Selection of sites and participant recruitment.
Right from the Start (RFTS) was a prospective cohort study of drinking water DBPs and pregnancy health conducted in three metro-politan areas of the United States (Savitz et al. 2005
). We selected these locations to include a wide range of individual DBP species and summary measures such as the sum of TCM, BDCM, DBCM, and TBM (ΣTHM). Sites 1 and 3 had moderate levels of chlorinated and brominated DBPs, respectively, and we chose them because they used chloramination rather than free chlorine for terminal disinfection. Chloramination results in minimal additional DBP formation within the distribution system; therefore, we would have expected all of the study participants within the same site to have similar tap water DBP concentrations for samples collected within the same week (Singer 1994
). Site 2 used free chlorine for the distribution system, but THM levels were so low that all consumers were exposed to low THM levels from residential tap water.
To be eligible to participate in the RFTS study, women had to be ≥ 18 years of age, reside and remain in one of the three metropolitan areas, use public drinking water, be able to speak and write English or Spanish, not have used assisted reproductive technology, and be trying to become pregnant or pregnant at < 12 weeks of gestation, with the intent to carry the pregnancy to term (Promislow et al. 2004
). Postpartum women who had participated in the RFTS study and were at least 30 days past delivery, not pregnant at the time of screening and enrollment, still residing in the study areas, and using public drinking water were eligible for the study of blood THMs. The institutional review boards at the University of North Carolina–Chapel Hill, University of Tennessee, and University of Texas approved the study protocols, and participants gave informed consent. Among the total of 238 women that were eligible and agreed to participate, 153 (64%) provided blood and water samples [see Supplemental Material, Figure 1
)]. The participation rate was 76% for site 1 and 57% for sites 1 and 2. To better assess seasonal variation in DBP levels, we collected water and blood samples from a subset of women in the summer and winter from site 1 (n
= 29) and site 3 (n
= 2). Blood and water samples were collected from January to March of 2004 (winter) for site 1, June to August of 2004 (summer) for sites 1–3, and December 2004 to January 2005 (winter) for site 3.
Blood and water sample collection and analysis.
Trained personnel scheduled morning home visits to collect blood and tap water samples before the participants had any contact with water. After signing a consent form, trained technicians (phlebotomists) collected a 10-mL blood sample from each participant via venipuncture into gray-top glass tubes (Vacutainer® Becton Dickinson, Franklin Lakes, NJ) that were specially treated before use to remove background THM contamination (Cardinali et al. 1995
). We mixed the blood samples to dissolve the anticoagulant immediately after the blood draw. The technicians collected a 12-mL water sample during the same home visit from a nonaerated, cold water tap. We kept all blood and water samples in coolers until they were shipped to the Centers for Disease Control and Prevention for analysis. We collected a total of 184 blood and water samples from 153 study participants. Seventy-four women provided blood samples in summer only, 48 in winter only, and 31 in both summer and winter. We excluded blood and water samples (n
= 4) from two participants from the analysis because of laboratory data quality concerns, and we excluded the water and blood samples for another participant who was exposed through a key water-use activity within 1 hr of sampling. In addition, we did not examine four water samples because of unacceptable headspace volume and/or freezing of vials. A total of 179 blood samples and 175 water samples from 150 women were available for analysis [see Supplemental Material, Figure 1
Isotope-dilution–based quantification of THM concentrations in tap water and blood samples was accomplished using solid--phase microextraction/gas chromatography (SPME/GC) with mass spectrometry (MS) (Cardinali et al. 2004
) and high-resolution MS (Bonin et al. 2005
), respectively. We added stable isotopically labeled analogs of the compounds of interest to 3 g blood and 5 mL water and sealed each sample in a 10-mL headspace vial. We heated (30°C for blood and 50°C for water) and agitated (350 rpm for blood and 500 rpm for water samples) samples using a CTC CombiPal® SPME autosampler (LEAP Technology, Carrboro, NC) to facilitate extraction of volatiles from the sample headspace onto an SPME fiber (Carboxen/PDMS, Supelco, Bellefonte, PA). After extraction, we inserted the fiber into a hot GC (5890 Series II; Agilent Technologies, Santa Clara, CA) inlet to desorb volatile compounds that were resolved chromatographically and then quantified in a high-resolution MS (Thermo Finnigan MAT 95; Thermo Finnigan, San Jose, CA) for blood and a quadrupole MS for water (Trace MS; Thermo Finnigan). Final quantification was based on daily seven-point calibration curves, and we normalized the concentrations according to sample weight. The limit of detection (LOD) in water was 0.93 µg/L for TCM, 0.21 µg/L for BDCM, 0.49 µg/L for DBCM, and 0.15 µg/L for TBM. The LOD in blood for TCM was 2.2 ng/L, 0.24 ng/L for BDCM, 0.21 ng/L for DBCM, and 0.58 ng/L for TBM. Out of 175 water samples, 5% were below the LOD for TCM, 2% for BDCM, 3% for DBCM, and 18% for TBM. Out of 179 blood samples, 6% were below the LOD for TCM, 8% for BDCM, 18% for DBCM, and 59% for TBM. Concentrations below the LOD were replaced with LOD/√
-2 (Hornung and Reed 1990
) for the analyses.
The participants self--administered a water-use activity diary 24 hr before the home visit, and we reminded them not to have any contact with water for at least 4 hr before their home-visit appointment. The 24-hr diary included information on water consumption practices (e.g., use of filters and other point-of-use devices), time, duration and location of showering and/or bathing, time spent bathing children, time spent washing dishes (and glove use), use of swimming pools, and use of fans and opening of windows while showering/bathing one’s self or children. Sociodemographic data collected from the main epidemiological study (Savitz et al. 2005
) were also available for this population [see Supplemental Material, Table 1
As part of the diary, we asked study partici-pants how many bottles of water and glasses/cups of cold tap water, hot tap water, and tap-water–based beverages (including juice, coffee, tea, and other beverages made from tap water) they consumed each day. We also asked that participants define their glass or cup sizes according to three options: small (0.1–0.3 L), medium (> 0.3–0.6 L), or large (> 0.6–1.0 L) for cold tap water beverages and small (0.1–0.3 L), medium (> 0.3–0.5 L), or large (> 0.5–0.7 L) for hot tap water beverages. We used the midpoint for each size range to estimate water consumption in ounces per day. We converted bottled water intake (spring water, mineral water, distilled water, sparkling water, or any water purchased in bottles or plastic jugs or obtained from a water cooler) to liters based on reported container sizes: small (8–12 ounces), medium (14–24 ounces), and large (26–34 ounces).
To help assess the primary determinants of blood THM levels, we developed a total daily exposure metric based on the main activities that impact ingestion, inhalation, and dermal absorption. We used six activities to calculate the daily exposure metrics for the 24 hr before sampling: a
) total tap water intake (liters), b
) total time showering/bathing themselves (minutes), c
) total time showering/bathing children (minutes), d
) total post-shower/bathroom time (minutes), e
) total time washing dishes (minutes), and f
) total time swimming (minutes). We summed intake of tap water and tap-water–based beverages to estimate ingestion exposure to THMs, and we used the reported other activities to estimate noningestion exposures. We applied a reduction of 70% in THM levels to the ingestion estimate for hot beverages, a 50% reduction for point-of-use filtration to filtered tap water, and a 50% reduction to those who reported using gloves while washing dishes (Forssén et al. 2007
; Krasner and Wright 2005
). We did not include bottled water consumption in the analy-sis because it typically contains very low levels of THMs (Weinberg et al. 2006
). Although we did not integrate data on the use of fans and the opening of windows/doors during showering/bathing because of the uncertainty associated with these specific practices on DBP levels (e.g., post-shower/bath levels), we performed sensitivity analyses (assuming 75% reduction) to assess the potential impact of these exposure modifying factors on the total estimate of ΣTHM exposure.
To integrate equivalent THM dose contributions from different exposure routes, we calculated a total exposure metric based on the summation of ingestion and noningestion activities. To allow for a common metric across disparate activities, we used liter--equivalents based on human biomonitoring data collected during controlled dermal or inhalation studies of TCM (Kerger et al. 2000
; Weisel and Jo 1996
) and previously applied in epidemiological studies (Dodds et al. 2004
; King et al. 2004
). The equivalency scores, including reductions applied to the afore-mentioned exposure modifying factors, were based on a presumed dose equivalency of 1 L total tap water intake, 5-min shower/bath, 15-min shower/bath for children, 15-min post-shower/bath time spent in the bathroom, 15 min of washing dishes by hand, and a 5-min swim. We examined ingestion and noningestion THM equivalency score tertiles in relation to blood THM concentrations. We restricted this analysis (n
= 150) to the first reported water diary entry for women with both summer and winter measurements (68% summer, 32% winter). Given potential toxicokinetic differences between specific compounds, we performed sensitivity analyses to assess the effect of estimated equivalencies (e.g., a 2-min shower/bath = 1 L ingested water) on the total exposure metric results.
We conducted statistical analyses using SAS statistical software (version 9.2; SAS Institute Inc., Cary, NC). We calculated descriptive statistics for blood, tap water, and sociodemographic characteristics of the study participants. We defined ΣTHM as the sum of TCM, BDCM, and DBCM, and TBM, concentrations in water (micrograms per liter) and blood (nanograms per liter). We defined brominated high-resolution THMs as the sum of BDCM, DBCM, and TBM in these two media. We did not weight the sums according to bromide content. Descriptive tests (skewness, kurtosis), histograms, and normal probability plots revealed deviations from a normal distribution for DBPs in blood and water [see Supplemental Material, Figures 2 and 3
)]; therefore, data were log10
transformed for the regression models and analysis of variance (ANOVA). We calculated the percentages of the brominated species in tap water and blood samples using the geometric means (GMs) for individual THM species and the ΣTHM GM for each site and season. We used Spearman rank correlation coefficients (rS
) to quantify the correlation between tap water and blood ΣTHM concentrations, between sites and seasons, and across individual THMs. We used paired t
-tests to compare mean water and blood THM concentrations between different seasons of sample collection. Because of the small sample size and limited number of samples, we restricted the intraindividual variability analysis to the 29 site 1 participants with repeated measures. We used linear regression to estimate the change in ΣTHM in blood per unit increase of ΣTHM in water. We adjusted the regression models for maternal age, ethnicity, education, smoking, marital status, body mass index, household income, season, study site, and reported water-use activities. We selected confounders based on percent change (> 10%) in regression coefficients from the univariate models. We performed trend analyses using one-way ANOVA to evaluate blood THM concentrations across the ingestion, noningestion, and total exposure metric tertiles. We defined statistical significance as a p
-value < 0.05 for the regression models, Spearman rank correlations, and ANOVAs.