Exposure to Trihalomethanes through Different Water Uses and Birth Weight, Small for Gestational Age, and Preterm Delivery in Spain

Background: Evidence associating exposure to water disinfection by-products with reduced birth weight and altered duration of gestation remains inconclusive. Objective: We assessed exposure to trihalomethanes (THMs) during pregnancy through different water uses and evaluated the association with birth weight, small for gestational age (SGA), low birth weight (LBW), and preterm delivery. Methods: Mother–child cohorts set up in five Spanish areas during the years 2000–2008 contributed data on water ingestion, showering, bathing, and swimming in pools. We ascertained residential THM levels during pregnancy periods through ad hoc sampling campaigns (828 measurements) and regulatory data (264 measurements), which were modeled and combined with personal water use and uptake factors to estimate personal uptake. We defined outcomes following standard definitions and included 2,158 newborns in the analysis. Results: Median residential THM ranged from 5.9 μg/L (Valencia) to 114.7 μg/L (Sabadell), and speciation differed across areas. We estimated that 89% of residential chloroform and 96% of brominated THM uptakes were from showering/bathing. The estimated change of birth weight for a 10% increase in residential uptake was –0.45 g (95% confidence interval: –1.36, 0.45 g) for chloroform and 0.16 g (–1.38, 1.70 g) for brominated THMs. Overall, THMs were not associated with SGA, LBW, or preterm delivery. Conclusions: Despite the high THM levels in some areas and the extensive exposure assessment, results suggest that residential THM exposure during pregnancy driven by inhalation and dermal contact routes is not associated with birth weight, SGA, LBW, or preterm delivery in Spain.


Study area
The study area in the Asturias cohort comprises 9 municipalities (Avilés, Castrillón, Corvera, Gozón, Muros de Nalón, Pravia, Cudillero, Soto del Barco and Illas) accounting for approximately 160.000 inhabitants in 2010, supplied by surface water through 11 distribution networks. The Gipuzkoa cohort comprises 2 counties (Goierri, Alto Urola) with 25 municipalities and 90.000 inhabitants, supplied by surface water from 3 reservoirs through 26 distribution networks. The Sabadell cohort is comprised by a single municipality with approximately 200.000 inhabitants and a single water distribution network from a surface source. The Valencia cohort includes 34 municipalities and part of Valencia City, accounting for approximately 300.000 inhabitants supplied with water from different sources (surface, ground and mixed) and 70 distribution networks. The Granada cohort includes 40 municipalities and part of Granada city, representing 475.000 inhabitants supplied by ground and surface water through 33 distribution networks. Chlorine is the main disinfectant used in all the study areas during the study period.

Swimming pool sampling campaigns
Public facilities with swimming pools open to the public were identified in the study areas, including sports facilities with indoor and/or outdoor pools and outdoor leisure swimming pools. One of the sampling campaigns was conducted in summer (July 2009) in order to sample simultaneously indoor and outdoor pools for comparison, since outdoor pools are only open in the summer season (mid June to mid September). Two criteria were followed to select swimming pools: municipality-owned swimming pools (since these tend to be the most frequented pools in Spain), and/or facilities with both indoor and outdoor pools. In Asturias, 3 sports complexes with swimming pools open to the public were identified in the study municipalities, 2 in Avilés and 1 in Castrillón. All of them were indoor pools sampled for the study in July 2009. A sample in the adult and infant pool was collected in the 2 swimming pools in Avilés, resulting in a total of 5 samples. Sabadell had 9 indoor swimming pools open to the public (7 of them had also an outdoor pool) and 6 outdoor swimming pools. The 9 sport facilities with indoor swimming pools were contacted and water samples were collected in

Procedure to model trihalomethane levels during pregnancy
In a first stage, bivariate descriptive analyses were conducted to identify the variables influencing THM levels (see variables in Supplemental Material, Table 1). Secondly, the THM variable was transformed when necessary to obtain a normal distribution and bivariate linear regressions were conducted. Variables associated with THM levels with statistical significance (p-value<0.05) or variables a priori relevant to THMs were included simultaneously in linear regression models and generalized additive models (GAMs) with different combinations of variables. GAM models were used to fit a smoothed function of month that was used to predict levels for months without observations and also a smooth function of coordinates. Both the R-squared and the Akaike information criteria were used to select the final model (see Supplemental Material, Table 2). The models with highest R2 and lowest AIC were the selected ones. We first modelled total THMs following this procedure.
For chloroform, bromodichloromethane, dibromochloromethane and bromoform models the same procedure was followed, with the difference that we started with the variables included in the total THM model. Non-significant variables were removed and relevant variables for temporal or geographical variability, not initially included, were tested and retained if appeared to be statistically significant. a. The 34 sampling sites were chosen to represent all municipalities (N=25) and distribution networks (N=26) in the Gipuzkoa cohort. Two sampling points were identified in municipalities with >5000 inhabitants (8 municipalities) and one sampling point in the rest.
b. Coordinates of the actual sampling points were estimated by geocoding the sampling address both from our own and regulatory measurements c. Population center in Valencia is the result of grouping the distribution systems (N=70) in categories with homogeneous water quality and urbanicity. All the study municipalities (N=34) are represented in the 49 population centers. Measurements of THMs were available in all the distribution systems and municipalities.