Chlorination Disinfection By-Products in Drinking Water and Congenital Anomalies: Review and Meta-Analyses

Objectives The aim of this study was to review epidemiologic evidence, provide summary risk estimates of the association between exposure to chlorination disinfection by-products (DBPs) and congenital anomalies, and provide recommendations for future studies. Data sources and extraction We included all published epidemiologic studies that evaluated a relationship between an index of DBP exposure (treatment, water source, DBP measurements, and both DBP measurements and personal characteristics) and risk of congenital anomalies. When three or more studies examined the same exposure index and congenital anomaly, we conducted a meta-analysis to obtain a summary risk estimate comparing the highest exposure group with the lowest exposure group. When five or more studies examined total trihalomethane (TTHM) exposure and a specific congenital anomaly, we conducted a meta-analysis to obtain exposure–response risk estimates per 10 μg/L TTHM. Data synthesis For all congenital anomalies combined, the meta-analysis gave a statistically significant excess risk for high versus low exposure to water chlorination or TTHM [17%; 95% confidence interval (CI), 3–34] based on a small number of studies. The meta-analysis also suggested a statistically significant excess risk for ventricular septal defects (58%; 95% CI, 21–107), but this was based on only three studies, and there was little evidence of an exposure–response relationship. We observed no statistically significant relationships in the other meta-analyses. We found little evidence for publication bias, except for urinary tract defects and cleft lip and palate. Conclusion Although some individual studies have suggested an association between chlorination disinfection by-products and congenital anomalies, meta-analyses of all currently available studies demonstrate little evidence of such an association.

volume 117 | number 10 | October 2009 • Environmental Health Perspectives Review Disinfection of drinking water has led to a major improvement in public health since it was first applied in the 20th century. It has now been > 30 years since it was discovered that by-products can be formed in small quantities as part of the chlorination process (Rook 1974). Chlorination disinfection byproducts (DBPs) are formed when water is chlorinated and organic matter in the water reacts with chlorine. Their formation and occurrence depends on many factors, including the chlorine dose, type of treatment, pH, temperature, residence time, and bromine levels [International Programme on Chemical Safety (IPCS) 2000; Nieuwenhuijsen et al. 2000a]; up to 600 DBPs have been identified (Richardson 1998;Richardson et al. 2008). Different mixtures of by-products may exist in different locations, depending on the various factors mentioned above, making it more difficult to assess any health effects of DBPs, particularly in epidemiologic studies.
Ingestion of water is thought to be the main route of exposure for nonvolatile DBPs such as halo acetic acids (HAAs). Exposure to volatile DBPs such as trihalomethanes (THMs) may occur through inhalation and absorption during activities such as showering, bathing, and swimming (Nieuwenhuijsen et al. 2000a).
Modeling of THM uptake has suggested that swimming may lead to the highest levels of THMs in the blood (Whitaker et al. 2003).
Various reviews have been conducted on the epidemiology of chlorination by-products and reproductive outcomes. These have concluded that there are still many problems to overcome and that the results are inconsistent and inconclusive, except perhaps for smallfor-gestational-age/intrauterine growth retardation (Bove et al. 2002;Gevecker Graves et al. 2001;IPCS 2000;Nieuwenhuijsen et al. 2000b;Reif et al. 1996;Tardiff et al. 2006). In addition, two meta-analyses by Hwang and Jaakkola (2003) and Hwang et al. (2008) reported evidence for an effect of exposure to chlorination by-products on the risk of neural tube defects, urinary system defects, and ventricular septal defects, but risks for other anomalies were considered hetero geneous and inconclusive. However, the meta-analyses by Hwang and Jaakkola (2003) included only five studies and did not include more recent studies, and the study by Hwang et al. (2008) generally focused on a few sub categories of anomalies and also did not include some recent studies. In this article we review the epidemiologic evidence and provide summary risk estimates of the association between chlorination disinfection by-products and congenital anomalies; we also provide recommendations for future studies.
We included all epidemiologic studies that evaluated the relationship between some index of chlorination by-products (treatment, water source, DBP measurements, and both DBP measurements and personal characteristics) and any congenital anomalies. We did not score the studies for quality, given the relatively small number of studies involved.
We obtained summary risk estimates for general water chlorination by conducting meta-analyses when three or more studies featured the same congenital anomalies. In these analyses we calculated risk estimates for high versus low exposure, using the exposure groups as defined in the original studies [e.g., high vs. low total trihalo methane (TTHM)] concentration, chlorination vs. non chlorination, or chlorination vs. chloramina tion]. When three or more studies had measurements of individual THMs or brominated fractions, we conducted meta-analyses, again comparing the highest versus the lowest exposure groups of the original studies. We tested for hetero geneity in risk estimates using the Q test (Cochran 1954). When the result of the Q test was statistically significant (p < 0.05), implying hetero geneity between the studies, we used random effects analyses, using the method of DerSimonian and Laird (1986). When the Q test was not statistically significant, we used fixed effects analyses, using the Mantel-Haenszel method (Mantel and Haenszel 1959). The summary estimates were weighted by the inverse variance of each study, taking into account whether a fixed or random effects model was used. We used STATA statistical software (StataCorp, College Station, TX, USA).
Where five or more studies, each with the same outcome and exposure classification (applicable only to TTHM) were split into three or more exposure categories, we conducted a meta-analysis to obtain exposureresponse risk estimates per 10 mg/L TTHM. All the included studies provided data on exposure in a categorical way. The midpoint of each given exposure category was used in the estimation of exposure-response slopes for each of the studies. If the highest exposure category was open-ended (i.e., greater than some concentration), a value of half the width of the previous exposure category was added to the cut point of highest exposure category, using the method described by Key et al. (2006). In the absence of data to the contrary, we assumed that the natural logarithm of the measure of effect varied linearly with concentration of TTHM in drinking water. Exposure-response slopes for those exposed to TTHM in drinking water were calculated using linear regression of log-transformed risk estimates. We performed the meta-analyses using R software (R Project for Statistical Computing 2009) and scripts adapted from those of Key et al. (2006). A random effects model was used for the meta-analysis of exposure-response slopes of individual studies. Regression slopes of exposure-response slopes derived from individual studies were plotted together with the summary slope produced from their meta-analysis.
We also produced forest and funnel plots (Light and Pillemer 1984). Forest plots show the odds ratio (OR) with its confidence interval (CI) for each study included in the metaanalyses and the estimation of the summary OR. The sizes of the square markers of OR in the plot represent the relative weight each study contributed to the regression.
Funnel plots show, for every study, the effect estimate (OR and relative risk) of the study against some precision meas ure (SE in this case) to examine possible bias due to publication. We also conducted a weighted Egger test, a linear regression in which the response is the estimated effect and the explanatory variable is a precision term (1/SE). A large deviation from zero of the slope term suggests publication bias (Egger et al. 1997).

Results
The extracted and evaluated studies are shown in Table 1. A third of these studies were conducted in the United States (Aschengrau et al. 1993;Bove et al. 1995;Klotz and Pyrch 1999;Luben et al. 2008;Shaw et al. 2003), followed by Sweden (Cedergren et al. 2002;Källen and Robert 2000), England and Wales (Iszatt N, Nieuwenhuijsen MJ, Toledano MB, Nelson P, Elliott P, unpublished data;Nieuwenhuijsen et al. 2008), Canada (Dodds et al. 1999;Dodds and King 2001), Norway (Hwang et al. 2002;Magnus et al. 1999), Australia (Chisholm et al. 2008), and Taiwan (Hwang et al. 2008). Five were case-control studies [Aschengrau et al. 1993; Iszatt N, Nieuwenhuijsen MJ, Toledano MB, Nelson P, Elliott P, unpublished data; Klotz and Pyrch 1999;Luben et al. 2008;Shaw et al. 2003 (includes two studies)], whereas the rest were cross-sectional, population-based studies, generally using registry data, which limited the opportunity to adjust for potential confounders. Most included live births, fetal deaths, and terminations. A number of the studies used only treatment method (chlorination vs. nonchlorination, chlorination vs. chloramination) as the index of exposure (Aschengrau et al. 1993;Hwang et al. 2002;Källen and Robert 2000;Magnus et al. 1999), whereas the other studies used a meas ure of DBP exposure level, mostly THMs (Bove et al. 1995;Cedergren et al. 2002;Chisholm et al. 2008;Dodds et al. 1999;Dodds and King 2001;Iszatt N, Nieuwenhuijsen MJ, Toledano MB, Nelson P, Elliott P, unpublished data;Klotz and Pyrch 1999;Luben et al. 2008;Nieuwenhuijsen et al. 2008;Shaw et al. 2003). Generally, studies used TTHM levels, but some studies also evaluated individual THMs or the brominated fraction (Dodds and King 2001;Iszatt N, Nieuwenhuijsen MJ, Toledano MB, Nelson P, Elliott P, unpublished data;Nieuwenhuijsen et al. 2008;Shaw et al. 2003). The levels of brominated THM were high in Australia and made up most of the TTHM there (Chisholm et al. 2008). Klotz and Pyrch (1999) also had a meas ure of halo aceto nitrile (HAN) and HAA exposure, and Luben et al. (2008) a measure of HAA exposure. Levels of DBPs in Scandinavia (Cedergren et al. 2002;Hwang et al. 2002;Källen and Robert 2000;Kuivinen and Johnsson 1999;Magnus et al. 1999) and Taiwan (Hwang et al. 2008) tended to be lower than in the other countries, with median levels around 10 µg/L TTHM, with approximately 20% > 20 µg/L. The Canadian (Dodds et al. 1999;Dodds and King 2001) and Australian (Chisholm et al. 2008) studies examined the highest TTHM levels (> 100 µg/L). Three studies examined personal behavioral activities such as ingestion, showering, and bathing (Iszatt N, Nieuwenhuijsen MJ, Toledano MB, Nelson P, Elliott P, unpublished data; Klotz and Pyrch 1999;Luben et al. 2008). The largest study was conducted in England and Wales and included > 2.5 million births and > 20,000 cases of congenital anomalies . Dodds et al. (1999) and Dodds and King (2001) used the same subjects but examined the effect of TTHM and individual THMs, respectively. The study by Hwang et al. (2002) included 3 years of subjects from Magnus at al. (1999), and therefore only the former was included in any of the meta-analyses. The report by Shaw et al. (2003) contained two case-control studies (study 1 and study 2), and they have been treated as separate studies.
Neural tube defects were one of the most commonly studied groups of congenital anomalies, and three studies found statistically significant positive associations between THM levels or chlorinated water and neural tube defects (Bove et al. 1995;Dodds and King 2001;Klotz and Pyrch 1999), whereas others did not (Chisholm et al. 2008;Dodds et al. 1999;Hwang et al. 2002;Källen and Robert 2000;Magnus et al. 1999;Nieuwenhuijsen et al. 2008;Shaw et al. 2003) [ Table 2; see also Supplemental Material 1, Tables 2a-2d (doi:10.1289/ehp.0900677.S1)]. There appeared to be little difference between the findings of using TTHM levels or brominated species as the exposure index. The metaanalyses for high versus low exposure for water volume 117 | number 10 | October 2009 • Environmental Health Perspectives chlorination (THM levels or chlorinated water) and bromo dichloro methane (BDCM) showed some excess of risk, but it was not statistically significant. We found no evidence for an exposure-response relation ship with TTHM levels (OR = 1.00; 95% CI, 0.94-1.07 per 10 µg/L TTHM) [see Supplemental Material 3, Figures 1-3 (doi:10.1289/ehp.0900677.S1)]. Klotz and Pyrch (1999) found a statistically significant association between TTHM levels in the water and neural tube defects, but not with levels of halo acetonitriles and halo acetates.
Inclusion of personal behavioral activities made little difference in the results. In addition, the effects were most pronounced in offspring from women who did not take supplementary vitamins, but these findings were not confirmed by Shaw et al. (2003). Findings of the sub categories of neural tube defects including anencephalus, spina bifida, and hydro cephalus were also mixed, and none of the studies found a statistically significant association except for Shaw et al. (2003) who found a statistically significant reduced risk. The meta-analyses showed some excess of risk for anencephalus and spina bifida, but it was not statistically significant. Cedergren et al. (2002), Chisholm et al. (2008), and Hwang et al. (2002) found statistically significant positive associations between chlorinated water with a higher color content, levels of TTHM > 10 µg/L, and high levels of DBPs (≥ 130 µg/L), respectively, and cardiovascular congenital anomalies, but other studies did not observe such an association (Bove et al. 1995;Dodds et al. 1999 (Hwang et al. 2002(Hwang et al. , 2008Nieuwenhuijsen et al. 2008), and a statistically significant excess risk was observed in the meta-analysis (OR = 1.59; 95% CI, 1.21-2.07) (Figure 2). However, only Hwang et al. (2002) showed some indication of an exposure-response relation ship. Ventricular septal defects showed little association with brominated THM levels in the study by Nieuwenhuijsen et al. (2008). No statistically significant associations were observed with atrial septal defects in the two studies by Hwang et al. (2002Hwang et al. ( , 2008. Two studies found a statistically significant positive association between chlorinated water and congenital anomalies of the respiratory system (Aschengrau et al. 1993;Hwang et al. 2002), but two other studies using TTHM levels did not (Chisholm et al. 2008;Nieuwenhuijsen et al. 2008) [ Table 2; see also Supplemental Material 1, Table 4 (doi:10.1289/ehp.0900677.S1)]. Results of the two Swedish studies were inconsistent (Hwang et al. 2002;Magnus et al. 1999). The meta-analysis showed some excess of risk, but it was not statistically significant.
Generally there was little evidence for publication bias, except for cleft lip and palate, specifically when comparing the exposure-response relationships (p = 0.04), and urinary tract defect (p = 0.002) [ Table 2; see also Supplemental Material 2, Figures 1-15 (doi:10.1289/ehp.0900677.S1)]. However, it should be noted that publication bias tests and funnel plots are not considered rigorous tests when the number of studies is small.

Discussion
The epidemiologic studies we reviewed showed inconsistent results for an association between drinking water chlorination by-products and risk of all congenital anomalies combined and of specific groups of anomalies. For all congenital anomalies combined, the meta-analysis gave a statistically significant excess risk for high versus low exposure to water chlorination or TTHM (17%; 95% CI, 3-34), based on a small number of studies. The meta-analysis also suggested a statistically significant excess risk for ventricular septal defects (58%; 95% CI, 21-107), but this was based on only three studies, and there was little evidence of an exposure-response relationship. We observed no statistically significant relationships in the other meta-analyses. We found little evidence for publication bias except for urinary tract defects and cleft lip and palate.
Major limiting factors in studies on chlorination by-products and congenital anomalies are crude exposure assessment (with the exception of more recent studies), small samples sizes, heterogeneity in outcomes, and, to a lesser extent, potential for bias and confounding and ability to detect susceptible groups. These factors together may explain some of the mixed results and possibly the lack of associations. Furthermore, combined with the small number of studies included in the meta-analyses, these factors also reduce the strength of any conclusions that can be drawn from meta-analyses. These analyses are therefore not meant to provide final conclusions on the subject, but instead evaluate the current status of this growing body of research and offer guidance for the way forward.
Use of ecologic water supply zone estimates as an exposure index may result in exposure misclassification (Whitaker et al. 2003). Furthermore, ingestion has generally served as the primary exposure route of interest, in spite of the fact that uptake through showering, bathing, and swimming could be considerable, specifically for THMs; such exposure has been considered in only a few studies (e.g., Iszatt N, Nieuwenhuijsen MJ, Toledano MB, Nelson P, Elliott P, unpublished data; Klotz and Pyrch 1999;Luben et al. 2008). Combining information on individual water use with water zone estimates could provide better exposure estimates, but the individual information should be evaluated for measurement error, because within-subject variability in questionnaire data may be substantial (Forssén et al. 2008) and attenuate risk estimates. Furthermore, exposure estimates have been based primarily on maternal residence at birth. This ignores any exposure that occurs outside the home (e.g., in the workplace) and also ignores the possibility that a mother has moved to another residence during her pregnancy. Therefore, exposure assessment based on maternal residence at birth may result in exposure misclassification. In addition, studies from Scandinavia (Cedergren et al. 2002;Hwang et al. 2002;Källen and Robert 2000;Magnus et al. 1999) and Taiwan (Hwang et al. 2008), for example, have generally shown low levels of DBPs with a small range, making risk estimation more difficult because of a higher probability of exposure misclassification, particularly where seasonal variability has not been taken into account. On the whole, epidemiologic studies have used THMs as a proxy for total DBP load, but THMs are not necessarily a good proxy meas ure. Only two studies have investigated other DBPs such as HAAs (Klotz and Pyrch 1999;Luben et al. 2008). The metabolism of different DBP species varies (IPCC 2000), so it is insufficient to analyze DBPs as a whole or to use TTHM as a proxy. Investigation of the relation between non-THM by-products and reproductive outcomes is required to understand whether any specific DBP may cause health effects. A detailed assessment of the DBP mixture is required to provide an understanding of any observed results.
Sample sizes have often been insufficient to produce stable results, and this may have resulted in chance findings and mixed results, although there are exceptions. For example, the studies by Hwang et al. (2002Hwang et al. ( , 2008 and Nieuwenhuijsen et al. (2008) have provided sufficiently large numbers of cases to create various exposure categories with more stable risk estimates, and were able to examine exposureresponse relationships to some extent. A related limitation of the presented meta-analyses is the relatively small number of studies and therefore the need to conduct relatively simple analyses, comparing high-versus low-exposure categories and combining what could be regarded as different and/or inconsistent estimates of exposure (e.g., high vs. low TTHM concentration and chlorination vs. nonchlorination). For instance, in Table 2, for all congenital anomalies Chisholm et al. (2008) used cut points of < 60 µg/L TTHM for low exposure, whereas the high-exposure group used by Hwang et al. (2008) was > 20 µg/L TTHM. Although we assume that the meta-analysis is statistically accurate, the biological basis for comparing studies with this degree of heterogeneity in the definition of exposure is more problematic, and this should be taken into account when interpreting the results. Ideally, all the exposure categories have the same cut offs, but in practice, this is often impossible because of the different local conditions. We also performed sensitivity analyses leaving out the studies with qualitative estimates (e.g., chlorination  The analyses of TTHM exposure-response relationships combined more comparable exposure levels and were therefore probably more informative, but they could only be conducted for a few end points because of the lack of a sufficient number of studies. The question here is whether TTHM may be the putative agent or a (not so good) marker for something else. Also, as a result of the small number of cases in the studies and to increase power, congenital anomalies have often been categorized into main groups such as NTDs, major heart defects, and abdominal defects, or as all congenital anomalies combined. These anomalies, however, are generally hetero geneous with respect to both pheno type and presumed etiology, and combining them may not be appropriate. For example, Nieuwenhuijsen et al. (2008) showed that focusing on isolated sub categories may result in different findings. Of course, this was possible only because of the large number of subjects in the study.
The retrospective and registry-based nature of many of the studies has meant that information on lifestyle factors, such as maternal smoking and alcohol consumption, has often been lacking. These factors are not necessarily potential confounders, as it seems unlikely that levels of chlorination by-products are related to such factors, nor are they established risk factors for all congenital anomalies. However, behavioral patterns such as ingestion of tap water may be associated with these factors (Forssén et al. 2007(Forssén et al. , 2008. Furthermore, registry completeness is a factor that may make studies incomparable. In some places, case ascertainment occurs immediately after birth, so some malformations may not be recorded, especially malformations such as hypo spadias and cardiac defects, where the less-severe forms may not immediately be recognized. In Norway and Taiwan, birth defects reported up to 7 days after birth were included in the study (Hwang et al. 2008;Magnus et al. 1999), whereas in California (Shaw et al. 2003) and Sweden (Cedergren et al. 2002), cases were diagnosed up to 1 year after birth. However, we anticipate that random under reporting would weaken any observed association rather than introduce a spurious effect. A further discrepancy between studies was whether the study population had been extended beyond live births. In the European Union, terminations accounted for 18% and fetal deaths for 2% of congenital anomalies [European Surveillance of Congenital Anomalies (EUROCAT) 2005]. However, whereas some study populations included terminations and stillbirths (i.e., Dodds et al. 1999;Klotz and Pyrch 1999;Nieuwenhuijsen et al. 2008), others did not (i.e., Cedergren et al. 2002;Chisholm et al. 2008;Hwang et al. 2008).
The strongest, albeit relatively modest, evidence for a possible association with DBPs was found for one specific type of cardiac defect: ventricular septal defect. This finding is of interest in light of studies of other environ mental exposures in which associations for this particular birth defect have been observed with, for example, air pollutants (Gilboa et al. 2005;Ritz et al. 2002). However, ventricular septal defects are also among the anomalies that are most subject to variable diagnosis and reporting in routine anomaly registries; all three studies that reported increased risks of ventricular septal defects in relation to high chlorination exposure were based on large, regional or nationwide, congenital anomaly registries in which variable reporting is common (Rankin et al. 2005) and may affect geographic comparison studies such as the DBP studies.
Only one study (Shaw et al. 2003) has examined gene-environment interaction and/or the presence of susceptible groups, and this study did not find any effect. There is some evidence, however, to suggest that cytochrome P450 2E1 (CYP2E1) or glutathione S-transferase T1 may play a role in the susceptibility to DBPs, and this should be further explored (Infante-Rivard 2004;Infante-Rivard et al. 2002).
Animal and cell studies have found some effects associated with DBPs. NTDs and cranio facial defects have been found with administration of di/trichloro acetic acid in rats (Smith et al. 1989a), and cardiac malformations have been induced at high doses of dichloro acetic acid (Smith et al. 1992). Hunter et al. (1996) observed changes in neural tube develop ment when mouse embryos were exposed to HAAs. Several chloro acetonitrile compounds have been shown to increase the rate of resorption, to reduce fetal body weight and survival (Smith et al. 1997), and to result in an increase in malformations of the cardiovascular, digestive, soft tissue, and uro genital systems (Smith et al. 1988(Smith et al. , 1989b. However, these adverse develop mental effects have only been demonstrated at high doses in conjunction with severe feto toxicity. Furthermore, the mechanisms for possible effects are still unclear, although some have been suggested. Low levels of folate have been associated with several congenital anomalies such as neural tube defects (EUROCAT 2003). Alston (1991) found that chloroform inhibited methionine bio synthesis in cell culture. Dow and Green (2000) showed that trichloro acetic acid interacts with vitamin B12, probably by a free radical mechanism, inhibiting both the methyl malonyl coenzyme A and methio nine salvage pathways in rats. As a result of the latter, a secondary folate deficiency develops because of the methyl folate trap, leading to a major impairment in formate metabolism. Folate and folic acid are forms of vitamin B that are involved in the synthesis, repair, and functioning of DNA and are required for the production and maintenance of cells (Kamen 1997). Folate plays an important role for cells under going rapid turnover, such as tissues in the developing fetus.

The Way Forward
Given the many studies conducted and the limited evidence of an association between chlorination by-products and congenital anomalies, we might ask whether there is a need to conduct more studies, and if so, what should they look at? Disinfection of drinking water is an important part of public health, and many people are exposed to chlorination by-products not only through ingestion but also through other activities such as showering and bathing. Ongoing surveillance of any possible adverse health effects is therefore warranted, even though the relative risks may be small. As mentioned above, the mixture of the by-products may differ by geographic area and time, for example, because of changes in water treatment methods, and generally only indicator substances such as TTHM have been used to examine the health risks. It is important that we understand the under lying mixture of the by-products, both in existing and new studies, and where possible, we should examine any possible health risks of specific DBPs or mixtures. Studies are needed to examine the potential effects of certain mixtures, such as brominated species, in more specific locations. For example, Perth, Australia (see Chisholm et al. 2008) or the Barcelona area in Spain (see Villanueva et al. 2006) might be suitable places because of their high levels of brominated compounds.
Furthermore, it would be worthwhile to examine the various exposure pathways and routes other than ingestion in more detail, specifically for volatile by-products such as THMs, as the level of exposure and metabolism may be different and the meas ures for exposure prevention are likely to be different. This can probably only be done in casecontrol studies; in such studies it would be necessary to estimate exposure in the most critical (early) periods of pregnancy, which could prove difficult. Prospective exposure assessment through a cohort design would be more appropriate, but practical and financial constraints preclude such a study, as the size of the cohort would need to be extremely large to study rare congenital anomalies.
Regarding the outcomes, the focus of future studies should be on subcategories of congenital anomalies, rather than on the whole group, and should focus on anomalies for which the ascertainment is reasonably complete and consistent if registry-based designs are used. Findings for ventricular septal defects should be followed up, preferably in well-designed case-control studies. The study by Nieuwenhuijsen et al. (2008) showed an excess risk for bromo form and gastro schisis; these findings may be worth examining in more detail and in a different population. One of the problems in that study was the low levels of bromo form in England and Wales, and future studies should be conducted in places where bromo form levels are higher (such as Perth).
Further work is needed on the relation between potential confounders such as smoking and alcohol intake, as well as the relation with by-products in the water and personal behavioral characteristics such as tap water ingestion (instead of bottled water), showering, and bathing to examine to what extent confounding may explain findings for registry-based studies where this information is missing.
There is some suggestion that some chlorination by-products may interfere with folate metabolism; this and other potential mechanisms such as oxidative stress and geno toxicity could be examined with biomarkers in pregnant women to assess to what extent this may be possible. Furthermore, geno typing may identify susceptible populations (e.g., those with CYP2E1).