Health Stream Article - Issue 50 - June 2008

DBPs And Birth Defects

In recent weeks some Australian newsletters have carried articles about two newly published studies reporting associations between maternal exposure to disinfection byproducts (DBPs) during pregnancy and the risk of some birth defects. One study was conducted in Perth, Western Australia (1) and the other in Taiwan (2). Both studies were based on analysis of birth registry data with exposure to DBPs in tap water inferred on the basis of the residential address of the mother at the time of delivery.

Birth defects are defined as abnormalities affecting the structure, function or body metabolism which are present at birth. Such defects are diagnosed in about 2-3% of newborn infants, and may range in severity from fatal or life-threatening through to relatively mild and easily treated conditions. In developed nations, birth defects are responsible for about 20% of deaths in newborn babies. Major structural defects are usually recognised at or soon after birth, but the rate of diagnosed defects continues to increase as infants grow and develop, and more abnormalities become apparent. By one year of age, the rate of diagnosed defects increases to about 6%, and by age five the rate reaches about 10%.

Given the relatively low rate of occurrence of birth defects, it is not considered feasible to undertake prospective epidemiological studies as it would be necessary to recruit around 30 pregnant women for every expected case of adverse outcome that would be apparent at birth. Thus in order to investigate risk factors for birth defects, only retrospective designs such as cross-sectional studies or case-control studies can be undertaken.

Western Australian Study This cross-sectional study examined birth defects for deliveries occurring over a five year period from 2000 to 2004 in a study area which spanned the northern suburbs of the city of Perth. Water sampling was conducted at 47 locations over a different time period (April 2005 to March 2006), with six samples collected from each site during this interval. Samples were analysed for the four individual trihalomethanes (THMs). The numbers of total births and birth defects for women residing in the postcodes corresponding to the study area were obtained from state registries. The coverage of these registries includes birth defects in live births and stillbirths and pregnancies terminated due to foetal abnormalities. The available information included postcode of maternal residence at the time of delivery, maternal age, and socioeconomic status, however the date of birth was not available for birth defects cases due to privacy constraints. The birth defects data was obtained in 2007 and thus included any birth defects registered since birth until that time. Pregnancy terminations were excluded from the study. The study examined all birth defects combined, and the seven most common categories of birth defects as classified under the British Paediatric Association International Classification of Diseases 9th Revision (ICD-9).

Total THM (TTHM) exposures were estimated by averaging the measurements for the six samples at each collection site. Postcodes were aggregated into areas approximately corresponding to water supply areas, and each of the resulting eight areas was assigned as having low (less than 60 microgram/L), medium (between 60 and 130 microgram/L) or high average TTHMs (130 micrograms/L or above). Women were assigned a TTHM exposure level according to their postcode of residence at the time of delivery. Associations between DBP exposure and birth defects were examined using binomial logistic regression models. Estimates were adjusted for maternal age at the individual level and socioeconomic status at the postcode level.

A total of 20,870 live births were included in the study, with birth defects being recorded in 1,097 individuals. Comparison of maternal demographic characteristics between the three TTHM exposure levels showed the average maternal age to be slightly lower in the low TTHM area. A much wider range in socioeconomic status existed within the high TTHM area than within the low or medium TTHM areas. The high TTHM area contributed 63% of total births in the study, with the low TTTHM area contributing 14% and the medium TTHM area 23%. The outcome analysis (summarised in the following table) showed a significantly increased Odds Ratio for any birth defect (combined category) when women living in high TTHM areas were compared with those living in low TTHM areas (OR=1.22, 95% CI 1.01-1.48). When the seven different categories of common birth defects were examined, a statistically significant increase was seen for cardiovascular defects (OR=1.62, 95% CI 1.04-2.51).

Summary of adjusted Odds Ratios and 95% CIs for association of TTHM exposure and birth defects for the Western Australian study (1)

The authors comment that the TTHM levels examined in this study are higher than those seen most other published studies on this topic. The Australian Drinking Water Guidelines currently specify a guideline value of 250 micrograms/L for total THMs with no guideline values set for the individual compounds. A number of other countries have set lower guideline or standard values, while the World Health Organisation has set separate guideline values for the four individual THMs and recommended a fractionation approach for calculating TTHM limits. TTHM values recorded in the sampling program for this study ranged from a minimum of 36 micrograms/L in the low TTHM area to a maximum of 190 micrograms/L in the high TTHM area. In contrast to most disinfected drinking water supplies where chloroform is the dominant THM, the brominated forms are more abundant in waters in Perth due to the high bromine content of local groundwaters. Shallow groundwaters in Perth also contain high levels of organic matter due to the porous nature of the soils.

Taiwanese Study This was a cross-sectional study of births occurring during the three year period 2001-2003. The study area was restricted to five water supply regions which were stated to have only one type of water treatment plant and to use chlorination for disinfection. The number of water treatment plants included in the study is not given, although it is stated that 90% of the Taiwanese population is supplied with disinfected water from 200 treatment plants, while the remaining 10% use private wells. A total of 396,049 infants were included in the study population, representing 55% of all babies born in Taiwan during the study period. Birth records and information on birth defects was obtained from a national registry. This registry records information only on defects diagnosed up to 7 days after birth, and does not record defects present in foetuses lost before 20 weeks of gestational age due to spontaneous or induced abortion.

The study assessed the eleven most common categories of birth defects including two types of defects affecting the brain (anencephalus, hydrocephalus), three types of heart defect (ventricular septal defects, atrial septal defects, Tetralogy of Fallot), three types of defect of the renal and urinary tract (renal agenesis and dysgenesis, obstructive urinary tract defects, hypospadias), cleft palate, cleft lip and chromosomal defects. The registry also contained information on the infant’s sex, maternal age category, whether the birth was single or multiple, and several conditions relevant to the mother’s health (eg diabetes mellitus, anaemia, cardiac disease). The population density for each municipality was used as a measure of urbanisation.

Exposure of the mothers to DBPs during pregnancy was estimated from water company records of total trihalomethane (TTHM) measurements for each water treatment plant. Such measurements were required at last four times per year at each plant under water quality regulations. Exposure for each woman over the whole pregnancy was estimated using a weighted average of TTHM measurements for the relevant time period for the water treatment plant(s) supplying the mother’s place of residence. TTHM exposure for each pregnancy was classified into four categories; 0-4 micrograms/L (reference category), 5-9 micrograms/L, 10-19 micrograms /L and 20 or more micrograms/L. The prevalence of each type of birth defect and all defects combined was calculated and logistic regression was used to generate Odds Ratios.

There were no significant differences between women in the four exposure categories in terms of maternal age categories, ratio of male to female infants, prevalence of maternal diabetes or single vs multiple births. Population density was lower in the highest TTHM exposure category compared to the reference category suggesting a less urbanised population. There were no significant differences in these characteristics between the included and excluded populations.

A total of 2,148 births (0.5%) with one or more of the selected categories of birth defects were reported from the 369,049 births included in the study. In the final statistical analysis Odds Ratios were calculated with adjustment for maternal age, single or multiple birth and population density. The adjusted Odds Ratios are summarised in the following table.

Summary of adjusted Odds Ratios and 95% CIs for association of TTHM exposure and birth defects for the Taiwan study (2)

In discussing the results, the authors highlight the increased adjusted Odds Ratios for anencephalus (OR=1.96, 95% CI 0.94-4.07), ventricular septal defects (OR=1.81, 95% CI 0.98-3.35) and cleft palate (OR=1.56, 95% CI 1.00-2.41) seen for comparison of the highest exposure group with the reference (lowest exposure) group. However, only the association with cleft palate reached borderline statistical significance at the p=0.05 level. This was also the only outcome which showed a consistent dose-response trend across the exposure categories. Both anencephalus and ventricular septal defects showed decreased risks in at least one of the intermediate exposure categories compared to the reference category.

The authors also carried out a meta-analysis of this study and five previous studies which have examined the relationship between DBP exposure and some of the seven most common categories of birth defects. These studies differed in their design (three cross-sectional, two case-control), exposure contrasts (high versus low TTHMs, high versus zero THMs, chlorine dioxide treated versus undisinfected water supplies), and index of exposure (THM measurements or high versus low colour). The number of studies addressing each type of defect ranged from two to four. This analysis resulted in a significantly elevated summary risk estimate for ventricular septal defects (OR=1.25 95%CI 1.08-1.46) based on three studies. For hydrocephalus and anencephalus, summary ORs were also significantly increased but the studies showed underlying heterogeneity. For atrial septal defects, cleft lip with/without cleft palate and for cleft palate, the meta-analysis showed consistent evidence of no effect.

Limitations of these studies Both studies used registries as a source of data on births and birth defects, and in both cases the coverage of pregnancies and accuracy of registries was said to be of high quality. However the exposure of pregnant women to TTHMs was simply inferred from their place of residence at the time of birth, and women were not interviewed to determine whether they resided there throughout their pregnancy, whether they drank tap water, or the extent of dermal or inhalational exposure to tap water (which may contribute substantially to exposure to volatile DBPs). While the Western Australian study collected water samples from 47 different sites in the distribution system, these were not uniformly spread over the study area and may not have adequately reflected the variability in THM levels. In addition the water samples were taken in a different time period from the pregnancies and the lack of information on birth dates for birth defect cases prevented adjustment even for seasonal variations. From the methodology described in the paper it appears all pregnancies during the five year period in each of the eight geographic zones were assigned the same TTHM exposure level. The Taiwanese study recorded THM values during the same time period as the pregnancies but it appears the measurements were taken at water treatment plants and would not have accounted for changes in THM concentrations that may have occurred during travel through the distribution system. Exposure was averaged over the entire pregnancy rather than being restricted to the first trimester when congenital defects occur.

The database used in the Taiwanese study provided some information on maternal health status but both studies lacked important information on maternal smoking, alcohol consumption, and the use of vitamin supplements or medications. A number of medicinal drugs are known to cause birth defects including some drugs used to treat epilepsy, depression and bipolar disorder, anticoagulants, angiotensin converting enzyme inhibitors (used to treat high blood pressure) and isoretinoin (used to treat severe acne). In addition, a wide range of commonly used prescription and over-the-counter drugs including antibiotics, analgesics, and anti-inflammatories are believed to increase the risk of birth defects. Illegal drugs are also suspected to cause birth defects although this is difficult to verify given the association of illegal drug use with other risk factors such as alcohol, smoking, poor nutrition etc. While many pregnant women avoid medication use as much as possible during pregnancy, it is estimated that even in developed countries such as Australia at least one-third of pregnancies resulting in live births are un-planned, opening the possibility of inadvertent exposure in the first few weeks after conception but before the pregnancy is recognised. Significant structural birth defects in the major organ systems are most likely to arise during the first 60 days of pregnancy. Other known risk factors for birth defects include maternal health conditions such as diabetes, maternal infections, genetic conditions and alcohol exposure in early pregnancy.

While both of these new studies reported increased risks of cardiac defects, a previous review of the weight of evidence on adverse reproductive and developmental effects found no evidence of association of cardiac anomalies with DBP exposure on the basis of eight epidemiological studies conducted prior to that time (3). This review also noted that animal toxicological studies of DBPs had shown adverse effects of foetuses only at levels that produced maternal toxicity, and that anatomical malformations had not been observed.

Overall, the findings of both of these new studies is limited by their cross-sectional design, inaccuracies in DBP exposure assessment and the lack of important information on other relevant maternal exposures which may affect the risk of birth defects. While it may be argued that lack of precision in assessing both DBP exposure and non-water risk factors is likely to be non-differential and thus result in a reduced Odds Ratio, the “ecological fallacy” is well known in field of epidemiology and it cannot be assumed that exposures assessed on the collective level will reflect individual exposures. Knowledge of the relationship (if any) between DBP exposures and birth defects is likely to be advanced only by studies of more analytical design which include detailed information on maternal exposures and health status at the individual level during the critical time window of early pregnancy.

(1) Chisholm K, Cook A, Bower C and Weinstein P (2008) Risk of birth defects in Australian communities with high brominated disinfection by-product levels. Environmental Health Perspectives doi:10.1289/ehp/10980
(2) Hwang BF, Jaakkola JJK, Guo HR (2008) Water disinfection by-products and the risk of specific birth defects: a population-based cross-sectional study in Taiwan. Environmental Health 7(23) doi:10.1186/1476-069X-7-23
(3) Tardiff, R. G., M. L. Carson, et al. (2006). Updated weight of evidence for an association between adverse reproductive and developmental effects and exposure to disinfection by-products. Regulatory Toxicology & Pharmacology 45(2): 185-205.