic variations of infant methylenetetrahydrofolate reductase and risk for spina bifida. Am J Epidemiol 1998; 148: 30–37.
19. Shaw GM, Rozen R, Finnell RH, et al.
Infant C677T mutation in MTHFR, maternal periconceptional vitamin use, and cleft lip. Am J Med Genet 1998; 80: 196–198.
20. Shaw GM, Finnell RH, Todoroff K, et al.
Maternal vitamin use, infant C677T mutation in MTHFR, and isolated cleft palate risk. Am J Med Genet 1999; 85: 84–85.
21. Croen LA, Todoroff K, Shaw GM. Maternal exposure to nitrate from drinking water and diet and risk for neural tube defects. Am J Epidemiol 2001; 153: 325–331.
22. Creasy MR, Alberman ED. Congenital malformations of the central nervous system in spontaneous abortions. J Med Genet 1976; 13: 9–16.
23. Whorton MD, Morgan RW, Wong O, et al.
Problems associated with collecting drinking water quality data for community studies: a case example, Fresno County, California. Am J Public Health 1988; 78: 47–51.
24. Backer LC, Ashley DL, Bonin MA, et al.
Household exposures to drinking water disinfection by-products
: whole blood trihalomethane levels. J Expo Anal Environ Epidemiol 2000; 10: 321–326.
25. Andelman JB. Inhalation exposure in the home to volatile organic contaminants of drinking water. Sci Total Environ 1985; 47: 443–460.
26. Andrews JE, Schmid J, Nichols H, et al.
Developmental toxicity of mixtures: the water disinfection by-products
dichloro-, dibromo- and bromochloro-acetic acid in embryo culture. Teratology 1999; 59: 384.
27. Mills CJ, Bull RJ, Cantor KP, et al.
Health risks of drinking water chlorination by-products: report of an expert working group. Chronic Dis Can 1998; 19: 91–102.
28. Smith MK, Randall JL, Read EJ, et al.
Teratogenic activity of trichloroacetic acid in the rat. Teratology 1989; 40: 445–451.
29. Smith MK, Randall JL, Read JL, et al.
Developmental toxicity of dichloroacetate in the rat. Teratology 1992; 46: 217–223.
Keywords:© 2003 Lippincott Williams & Wilkins, Inc.
birth defects; case-control; disinfection by-products; environment; epidemiology