The PELAGIE cohort. The PELAGIE cohort included 3,421 pregnant women from Brittany from 2002 through February 2006. Enrollment started in the district of Ille-et-Vilaine, continued in Finistère, and ended in Côtes d’Armor. Gynecologists, obstetricians, and ultrasonographers recruited women during consultations in early pregnancy, informing them about the nature of the study and asking them to participate (after providing written consent). Women were enrolled before the 19th week of gestation after providing written informed consent and completing a questionnaire at home concerning family social and demographic characteristics, diet, and lifestyle. Women were asked to return the questionnaire by mail, along with a first-morning-void urine sample that they collected and transferred into two vials containing nitric acid (to inhibit bacterial proliferation). Samples were mailed to the study laboratory (INSERM U625, Rennes, France) in a prestamped package at ambient temperature, with routine delivery taking from 1 to 3 days. On receipt, samples were frozen and stored at –20°C. Midwives and pediatricians at the maternity units provided study staff with medical information about the pregnancy, delivery, and neonatal health for 3,399 women (99.4% of the cohort; the other 22 women were lost to follow-up). Birth weight, length, and head circumference were obtained from hospital records by pediatricians. Gestational age was based on both date of the last menstrual period and ultrasound examinations. The appropriate national ethics committees approved the study.
Live-born singleton children of the cohort without major congenital anomalies were classified as fetal growth restricted (FGR) if their birth weight was below the 5th percentile of the distribution of expected birth weight modeled within the cohort according to gestational age (third-degree polynomial), sex, parity (third-degree polynomial), maternal weight (third-degree polynomial), height, and age (second-degree polynomial) (Mamelle et al. 2001
). Small head circumference (SHC) was defined by a head circumference below the 5th percentile of the birth head circumference distribution for a given gestational age and sex, according to French reference curves (Mamelle et al. 1996
). Major congenital malformations in live-born infants were diagnosed by hospital staff pediatricians at the maternity unit, based on findings at the clinical examination at birth. In addition, pediatricians were asked specifically about male genital anomalies (hypospadias, undescended testis, and micropenis). Diagnosis validation by pediatric surgeons was obtained for half of these male genital anomalies (n
= 12 of 26) up to 2 years after birth. For detailed information on the classification of congenital malformations in this cohort, see Garlantézec et al. (2009)
Data on corn agricultural activities and atrazine water contamination. Data on agricultural activities were collected for each mother’s municipality of residence at enrollment from the national agricultural census conducted in 2000; specifically, we recorded the proportion of the municipality’s area used for corn crops. Drinking water contamination by atrazine was routinely assessed until its year of ban by the office of Social and Sanitary Affairs of Brittany (Rennes, France). We collected these data from 2000 to 2002 for municipalities in Ille-et-Vilaine and used them to estimate the average atrazine level in water during the first trimester of pregnancy for participants residing in these locations. Exposure to atrazine via tap water was then estimated by multiplying the average concentrations by the tap water consumption reported in the questionnaire.
The present study was based on a case-cohort design (Kass and Gold 2004
) that included 601 children randomly selected from the entire live-born singleton cohort (18%), referred to as the subcohort, and groups with adverse birth outcomes, including congenital anomalies (88 major anomalies, 26 male genital anomalies), FGR (n
= 180), and SHC (n
= 105), representing all of the cohort members with these outcomes.
Pesticide determination in urine samples.
Pesticides were measured in urine samples from 579 women in the subcohort (96%), all mothers of babies with major congenital anomalies or male genital anomalies, 178 mothers of children with FGR, and 103 mothers with children with SHC. First-morning-void urine samples, which are very likely to contain recent traces of exposure to the nonpersistent pesticides we targeted (Barr et al. 2005
), were used to determine the presence of 12 triazine compounds: atrazine and its glutathione-derived metabolite atrazine mercapturate, simazine and its metabolite simazine mercapturate, three dealkylated triazine metabolites that are likely to be formed in water systems (desethylatrazine, desisopropyl atrazine, and 2-chlorodiaminoatrazine), and five hydroxylated triazine metabolites, mostly formed in plants [hydroxyatrazine, hydroxysimazine, hydroxydesethylatrazine, hydroxy-desisopropyl atrazine, and hydroxy-2-chlorodiaminoatrazine (ammeline)]. The chloroacetanilide herbicides alachlor, metolachlor, and acetolachlor were also measured, together with 2,6-diethylaniline, an alachlor metabolite.
Chemical analyses were performed by the Idhesa Institute (Plouzané, France) on urine samples (maximum, 10 mL) with liquid chromatography/triple-quadrupole mass spectrometry (LC/MS-MS) after solid-phase extraction (SPE). Triazine and chloroacetanilide herbicides were analyzed simultaneously. Reference standards for pesticides and their metabolites were purchased from the Ehrenstörfer laboratory (Augsburg, Germany) and from Promochem (Teddington, UK). Other purchased chemicals included LC/MS-grade acetonitrile and methanol from Fisher Scientific (> 99%; Loughborough, Leicestershire, UK), analytical-grade formic acid from Baker (98%; Deventer, Netherlands), and gas from Air Liquide (> 99%; Paris, France). Standard solutions were prepared in methanol or acetonitrile and stored at –20°C. Simazine-d10 was used as internal standard for extraction and detection controls. Calibration standards were prepared by adding appropriate working standard solutions to 10-mL fresh samples of pesticide-free human urine before extraction to obtain concentrations in the calibration range. Isotope-labeled standards and surrogates were added to calibration standards to obtain a final concentration of 1 μg/L. The urine samples were thawed and shaken. Preconcentration and extraction (5 mL) were performed with an online SPE high-volume symbiosis system (Spark Holland, Emmen, Netherlands)and the Hysphere C18 HD cartridge (2 × 10 mm; Spark Holland), and analytes were eluted during the mobile phase. Separation was performed with Synergy fusion-RP analytical column (250 mm × 2.0 mm, 4 μm particle diameter; Phenomenex, Torrance, CA, USA) preceded by a guard column (4 mm × 2 mm) of the same packing material from Phenomenex. The mobile phase was a gradient of a mixture of 5 mm ammonium formate/formic acid 0.01% and acetonitrile/formic acid 0.01%. The flow rate was 0.2 mL/min, and the temperature 35°C. LC/MS-MS analyses were performed with a system composed of a Waters alliance 2690 LC pump equipped with an autosampler and connected in series with a Quattro Ultima triple-quadrupole mass spectrometer (Waters-Micromass, Manchester, UK), equipped for electrospray ionization. Acquisition took place in multiple-reaction monitoring mode, with two transitions per compound (one for quantification and one for confirmation) in positive ionization mode. All validation procedures were performed with fresh samples of herbicide-free human urine. The limit of detection (LOD) was defined as a signal three times the background noise at the lowest analyte concentration assayed, and the limit of quantification (LOQ) as 10 times the background noise. LOQs ranged from 0.001 to 1.7 μg/L. Concentration ranges were linear from 0.010 to 10 μg/L for herbicides with the lowest LODs (metolachlor, desethylatrazine, alachlor, 2,6-diethylaniline) and were linear from the LOD for the others. Average recoveries were 100% ± 20%, with coefficients of variation ranging from 0.1% to 13.9%. Stability tests showed, among doped human urinary samples, a slight decrease in levels after 32 hr at ambient temperatures (respectively, –7% and –9% for atrazine and atrazine mercapturate) and no changes according to the presence of nitric acid. This method is reported to provide minimal sample handling, good extraction recovery, and satisfactory LODs and LOQs (Panuwet et al. 2008
Definition of exposure profiles from urinary concentrations.
Most human biomonitoring studies assess atrazine exposure by measuring atrazine mercapturate in urine samples. This metabolite is formed directly from atrazine and cannot result from other atrazine degradation processes in the environment or organisms (see Barr et al. 2007
, their ). However, measurement of this metabolite alone misrepresents total atrazine exposure (Barr et al. 2007
). In our study, mothers were considered exposed to atrazine when atrazine or at least one of its specific metabolites (i.e., formed only from atrazine degradation processes: atrazine mercapturate, desethylatrazine, hydroxyatrazine, or hydroxydesethylatrazine) was quantified in their urine. Among them, mothers with quantifiable levels of atrazine or atrazine mercapturate were classified as “directly exposed” to atrazine. We classified mothers as unexposed if there was no trace (i.e., < LOQ) of atrazine or any of its four specific metabolites in their urine sample. We used a similar strategy to classify exposure to simazine and its specific metabolites. We also classified women as exposed to dealkylated or hydroxylated triazine metabolites if at least one metabolite in the respective group was quantified in the urine sample, and as exposed to alachlor if alachlor or its metabolite 2,6-diethylaniline was quantified. Molar concentrations (nanomoles per liter) were summed to compute combined biomarker levels for groups of related compounds. For this computation, nonquantified values were not imputed (or were set at zero), and molar concentrations were computed by dividing the chemical determination value, expressed in micrograms per liter, by the molar mass.
Figure 1 Possible determinants of triazine and chloroacetanilide herbicide
exposure shown by urinary biomarkers in 579 pregnant women randomly selected from
the PELAGIE cohort (France, 2002-2006). Dealkylated triazine metabolites are
desethylatrazine, (more ...) Statistical analyses.
Logistic models were used to estimate associations between biomarkers and adverse birth outcomes. We did not incorporate case-cohort sampling probabilities because the case-cohort odds ratio (OR) approximates the case–control OR when the outcome is rare (Kass and Gold 2004
). Covariates considered in multivariate analyses were known or suspected risk factors and factors that predicted outcomes or urinary concentrations in this cohort. An initial list of covariates, used for the FGR models, included year of enrollment (2002–2003, 2004, 2005–2006), residence district (Ille-et-Vilaine, Côtes d’Armor, Finistère, or other), maternal age (< 25, 25–30, 30–35, > 35 years), education level (primary/secondary school, baccalaureate, technical school/postgraduate), smoking status at enrollment (nonsmoker, nonsmoker who smoked at conception, smoker), alcohol consumption at enrollment (abstinent, occasional, ≥ 1 drink per day), high blood pressure before/during pregnancy, season at conception, thawing and refreezing process (a subset of 52 samples), and urine creatinine level (milligrams per liter). For SHC models, we also considered cesarean delivery (because passage through the birth canal may influence head circumference at birth), parity (1, 2, ≥ 3), and prepregnancy body mass index (BMI; < 18.5, 18.5–25, 25–30, ≥ 30 kg/m2
). For congenital anomalies models, in addition to the initial list, we considered folic acid intake during the 3 months before enrollment (through vitamin supplementation), maternal fever in early pregnancy (before enrollment), child’s sex, parity, prepregnancy BMI, mother’s occupational exposure to solvents (no, yes, no job) [in accordance with previous associations observed from this cohort (Garlantézec et al. 2009
)], and gestational age at birth (in weeks). Linear models were used to estimate associations between biomarkers and birth weight, birth length, and head circumference as continuous outcomes, using the same initial list of covariates plus child’s sex, parity, prepregnancy BMI, gestational age at birth, and squared gestational age, as well as fish intake (three categories) and shellfish intake (three categories) according to previous results from this cohort (Guldner et al. 2007
) and, for head circumference, cesarean delivery. For all outcomes, the list of covariates also included exposures to herbicides, as categorical variables as defined above.
Final models were obtained using a backward selection process of the covariates (p < 0.20) and forcing in models the exposure of interest. Additional analyses considered biomarker levels as continuous in models and were thus restricted to samples with quantifiable levels of the compound of interest. SAS software was used for all calculations (SAS/STAT version 9.1; SAS Institute Inc., Cary, NC, USA).