Chlorine is the most common disinfectant used in drinking water and is a cost-effective method to prevent the spread of waterborne diseases. In the 1970s, it was recognized that chlorine reacted with naturally occurring organic matter in surface waters to produce chemical byproducts.1 Of these, trihalomethanes (THMs) occur in the highest concentrations and, until recently, have been the only routinely monitored byproduct in public water supplies in the United States and Canada.
In the past decade, several epidemiologic studies have reported relationships between THMs and adverse reproductive outcomes. Among the studies that have specifically examined intrauterine deaths, all but one2 have reported increased risks of spontaneous abortion3,4 or stillbirth5,6 in relation to total THMs. Two studies have reported on the risk of intrauterine death associated with individual THM compounds,4,7 and in both studies, higher risks were observed with exposure to bromodichloromethane (BDCM) than the other THM compounds. As well, a previous study conducted in Nova Scotia, Canada, reported higher risks associated with stillbirths resulting from asphyxia-related deaths and exposure to THMs.7
Although the conclusions of some of these studies are similar, they provide only suggestive evidence of a causal relationship. Only 2 studies of intrauterine deaths3,4 collected individual-level data on their subjects, and both of these were studies of spontaneous abortion rather than stillbirth; the remainder relied on existing databases for all outcome and risk factor measurement and thus had limited ability to control for potential confounders. An expert advisory panel from the Environmental Protection Agency (EPA) has recommended that future epidemiologic studies on chlorination byproducts make efforts to improve on the exposure assessment by collecting individual-level information on exposure to residential water and on byproduct measures from residential tap water.8 With this in mind, we undertook a population-based case-control study to examine the relationship between THM exposure during pregnancy and stillbirth.
The Research Ethics Boards at all of the participating hospitals in Nova Scotia and Eastern Ontario, Canada, reviewed and approved this study, and each subject provided informed consent.
We identified all stillbirths occurring in Nova Scotia and Eastern Ontario (which includes 13 counties in the eastern most part of the province) between July 1, 1999, and December 31, 2001, through population-based perinatal databases. Cases were women who had stillborn infants, defined as death of a fetus weighing 500 g or more at delivery. Women whose pregnancy was terminated for a fetal anomaly were excluded. Controls were women who delivered a live-born infant during the same 3-month time period as the cases and were randomly selected from the perinatal database from the same study area. Approximately 4 times the number of controls as cases were selected to maximize statistical power within the constraint of limited case availability. Women were eligible to participate if they lived in the study area for at least the first 5 months of their pregnancy, delivered in the study area, and were residents in the study area at the time of recruitment. The attending physician for each subject was asked to screen the patient for eligibility and to exclude those for whom participation in the research would possibly cause emotional stress. We delayed initial subject contact for 6 months after delivery because we felt that women who had experienced a stillbirth might be reluctant to talk about their pregnancy soon after delivery.
Subjects completed a telephone interview that focused on water use behaviors at approximately 3–4 months’ gestation and measurement of other stillbirth risk factors that may confound the relationship of interest. Information collected on exposures to THMs included consumption of beverages made with tap water, description of any water filters used, consumption of bottled water, and length of time spent showering and bathing.
Residential tap water samples were collected from all subjects who lived in an area served by a public water supply. Ideally, water sampling would have occurred at approximately 15 weeks’ gestation to give an estimate of exposure during the first and early second trimester of pregnancy. As Klotz et al.9 had done in a previous case-control study of chlorination byproducts, residential water was sampled 1 year later to approximate the byproduct level during the same season as the critical point in pregnancy. Water samples were not collected for subjects with a private well source because the THM levels were assumed to be zero. Subjects were provided with simple written instructions as well as coolers containing an ice pack and 2 bottles. Overnight courier delivered residential samples collected by subjects to the Analytical Services Unit of Queen's University. A purge and trap gas chromotographic/mass spectrometer method was used for THM analyses according to standard methods (EPA Method 624). If a subject had moved after the fourth month of pregnancy, a surrogate sample was collected from a public building close to the subject's former residence.
At the completion of subject recruitment, the hospital records of stillbirth cases were reviewed by an obstetrician and a neonatologist. The categories for the underlying cause of death were adopted from previously described methods.10,11 The categories included congenital anomalies, immaturity, intrauterine asphyxia, infections, other specified deaths, and unexplained deaths. Asphyxia deaths included complications with the umbilical cord, the placenta, and labor and delivery.
Measurement of Exposure
To account for water use behavior, measures were developed that integrated THM levels from residential tap water with estimates of daily exposure to water through ingestion, inhalation, and absorption. Ingestion was defined as the total liters of water consumed daily through cold and hot tap water-based drinks. Water consumed at work was also included in the total ingestion calculation. Because a water sample was not collected from the workplace, THM levels were estimated from a seasonally adjusted average of study samples obtained for all subjects for each municipality. For workplace municipalities without sampling data (n = 21 subjects), an estimate of THMs was obtained from routine monitoring data. For subjects who used a carbon filter (for example, Brita, Toronto, Ontario), we applied a 50% reduction in THM intake through cold tap water-based drinks.12 A reduction of 70% was applied to the THM ingestion estimate for boiled hot water drinks.13,14 We assumed that bottled water would not contribute to the THM exposure by ingestion, and it was therefore not included in the analyses. The number of minutes spent showering and bathing daily was ascertained to account for exposure from inhalation and dermal absorption. Lastly, an estimate of total daily THM exposure was created to incorporate ingestion, showering, and bathing. This total exposure metric assumed an equivalency of absorbed dose from the different sources of exposure such that 1 L of ingested water was equivalent to a 5-minute shower15 and a 15-minute bath.16
The odds ratio was used as an estimate of relative risk and, for ease of presentation, the results are described in terms of relative risk. We calculated odds ratios (ORs) and 95% confidence intervals (CIs), adjusted for potential confounders, using logistic regression. Information on potential confounding factors included factors pertaining to pregnancy history, information pertaining to the index pregnancy, information on jobs held during pregnancy, socioeconomic factors, and other potential risk factors (for example, exposure to pesticides during pregnancy and smoking during pregnancy). Potential confounders were entered into a multivariate model along with the THM exposure variables. Factors were kept in the model and considered confounders if the change in effect size for the THM exposure variable was 10% or greater. In addition, all multivariate models were adjusted for age and province. Analyses were conducted separately for total THMs, chloroform, and BDCM and for subgroups according to the cause of death categories. The 2 cause of death subgroups of primary interest in relation to exposure to THMs were asphyxia-related deaths and unexplained deaths.
We identified 271 stillbirths and 814 potential controls in the perinatal databases. The attending physician responded for 234 (87%) cases and 684 (84%) controls. Among those with a physician response, participation was deemed inappropriate for 12% of cases and 6% of controls. Another 8% of cases and 8% of controls were excluded because subject contact information was not known or because the physician reply was received after subject recruitment was completed. We sent an invitation to participate to mothers of 188 stillbirth cases and 589 live birth controls who met the eligibility criteria. Of these, a positive reply card was received by the study investigators, and an interview was completed for 60% of cases (n = 112) and from 68% of controls (n = 398). Among the cases, 25 subjects (13%) indicated they did not want to participate and a reply card was not received by the remaining 50 cases (27%). Among the eligible controls, 52 subjects (9%) indicated they did not want to participate and the reply card was not received by the remaining 127 controls (22%). One case and 12 controls sent a positive reply card but either moved or were lost to follow up before the interview and water collection took place.
Cause of death for the stillbirth cases was assigned in 103 subjects (9 subjects did not give consent for their records to be reviewed). The majority of stillbirths were unexplained (51%). The remaining causes of death were: asphyxia-related disorders, 21%; congenital anomalies, 10%; other specified disorders, 7%; immaturity-related disorders, 6%; and infections, 5%.
Table 1 provides the univariate analyses for selected demographic factors, potential confounders, and water use characteristics. Seventy percent of cases and 62% of controls had a chlorinated household water supply. The mean total THM level measured in residential water samples for subjects with a chlorinated water supply was 57 μg/L among cases and 55 μg/L among controls. The maximum total THM, chloroform, and BDCM values among the subjects were 318 μg/L, 315 μg/L, and 21 μg/L, respectively. A surrogate sample was collected for 54 subjects whose residence at the time of recruitment was different than their residence at 4 months’ gestation.
Table 2 presents the crude and adjusted risks for stillbirth and level of THMs as measured from the residential water samples. Subjects whose total THM level was 80 μg/L or higher had a 2-fold greater risk of stillbirth (OR = 2.2; 95% CI = 1.1–4.4) relative to women whose residential water did not contain THMs. For total THMs, chloroform, and BDCM, the risk of a stillborn infant was increased at most of the exposure levels above zero, but the risks did not increase in a monotonic fashion. Nevertheless, the largest risks were consistently seen in the highest category of exposure.
Table 3 provides results for the total exposure metric, which incorporates residential THM level, tap water ingestion at home and at work, and showering and bathing activities. Among the subjects with some THM exposure, quintiles were formed based on the distribution among controls. A quintile representation was selected for comparability to the residential exposure level with respect to the proportion of controls in the top exposure category. Subjects could be in the top quintile because of a high THM level or because of high values for ingestion or showering/bathing. For example, a subject would be at the cut point for the highest quintile of total THM exposure with a residential THM level of 50 μg/L, consumption of 4 glasses of tap water per day, and showering 5–10 minutes 5 times per week. Using the total exposure metric for THMs, subjects in the highest quintile of exposure had a 2.4-fold risk of stillbirth (95% CI = 1.2–4.6). For chloroform, the highest quintile of exposure was associated with an odds ratio of 2.0 (95% CI = 1.0–4.0). The highest quintile for the total exposure metric to BDCM was associated with a 2.5% increase in risk of a stillbirth (95% CI = 1.3–4.9) compared with women who were not exposed to BDCM.
The individual contribution of tap water ingestion and residential THM level on stillbirth risk was further explored through examination of the joint effects of cold water ingestion and total THM level (Table 4). Women who consumed 5 or more cold tap drinks per day with total THM levels of 50 or more μg/L had a 4-fold risk of stillbirth (95% CI = 1.4–11) compared with subjects with a total THM level of zero and low consumption of tap water. Women in the intermediate THM category (1–49 μg/L) who consumed 5 or more tap water drinks a day were also at increased risk of having a stillborn infant (OR = 2.4; 95% CI = 1.1–5.3). Adjustment for showering and bathing minutes did not alter these results (results not shown).
The joint effects of time spent bathing or showering and residential THM level on the risk of stillbirth is shown in Table 5. In this analysis, bathing minutes were divided by 3 for equivalency to showering minutes. Among women with a residential THM level of 50 μg/L or more, the stillbirth risk was increased only among those with more than 15 minutes showering/bathing per day (OR = 2.6; 95% CI = 1.1–5.8). Adjustment for tap water ingestion did not alter these results (results not shown).
Table 6 shows the relationship between total THMs, as measured from residential water samples, and the risk of unexplained stillbirths and stillbirths resulting from asphyxia-related disorders. The risks observed in each of the stillbirth subgroups are similar to those observed for all stillbirths combined, but the small number of cases in the subcategories resulted in low power and wide confidence intervals.
Our data support an association between exposure to chlorination byproducts and stillbirths. A residential total THM level of 80 μg/L or more was associated with a doubling of stillbirth risk compared with those with zero exposure. Women in the highest quintile of total exposure to BDCM had 2.5 times the risk of stillbirth compared with those with zero exposure. This was the first study of stillbirths related to exposure to chlorination byproducts that examined individual-level exposure and residential water samples. The magnitude of risk observed in this study is somewhat larger than risk estimates observed in the previous Nova Scotia study based on existing databases.6 In this study, we were able to quantify exposure more precisely, which likely improved our exposure assessment and reduced misclassification. The largest odds ratio (4.0; 95% CI = 1.4–11) was observed for those consuming 5 or more glasses of tap water per day with a residential THM level of 50 μg/L or more. The combined effect of showering/bathing and residential THM level suggests that exposure by inhalation and absorption may also contribute to stillbirth risk.
Previous studies of spontaneous abortion are also relevant to this discussion, because they may have a similar biologic response with respect to the toxicity of chlorination byproducts. Two studies have examined risk for spontaneous abortion in relation to the level of THMs in the water distribution system. Our results are consistent with Waller et al.4 with respect to total THM level and number of cold tap water drinks consumed per day. In both studies, consuming 5 or more tap water drinks per day increased risk when total THM level was high, whereas consuming less than 5 tap water drinks per day was not associated with increased risk, regardless of THM level. However, the results of this study and the study by Waller et al.4 differ with respect to the effect of showering or bathing. Although neither found that adjustment for showering or bathing affected the results for ingestion, our study showed some evidence of an independent effect of showering or bathing when the THM level was elevated. In a study by Savitz et al.,3 increased risk of spontaneous abortion was seen only in the highest sextile of THM concentration.
Analyses by cause of death categories may help in understanding biologic mechanisms related to the potential effect to the fetus from chlorination byproducts. In this study, the point estimates for the asphyxia-related deaths (of which abruptio placentae was the most common condition) were higher than for unexplained stillbirths, but the numbers were small and the confidence intervals wide. This lends some support to the results found in a previous Nova Scotia study in which risk estimates associated with THM exposure were larger for asphyxia-related deaths compared with unexplained stillbirths and stillbirths overall.2,7 It has been observed previously that defects in the methionine–homocysteine metabolic pathway, which can be the result of low folate levels and result in elevated homocysteine levels, may be a contributing factor for abruptio placentae.17 A study by Alston18 found that chloroform inhibited methionine biosynthesis in cell culture. It may be possible that folate level modifies the effect of chlorination byproducts on the risk of stillbirth. More than 90% of subjects in this study took vitamin supplements during pregnancy, so we were unable to test this hypothesis. These observations and theories regarding a potential causal mechanism need to be tested in both toxicologic and epidemiologic research.
The rate of stillbirth is low (approximately 4 per 1000 births), which makes a prospective design impractical. The case-control design requires far fewer subjects but, as a result of the retrospective manner in which exposure is assessed, it has the potential for recall bias. In this study, subjects were asked about exposure to many factors, not only water-related factors, and they were not aware of our specific hypotheses. Furthermore, the measurement of residential level of chlorination byproducts is not subject to recall bias, and women did not know the level of chlorination byproduct in their residential water at the time they were questioned about consumption and use. The pattern of risk that we observed helps to rule out recall bias as an explanation for the increased risk. For example, in the absence of THMs, there was no evidence that drinking 5 or more glasses of water per day was associated with increased stillbirth risk.
Because information was collected retrospectively, sampling water during the pregnancy was not possible. The sampling approach used in this study (sampling 1 year after the “critical” time period during pregnancy) takes advantage of seasonal patterns of disinfection byproduct occurrence. To verify the validity of this sampling method, some water samples were retested from the same sampling location 1 year after the first sample was collected. For total THMs, the original sample and retest sample had a correlation of 0.87 and a mean difference of 10 μg/L. This supports the validity of the “1-year later” sampling method. Other sources of misclassification in THM levels are the estimates of water ingestion at the workplace, and the surrogate samples collected when women had moved.
A concern with exposure representation in this study is that the referent category for all analyses contains subjects served by a private well, and therefore the excess risk observed may represent a private versus public water effect. We repeated the analyses restricted to those served by a public water supply and observed a similar pattern of risk, but with lower risk estimates for the highest exposure category and wider confidence intervals (eg, adjusted OR for the highest quintile of THMs, using the total exposure metric, was 1.9, 95% CI = 0.8–4.3). The analysis strategy considered the potential confounding effects of known and suspected risk factors for stillbirths. Control for these factors had little impact on relative risk estimates for THM exposure because very few factors were both strong risk factors for stillbirth and factors related to THM exposure.
Low subject response rates are a concern. We feel the response rate is unlikely to have affected the validity of the results, because potential participants were unlikely to be aware of either the specific study hypothesis or the level of chlorination byproducts in their water. Nonresponse was more likely the result of the restriction of having to invite potential subjects through a letter from their physician rather than through more active recruitment. Response rates for controls in rural areas may be higher than those in urban areas, and thus bias results through overrepresentation of rural residents with low THM exposure in the control group (because rural residents are more likely to not have municipal water service). Although overall participation rates were higher for rural than for urban participants, the ratio of rural–urban participation rates was similar among both cases and controls. Therefore, we do not think our results are explained by a higher response rate among controls from areas with low THM levels.
The lack of a monotonic dose–response relationship in the risks related to residential THM levels, or with the total THM exposure metric, diminishes the likelihood of a causal relationship. Even so, the highest risks were consistently observed in the highest exposure categories.
These results support evidence from other studies that exposure to chlorination byproducts through ingestion, showering, and bathing may contribute to the risk of stillbirth. In this study, mothers with high total exposure to total THMs had over twice the risk of stillbirth compared with those who were not exposed. If this relationship is causal, consuming filtered or bottled water will not completely eliminate risk because showering and bathing contributed to the overall exposure. A decrease in population exposure will be most effectively achieved at treatment plants rather than the household level. However, changes in water treatment should be approached carefully. The disinfection process can result in the formation of a number of byproducts besides THMs, including haloacetic acids, haloacetonitriles, and haloketones. THMs may be acting as a surrogate for the occurrence of other byproducts, which may in fact be the causal factor for an adverse reproductive event. Efforts to reduce specific chlorination byproducts must therefore be done without increasing the others.
We thank the study coordinators, Adelia Trenchard and Celine Zakos, and the clinical collaborators, Graham Smith, Kent Dooley, Ann Houstoun, and Ted Luther. We are grateful for assistance from the Reproductive Care Program of Nova Scotia, the Perinatal Partnership Program of Eastern and Southeastern Ontario, and the Analytical Services Unit of Queen's University. We thank all the participating physicians, hospitals, and study participants.
1.Bellar TA, Lichtenberg JJ, Kroner RC. The occurrence of organohalides in chlorinated drinking waters. J Am Water Works Assoc.
2.Bove FJ, Fulcomer MC, Klotz JB, et al. Public drinking water contamination and birth outcomes. Am J Epidemiol
3.Savitz DA, Andrews KW, Pastore LM. Drinking water and pregnancy outcome in Central North Carolina: source, amount, and trihalomethane levels. Environ Health Perspect
4.Waller K, Swan SH, DeLorenze G, et al. Trihalomethanes in drinking water and spontaneous abortion. Epidemiology
5.Aschengrau A, Zierler S, Cohen A. Quality of community drinking water and the occurrence of late adverse pregnancy outcomes. Arch Environ Health
6.Dodds L, King W, Woolcott C, et al. Trihalomethanes in public water supplies and adverse birth outcomes. Epidemiology
7.King WD, Dodds L, Allen AC. Relation between stillbirth and specific chlorination by-products in public water supplies. Environ Health Perspect
8.EPA Panel Report and Recommendations for Conducting Epidemiological Research on Possible Reproductive and Developmental Effects of Exposure to Disinfected Drinking Water.
Research Triangle Park, NC: Environmental Protection Agency; 1998.
9.Klotz JB, Pyrch LA. Neural tube defects and drinking water disinfection by-products. Epidemiology
10.Cole S, Hartford RB, Bergsjo P, et al. International collaborative effort (ICE) on birth weight, plurality, perinatal and infant mortality III: a method of grouping underlying causes of infant death to aid international comparisons. Acta Obstet Gynecol Scand
11.Alberman E, Blatchley N, Botting B, et al. Medical causes on stillbirth certificates in England and Wales: distribution and results of hierarchical classification tested by the Office for National Statistics. Br J Obstet Gynaecol
12.Fit to drink? Consumer Reports
13.Kuo HW, Chiang TF, Lo II, et al. VOC concentration in Taiwan's household drinking water. Sci Total Environ
14.Wu WW, Benjamin MM, Korshin GV. Effects of thermal treatment on halogenated disinfection by-products in drinking water. Water Res
15.Weisel CP, Jo WK. Ingestion, inhalation and dermal exposures to chloroform and trichloroethene from tap water. Environ Health Perspect
16.Kerger BD, Schmidt CE, Paustenbach DJ. Assessment of airborne exposure to trihalomethanes from tap water in residential showers and baths. Risk Anal
17.Ray JG, Laskin CA. Folic acid and homocyst(e)ine metabolic defects and the risk of placental abruption, pre-eclampsia and spontaneous pregnancy loss: a systematic review. Placenta
18.Alston TA. Inhibition of vitamin B12-dependent methionine biosynthesis by chloroform and carbon tetrachloride. Biochem Pharmacol