Drinking water disinfection by-products exposure and health effects on pregnancy outcomes: a systematic review

The use of disinfectants as a drinking water treatment step in developing countries has led to an effective decrease of waterborne diseases (World Health Organization (WHO) 2011). Since then, chlorine and other related compounds are being used globally because of successes and milestones in public health protection. It has been known for more than 20 years that chlorination of surface water produces chloroform and other toxic compounds that are health risks (Rook 1974). More than 600 DBPs have been identified to date (Richardson et al. 2007; Nieuwenhuijsen et al. 2009a). The most commonly studied DBPs are THMs and haloacetic acids (HAAs) as they occur in higher concentrations in tap water compared to others.

The health effects of drinking water DBPs on adverse pregnancy outcomes have been previously reviewed; however, the conclusions of these reviews varied widely, from indicating association to suggesting no association of DBPs on pregnancy outcomes (Bove et al. 2002; Hwang & Jaakkola 2003; Nieuwenhuijsen et al. 2009b, 2010; Grellier et al. 2010). In addition, previous reviews published on this subject did not assess the risk of bias among the articles studied.

The aim of this study was to systematically review the evidence on the risks of spontaneous abortion (miscarriage), preterm or premature birth (PTB), low birth weight (LBW) and SGA associated with exposure to different drinking water DBPs.

The current study followed Preferred Reporting Items for Systematic Reviews and Metal-Analyses (PRISMA) guidelines and guidance of observational studies in epidemiology criteria (Moher et al. 2009). The study received ethical approval from the Faculty of Health Sciences, University of Pretoria (reference no. 115/2016).

The pertinent object of this study was to evaluate the evidence/literature on risk of spontaneous abortion (miscarriage), premature birth, LBW and SGA among women exposed to different drinking water DBPs during pregnancy.

Observational studies that measure the association of drinking water DBPs and maternal exposure and adverse pregnancy outcomes were included in this review. Published studies in peer-reviewed journals from 1986 to 2016 formed part of the investigation. We limited ourselves to articles published in English with females as participants, and articles conducted in communities or populations. Studies with information on (1) THMs, (2) HAAs, (3) haloacetonitriles (HAs), (4) haloketones (HANs), (5) bromate and (6) chlorate as selected drinking water DBPs were included. The studies, which provide rational information on the method of measuring the maternal exposure to DBPs and their effects on selected adverse pregnancy outcomes, were considered for inclusion in the review. Studies with less than 30 participants, and summarized publications and studies using chlorination and monochloramination in other matrices or contexts other than drinking water disinfection, were excluded from this review.

Peer-reviewed observational studies, retrospective or prospective cohort studies, case-control studies, and cross-sectional studies were included in the review.

In this review, we included studies identifying pregnant women exposed to various drinking water DBPs during the time of pregnancy (at any stage of gestation).

Data regarding maternal exposure to drinking water DBPs obtained via three sources were eligible for inclusion: (1) measurement of participants' urine TCAA levels as a biomarker; (2) measuring DBPs from drinking water source (i.e., residential tap water); and (3) linking data obtained from municipal or national monitoring database on water quality measurements, resulting in an estimated value for women's exposure during time of pregnancy.

We included any studies that reported data on four of the following outcomes: (1) pregnancy loss or spontaneous abortion (miscarriage); (2) premature/preterm birth or premature delivery (PTB or PTD), defined as gestational age of less than 37 weeks; (3) LBW, defined as birth weight of less than 2.5 kg; and (4) SGA, defined as birth weight below the 10th percentile for gestational age.

Searches were performed using PubMed, Medline and Google scholar electronic databases using the terms and key words and a combination of the key words. Additional data were extracted from grey literature, which include but are not limited to World Health Organization, ProQuest dissertations as well as theses, databases and conference proceedings. Water industry bodies, such as the South African Water Research Commission (WRC), were also used to search for data. The phrases and key words used were based on the terminology commonly used in this subject, including: ‘drinking water disinfection by-products’, ‘chlorination of water’, ‘monochloramination of water’, ‘chlorination disinfection by-products’, ‘chloramination disinfection by-products’, ‘exposure to disinfection by-products’, ‘disinfection by-products health effects’, ‘haloacetonitriles’, ‘haloketones’, ‘trihalomethanes’, ‘haloacetic acids’, ‘bromate’, ‘chlorate’, ‘birth outcomes’, ‘adverse pregnancy outcomes’, ‘gestational age’, ‘premature birth or preterm birth’, and ‘birth weight’ (Table 1). The searched articles were screened from their titles and abstracts to select the eligible studies.

Two reviewers independently evaluated each of the full texts of eligible studies and any disagreement was resolved via discussion, with the help of a third reviewer. Data extracted from each were piloted on to the data collection form without modifying its origin. Lists of confounders were also collected before being adjusted for analysis in the studies. The included articles were then assessed for quality.

The tool to assess the quality was adopted from Shah & Zao (2009) (Table 2) with the scale from none to high bias. The table includes the following characteristics: selection, exposure assessment, outcome assessment, confounding factor, analytical and attrition bias.

A systematic review of these was made rather than performing meta-analyses, as heterogeneities were recognized in previous reviews. We did not assess statistical heterogeneity and publication bias, as the meta-analyses were not performed. However, studies were assessed for methodological differences and data were reported.

Search results and the number of studies are summarized in Figure 1, along with the reasons for excluding the studies from this review. Detailed characteristics of all included studies are reported in Table 1.

Risks of bias assessment of the included studies were studied. Of the 32 studies included, 26 (81%) had an overall low risk of bias, whereas six (19%) had an overall moderate risk of bias. In exposure assessment, moderate risks of bias were assigned to the studies due to indirect use of exposure methods employed. Most studies had no risk of bias for outcome assessment, confounder adjustments and participant selection (see Table 3 for details).

Adverse pregnancy outcomes according to various drinking water DBPs were assessed. Studies included in the review reported various drinking water DBPs; however, the majority of studies reported on THMs followed by HAAs, or both. The results below are classified according to individual DBPs. The details of level of exposure, time of exposure and birth outcomes are reported in Tables 2–6.

Twenty-nine studies reported data on adverse pregnancy outcomes following THM exposure (Table 2). Studies reported on various levels of exposure, ranging from 0 to 108.8 μg/L. Sixteen studies investigated the association between THM exposure and SGA, and 13 studies reported on either PTB, PTD or LBW and pregnancy loss, or all three. Ten studies reported on exposure to THMs during the third trimester, 13 studies during the entire pregnancy, five studies on both first trimester and entire pregnancy, or second trimester and third trimester, or only the second trimester's maternal exposure. One study reported on first, second and third trimesters and entire pregnancy) THM exposure. Nine studies reported on evidence of association between maternal THM exposure and adverse pregnancy outcomes (Dodds et al. 1999; Aggazzotti et al. 2004; Maclehose et al. 2008; Grazuleviciene et al. 2011; Levallois et al. 2012; Rivera-Núñez & Wright 2013; Kumar et al. 2014; Iszatt et al. 2014; Cao et al. 2016). These studies were conducted in countries including the USA, UK, Canada, China, Italy and Europe. Twenty studies reported no association between THM exposure and adverse pregnancy outcomes (see Table 2).

The exposure to HAAs and adverse pregnancy outcomes was explored in 12 studies (Table 3). The levels of exposure to HAAs ranged from 0.1 to 75.9 μg/L among the studies. Five studies reported on exposure to HAAs and SGA, four studies reported on exposure to HAAs and PTB/PTD, and five studies on either LBW, BW or BWT. Two studies reported on exposure to HAAs and pregnancy loss. Hoffman et al. (2008b), Hinckley et al. (2005), Wright et al. (2004) and Zhou et al. (2012) reported exposure to HAAs during the third trimester; six studies reported on exposure during the entire pregnancy (Savitz et al. 2006; Maclehose et al. 2008; Costet et al. 2011; Grazuleviciene et al. 2011; Horton et al. 2011; Danileviciute et al. 2012); Hoffman et al. (2008a) reported on the first trimester and entire pregnancy; and Rivera-Nunez & Wright (2013) reported on the second and third trimesters. Eight studies were conducted in the USA, two in Europe, one in China and one in France. A slightly positive association between adverse pregnancy outcomes and high level of HAA exposure was reported among the studies (see Table 3).

Two studies (Danileviciute et al. 2012; Ileka-Priouzeau et al. 2015) reported on adverse pregnancy outcomes following HA exposure (Table 4). The levels of exposure to HAs ranged from 1.0 to 9.00 μg/L. Both articles studied HA exposure and SGA or LBW, or both. Studies were conducted in Europe and Canada, respectively. No association between HA exposure and SGA or LBW was reported.

Three studies (Aggazzotti et al. 2004; Danileviciute et al. 2012; Ileka-Priouzeau et al. 2015) reported on pregnancy outcomes following chloropicrin and chloral hydrate exposure (Table 5). The levels of exposure were ≥200 μg/L for chlorate or chlorite concentrations. Two studies reported on exposure to HANs during the entire pregnancy and one during the third trimester. Studies reported on SGA and PTD, SGA and LBW, and SGA, respectively. Studies were conducted in Italy, Europe and Canada. None of the studies reported an association.

Four studies (Savitz et al. 2006; Hoffman et al. 2008b; Maclehose et al. 2008; Horton et al. 2011) reported data on TOX and pregnancy outcomes (Table 6). The levels of exposure to TOX ranged from 18.7 to 186 μg/L. Three studies were conducted during the entire pregnancy and one during the third trimester. The adverse pregnancy outcomes reported were either pregnancy loss, SGA or PTB. A slightly positive association between pregnancy loss with an increased TOX exposure was reported (see Table 6).

We used a systematic review of 32 studies to assess the associations between exposure to drinking water DBPs and adverse pregnancy outcomes. We identified various DBPs on adverse pregnancy outcomes of spontaneous abortion (miscarriage), preterm or premature birth, LBW and SGA. Various DBPs found in drinking water include THMs, HAAs, HAs, HANs and TOX.

In this review, 38% of the studies included in the review reported on evidence of association between maternal exposure to drinking water DBPs and adverse pregnancy outcomes. This was consistent with the findings from Grellier et al.’s (2010) review, where 40% of studies included were statistically associated with birth outcomes. THMs were associated with SGA and slightly with LBW or pregnancy loss. Higher concentrations of HAA and TOX exposures were slightly associated with SGA and pregnancy loss, respectively. The evidence of any association between other drinking water DBPs (HAs, HANs) and adverse pregnancy outcomes is still inconclusive. The examination of drinking water DBPs and adverse pregnancy outcomes or birth outcomes is still a challenge in most epidemiological studies.

Our reviewed articles demonstrated different exposure assessment methods used, of which 31% of the studies used exposure data obtained from national or local databases housed by water utilities/industries, 34% used water sampling campaigns to measure the DBP concentrations in residential areas, and 22% used both water sampling campaigns and national or local databases from water utilities/industries to assign the exposure to the participants with some instances relying on questionnaires for personal habits. The indirect approach of measuring exposure assessments is still the most commonly applied methodology because it is less costly.

Recently, other methods of measuring exposure in epidemiological studies have been explored. For instance, the use of blood samples to measure the internal exposure was reported by Danileviciute et al. (2012) in this review. Blood THMs decrease within minutes to hours after exposure; however, slower partitioning out of adipose tissue and the relatively high (e.g., daily) frequency of exposure events such as showering/bathing are thought to produce steady-state blood concentrations. Costet et al. (2011) and Zhou et al. (2012) also explored the use of urine samples to measure the TCAA levels. TCAA is one of the major HAAs and is used as a biomarker because it is stable, unmetabolized in urine and is not readily degraded through the collection and storage processes (Smith et al. 2013). The eradication half-life of TCAA is between 2 and 6 days, which provides enough information on urinary concentration (Savitz 2012). Both studies showed positive evidence of using a biomarker for exposure assessment.

In this review, included studies were carried out in well-developed countries. The effects of exposure on adverse pregnancy outcomes can vary according to the country in which the study is being conducted, as the regulatory standards differ. Countries in the European Union, the USA, the UK, Australia, China and others (e.g., Canada) have set standards which benchmark against the WHO drinking water quality guidelines of 2011. For instance, the EU set the drinking water quality standard for total THMs to 100 μg/L (WHO 2011). The USA set a regulatory standard for THMs to 80 μg/L and 60 μg/L for five HAAs, 10 μg/L for bromate and 1,000 μg/L for chlorite (USEPA 2011). Canada set a limit of 80 μg/L for THMs with provisional maximum acceptable concentrations of 100 μg/L level of THMs according to the 2003 guidelines for Canadian drinking water quality (Rodriguez et al. 2004).

A previous review by Bove et al. (2002) found moderate association between THM exposure and birth outcomes (SGA and spontaneous abortions). Their results correlate with the findings from the review conducted by Grellier et al. (2010), where SGA was associated with exposure to total THMs; of nine studies included in Grellier et al.’s (2010) review, none found statistically significant association of DBPs with preterm birth (Kramer et al. 1992; Savitz et al. 1995; Gallagher et al. 1998; Dodds et al. 1999; Wright et al. 2003, 2004; Lewis et al. 2007; Yang et al. 2007; Hoffman et al. 2008a). Our results have both similarities and differences compared with the previous reviews. However, none of the previous reviews have used the PRISMA guidelines. In addition, impact based on individual drinking water DBPs was not observed from previous reviews.

This review has demonstrated several strengths. To our knowledge, it is the first review to assess associations of adverse pregnancy outcomes using PRISMA guidelines. The method of reviewing also assesses the maternal exposure to individual drinking water DBPs. Risk assessment of biases in the included studies and analyses of exposure-outcome measurement also gives strength to this review. However, it also has its limitations. We did not retrieve the raw data for studies included in the review. We also limited our search strategies to English language publications so there may be a low possibility of obtaining different results in non-English language articles.

Health determinant factors that contribute to adverse pregnancy outcomes should be considered when interpreting the results. Therefore, our data in most studies included in the review were extracted after adjusting the confounders. Another limitation, as in other reviews, is that the adverse pregnancy outcomes' definitions are not the same across the studies. These limitations are important to bear in mind when considering the conclusions of this review. The purpose of this article was to assess associations or risks, not to disprove or prove causality.

Mothers' exposures to common (THMs, HAAs) drinking water DBPs are associated with adverse pregnancy outcomes. In addition, the concentration levels of DBPs studied varied between studies. Most studies are being conducted in developed countries where the set standards are well established and regulated. Evidence of any association between other drinking water DBPs (HAs, HANs) and adverse pregnancy outcomes is still inconclusive. However, the absence of association results does not demonstrate the absence of health effects on pregnancy outcomes. Likewise, a statistical significance does not always suggest clinical importance. Difficulty in measuring exposure, inappropriate time of measurement and interaction between drinking water DBPs may have resulted in absence of association in most studies. The use of urinary TCCA biomarkers as a direct exposure assessment method is deemed to be the future in this field.

Health impact associated with DBPs is a global issue for public health perspectives. The findings of this review underline the need for action to be taken in the reduction of exposure to DBPs, especially during pregnancy. The association of THMs and pregnancy outcomes indicates that exposure to high THM concentration during pregnancy is harmful to the foetus. The association of other health determinants factors and birth outcomes are important, as indicated in other studies. National, regional and local water industry efforts are needed to reduce the production of DBPs in drinking water. Even though common DBPs are regulated internationally, exposure to them can vary according to individual actions, such as water activity habits (especially during pregnancy) as many pregnant women tend to consume more water than normal.

Future studies need to focus on the underlying biological mechanisms to understand the impact of individual contaminants as well as the interactions between them. Previous studies have underlined the key areas where research is needed to improve the understanding of the association between DBPs in drinking water and adverse pregnancy outcomes (Villanueva et al. 2015). These include the use of cohort study design with the use of biomarkers and understanding the biological mechanism of action between DBP exposure and adverse pregnancy outcomes. Studies with large sample sizes are needed to achieve sufficient statistical power (Villanueva et al. 2014) and a better understanding of the pathways by which DBPs or contaminants cause birth outcomes (Ferguson et al. 2013). Developing countries must also form part of this assessment in this field in order to add to the existing knowledge.

The authors would like to thank Mrs Groebler, Librarian, for her assistance in searching and retrieving some of the articles. This study was sponsored by the South African National Research Foundation (NRF) Grant (SFH150625121049). The NRF did not play any role in the analyses, writing of the report, interpretation of data or decision to submit the manuscript. The views expressed in this article are those of the authors and do not reflect the views of the NRF. No conflict of interest is declared.