Study population
This study was based on the Taiwan Birth Panel Study (TBPS), a prospective cohort study, which was conducted between April 2004 and January 2005 (Hsieh et al., 2011). Informed consents were obtained from study participants before delivery to collect umbilical cord blood at birth and stored at -80℃ until laboratory analysis.
After delivery, all subjects were interviewed by trained interviewers to obtain the information during prenatal life style. The protocols used in this study were approved by the Ethical Committee of National Taiwan University Hospital.
In this study, five mothers who reported smoking over 5 months during pregnancy, four mothers who reported drinking more than 100 mL during pregnancy, four infants who were very preterm (gestational age < 32 weeks), and six twin babies were excluded from the analysis. Combined with the appropriate cord blood for phenols measurement, finally, a total of 401 mother-infant pairs were remained in our study.
Measurement of BPA, NP, OP in cord blood
Chemicals and reagents
Bisphenol A was supplied by AccuStandard (New Haven, Connecticut, USA).
Bisphenol A-D16, 4-tert-octylphenol and the technical mixture of nonylphenol were obtained from Sigma/Aldrich (Saint Louis, MO, USA). 4-n-Octyl-d17-phenol was supplied by C/D/N Isotopes (Pointe-Claire, Quebec, Canada; purity > 98%). Stock solutions of each compound were prepared at a concentration of 500 μg/mL in methanol and stored at -20℃. Milli-Q water was obtained from a Millipore water purification system (Milford, MA,USA). N-methylmorpholine (purity > 99.5%) were provided by J.T. Baker (Phillipsburg, NJ, USA). Solvents including methanol and acetonitrile were all LC/MS grade (J.T. Baker). Bovine plasma were purchased from Sigma-Aldrich (St.
Louis, MO, USA).
Sample preparation and calibration experiments
All glassware was rinsed with methanol before being used for experiments. The concentration of BPA, NP, and OP in the cord blood were quantified using modified analytical methods previously described (Anari et al., 2002). We conducted a small
experiment to compare two methods of derivatized with Dansyl chloride and underivatized analysis. We chose the better method by comparing the linear range of calibration and the feasibility between derivatization and non derivatization (the results showed in Appendix1). Finally, the underivatization method was used in this study.
A 50 μL aliquot plasma was diluted with 0.5 mL of water and added in 50 μL of internal standard (Bisphenol A-D16 and 4-n-Octyl-d17-phenol; 200 ng/mL in methanol).
After gentle mixing, samples were added in 2 mL of ethyl acetate to each test tube. The samples were vortex-mixed and the organic phase (1.5 mL) of the samples was transferred to another tube, then filtrated through 0.22 μm PVDF syringe filters into a 2 mL auto-sampler vial, and evaporated to dryness by SpeedVac concentrator (Thermo Savant SPD 1010, Holbrook, NY, USA). The residues were reconstituted with 50 μL of methanol and transferred to 150-μL insert for UPLC/MS/MS analysis.
Matrix-matched standard calibration solution was prepared in bovine plasma and through the same procedure of sample preparation. The linear range were 0.5-500 ng/mL for BPA, 10-750 ng/mL for NP, 2.5-200 ng/mL for OP, and spiked 200 ng/mL of internal standard (BPA-d16 and 4-n-Octyl-d17-phenol ) in each solution.
Instrumental analysis
We measured the concentration of bisphenol A (BPA), nonylphenol (NP), and octylphenol (OP) in cord bloods using ultra-performance liquid chromatography tandem mass spectrometry (UPLC/MS/MS). The UPLC/MS/MS was performed using a Waters Acquity UPLC system (Waters Corporation, Milford, MA, USA) and controlled by MassLynx V4.1 with QuanLynx Application Manager. An ACQUITY UPLC BEH C18 column (2.1 mm 50 mm, 1.7μm) was used, the temperature and the flow rate were maintained at 60℃ and 0.5 mL/min, respectively. The mobile phase was composed of 10 mM N-methylmorpholine (pH 9.5) and acetonitrile. The initial composition of gradient program was 40% acetonitrile for 0.5 min, followed by a linear gradient to 60% acetonitrile in 0.5 min, then 95% acetonitrile in 2 min and held at 95% acetonitrile for 1 min before being returned to the initial condition. At last, the column was re-equilibrated at 40% acetonitrile for 1 min. The total run time of gradient program was 5 minutes and the sample injection volume was 5 μL.
To achieve maximal analyte signal intensities, the instrumental parameters were referenced to previously described (Lien et al., 2009). The mass spectrometer was performed in native electrospray ionization (ESI-) and the capillary voltage was maintained at 3.0 kV. The desolvation gas flow, con gas flow, desolvation temperature,
and source temperature were set at 900 L/hr, 50 L/hr, 400℃ and 120℃, respectively.
Extractor voltage was 3.0 V and RF lens voltage was 0 V. Collision gas was argon at 3.13×10-3 mbar. Ion energy 1 and 2 were set at 0.3 and 3, respectively. Both LM 1 and LM 2 resolution were set at 12. The multiplier voltage was set at 650 V. The dwell time for NP and OP were 0.08 second, then for BPA was 0.1 second. Ions were monitored by selected reaction monitoring (SRM) as shown in Table 1 (p55) as well as the individual collision energy and cone voltage.
Evaluation of matrix effect and extraction efficiency of sample pretreatment
Three duplicates of bovine plasma which were spiked three different levels of BPA, NP and OP (10, 25 and 100 ng/mL) before or after the procedure of sample preparation. For matrix effect, the same levels of all analytes were spiked into solvent solutions. The percent matrix effect was calculated by the following formula: matrix effect% = (area of post-extraction spike / area of standard) × 100. Then, the extraction efficiency was tested to measure the analytes loss during sample preparation. We used the following formula to calculate the extraction efficiency: absolute recoveries% =
Method validation and quantification
The sample preparation of plasma with UPLC/MS/MS method was validated regarding the precision, accuracy, and detection limit. For intra- and inter-day precision and accuracy, the calibration standards in bovine plasma were analyzed on the same day (n=3) and on three different days. The recovery of the method was determined by three duplicates of bovine plasma spiked known amounts (10, 25 and 100 ng/mL) of BPA, NP and OP with fixed levels of internal standard, and was calculated by dividing the measured quantities with the theoretical (spiked) quantities. BPA and NP were found in blanks, almost all the human plasma were detected the peak but some signal intensities were below the background level, it is impractical to calculate their limits of detection (LODs), therefore we used the lowest concentration of calibration curve to define their LOQs. In addition, limits of detection (LODs) and quantification (LOQs) were defined as the minimum concentration of OP in the calibration curve by using the peak-to-peak option of MassLynx software and was defined as signal-to-noise ratios (S/N ratios) equaling to 3 and 10.
For quantification accuracy, the quality control (QC) samples were prepared from a plasma pool obtained from multiple cord blood plasma which provide from those were not recruit in this study. The plasma pool was divided into four subpools, one
subpool was used to analyze the phenols levels of the samples (no spike), and the three different concentration (5, 20 and 100 ng/mL) were spiked into the other three subpools.
All glassware was rinsed with methanol before being use for experiments. A solvent blank sample spiked fixed level of internal standard with each batch of samples to check experimental contamination and background level of native analytes. The quality control (QC) samples were prepared from human plasma, and the matrices were mixed uniformly and divided into three subpools. Two subpools were used to spiked with fixed level of internal standard to detect the level of analytes, then another subpool was spiked with known level (50 ng/mL) of BPA, NP and OP standard to check the stability of method. These QC samples were chose from every twenty-five samples.
The linear ranges in plasma with 1/ x weighted were as follow: 0.5-500 ng/mL for BPA; 10-750 ng/mL for NP; 2.5-200 ng/mL for OP. The square of the correlation coefficient (R2) was equal to or greater than 0.997. The data acquired and processed using MassLynx V4.1 Software.
Infant’s birth outcomes and child growth
Birth outcomes used in this analysis included infant’s gestational age (weeks), birth weight (grams), length (centimeters) and head circumference (centimeters), which were obtained by researchers abstracting from medical records. Growth data were recorded from Child Healthcare Handbook which was created and published by Bureau of Health Promotion, Department of Health, Taiwan. The height, weight and head circumference of the child were measured at visits to clinics for health examinations and were recorded in the Child Healthcare Handbook. We also measured child’s height and weight in every times of follow-up at different age.
Statistical analysis
The backgrounds of BPA and NP were deducted, then phenols concentrations were natural log-transformed to fit the normal distribution before linear/mixed regression analysis. We used t-test and ANOVA to assess the phenols concentration distribution in different characteristics of mothers and infants. The univariable linear/mixed models analysis were used to define the potential confounders of birth outcomes and child growth parameters. Furthermore, any potential confounders which were significantly related to at least one of the growth parameters or phenols levels in
cord blood were included in the multivariable models.
Linear regression models were used to assess the association between prenatal phenols exposure and birth outcomes of newborns. According to the statistics and literature review, the potential confounders were infant gender, gestational age, lead and cotinine level in cord blood, maternal education, maternal BMI during pregnancy, and annual household income.
Mixed models were used to assess the effect of prenatal phenols exposure and child growth from birth to 6 years of age. All pairs, no matter how many growth measurements from birth to 6 years they had, were included in the analysis. We also analyzed age- and sex-specific height and weight z-scores according to the growth norm of preschool children in Taipei City (Lee et al., 2009). Because of the growth measurements time of every subjects are not consistent, we adjusted the measurements time in the mixed models. Other potential confounders in models included infant gender, gestational age, lead and cotinine level in cord blood, and maternal education.
Furthermore, the phenols levels in cord blood were classified by quartile to assess the dose-response of prenatal phenols and child growth.
Statistical analysis was conducted with SAS version 9.2. All tests were two-sided
Results
Method performance, matrix effect, recovery and method validation
The matrix effects in different spiking levels in bovine plasma of the analytes were shown in Table 3. For NP, the matrix effect was from 81.5% to 97.9%, and ion enhancement effect was found of BPA and OP. The extraction efficiencies of the three phenols in bovine plasma were 64.4% to 123.3% (Table 3, p57). Using the quantitative method of matrix-matched calibration with one internal standard, the recoveries of three analytes were 94.3%-107.0% as shown in Table 3. The accuracy and precision of intra- and inter-day of matrix-matched calibration were shown in Table 4 (p58).
Three different concentrations (5, 20, 100 ng/mL) were spiked into the human cord blood samples to evaluate the method accuracy and precision. We found that the measured concentrations were very close to the spiked levels if the backgrounds (no spike) were not deducted (Table 5, p59).
Because BPA and NP were found in the blanks, their LOQs were defined as the lowest concentration of calibration curve, 0.5 and 10 ng/mL, respectively. Moreover, the LOD and LOQ of OP were 0.89 and 2.96 ng/mL, which was defined as the signal-to-noise (S/N) ratios equaling to 3 and 10.
Phenols levels in cord blood
A total of 401 subjects’ cord bloods were used in this study and the detection rate of BPA, NP and OP were 55.86%, 77.56% and 68.33%, respectively. If BPA or NP levels were below the background level, half of LOQ (0.25 ng/mL for BPA; 5 ng/mL for NP) would be regarded as the BPA or NP levels in cord blood. Also, half of LOD (0.45 ng/mL) would be regarded as the OP levels if OP levels were lower than the LOD. The range of BPA was from 0.25 to 211.40 ng/mL, NP was from 5.0 to 561.57 ng/mL, and OP was from 0.45 to 183.4 ng/mL. Then, the median of BPA, NP and OP were 1.50, 60.97, and 2.30 ng/mL, respectively (Table 6, p60).
The phenols concentration distribution in different characteristics of mothers and infants were shown in Table 7 (p61). For NP levels, young mothers had lowest geometric mean concentration than other older mother; mothers who had low annual household income had higher NP concentration. BPA concentration of the mothers who had ETS-exposure during pregnancy was higher than those who didn’t have prenatal ETS-exposure. But, we didn’t find the correlation between cotinine level in cord blood and prenatal phenols exposure in this study.
Prenatal phenols exposure and child growth
A total of 401 mother-infants pairs were recruited in this study (Figure 1, p51).
The characteristics of these 401 mother-infants pairs were shown in Table 8 (p62). Half of mothers had lower education level, and about 30% of mothers had ETS-exposure during pregnancy. The geometric means of lead level and cotinine level in cord blood were 1.07 and 0.15 ng/ml, respectively. We excluded the vary preterm infants, so the mean of gestational age and birth weight are all within normal limits.
The relationship between infant and maternal factors, and growth parameters were shown in Table 9 (p63). Infant’s gender, gestational age, cotinine level in cord blood, maternal age at delivery, and maternal education were important determinants of birth size at birth and growth from birth to 6 months. Because maternal age was highly correlated with maternal education, we only adjusted maternal education in multivariable models to avoid over-adjustment.
Table 10 shows that the linear regression models of natural log-transformed phenols concentration and birth outcomes. After adjusting, there was a negative association between ln-NP in cord blood and head circumference at birth. Head circumference decreased by 0.15 cm (95% CI: -0.25, -0.05) in association with a 1-unit increase in natural log-transformed NP concentration (Table 10, p64).
Table 11 to 14 shows that the effects of prenatal phenols exposure on child growth between birth to different age. After adjusting the growth measurements time and other confounders, there was a significantly negative association between cord blood NP and head circumference up to 18 months (Table 11 and Table 12, p65 and 66). Regarding other growth parameters, an increase in cord blood NP was significantly associated with a decrease in height z-score from birth up to 6 years when controlling the confounders (Table 12 to Table 14, p66-68). A 1-unit increase of ln-NP was significantly associated with a decrease in head circumference (β = -0.12, 95% CI:
-0.20, -0.04 for 0-6 months; β = -0.11, 95% CI: -0.21, -0.02 for 0-18 months), and height z-score (β = -0.06, 95% CI: -0.12, -0.01 for 0-6 years). Although the estimates suggested that no significant dose-response between prenatal NP exposure and child growth, children who in the high dose group had smaller head circumference and height than those children who in the low dose group ( Table 15, p69).
Discussion
Our findings indicated that prenatal exposure to nonylphenol was adversely related to the birth outcome and child growth at early childhood, especially for head circumference and height. However, no effects were found for exposure to bisphenol A and octylphenol.
Compared with previous studies, the BPA, NP and OP levels in cord blood in our study were slightly higher than other countries and the general population in Northern Taiwan (Chou et al., 2011; Chen et al., 2008; Lee et al., 2008; Schönfelder et al., 2002).
In the US population, concentrations of BPA for those in low annual household income were higher than those in high household income (Calafat et al., 2008). Although we didn’t find the significant inverse trend with annual household income, the mothers whose annual household income was below NT$ 1,000,000 had high BPA level in this study. Furthermore, we found the significant inverse trend for NP concentration with mother’s education level and annual household income.
To our knowledge, this is the first study to investigate the association between NP levels in cord blood and child growth. Our study found that head circumference and height were decreased with the increase in the prenatal NP exposure. Similar results were found in animal studies, Hirano et al found an adverse effect of NP exposure on
body length among aquatic invertebrate species, and the results indicated that the body length to be the most sensitive to NP exposure (Hirano et al., 2009). Another study found that birth length was decreased with the increase in the gestational exposure of NP concentration (Jie et al., 2010). After categorizing NP exposure in quartile, we observed that children’s head circumference and height in higher exposure group were smaller and shorter than low exposure group. These findings suggested that higher prenatal NP exposure may lead to the adverse effect on child growth. Although the half-life of NP from the blood was 2-3 hour (Müller et al., 1998), the exposure rout of NP is ubiquitous in the environment and repeated exposure of fetus due to increased amount taken by pregnant mothers is possibly lead to the adverse effect on children’s growth and development.
We didn’t find the prenatal BPA exposure affect birth sizes and child growth, also the inconsistent findings were shown in previous studies either in rodent or in human.
Perinatal low dose BPA exposure to rate was associated with increase in body weight after birth and continued into adulthood (Rubin et al., 2001), another study found the effect on perinatal BPA exposure with decrease in body weight in postnatal days (Honma et al., 2002). Moreover, previous human studies had shown that prenatal BPA
Regarding the BPA exposure and obesity, the association between BPA exposure and obesity in adults and elderly adults had been shown in previous studies (Wang et al., 2012; Carwile and Michels, 2011). So far the association was found in adults rather than the young children. Furthermore, the half-life of BPA from blood was less than 6 hours and rapidly excreted with urine (Völkel et al., 2002).
Several potential limitations in our study should be discussed. First, the most important limitation in our study, a single plot blood collection has the potential to misclassify exposure. The concentration of BPA in urine is variable over time during pregnancy, and many environmental factor are related with BPA exposure (Braun et al., 2011). In this study, we assumed the phenols exposure sources of theses pregnant mothers are regular and sustained during pregnancy, although the half-life of phenols is short. In addition, all the cord blood was collected at delivery. We standardized the timing of cord blood collection to reduce the variability over time. Second, we didn’t measure the postnatal exposure to phenols of these children. Phenols exposure routes of young children are ubiquitous in the environment such as dietary ingestion, non-dietary ingestion, and inhalation route. Postnatal exposure is also important for exposure assessment of children and may lead to the adverse effect on child growth.
Additionally, measurement errors of birth outcomes may exist in this study, especially for birth length and head circumference. The measurement of head circumference is
likely to have errors due to head modeling. The measurement errors were expected to be random, and we excluded the irrational measurements of growth data for each subjects.
However, this is a prospective and longitudinal birth cohort study, and we had repeated measurements of child growth at different age to follow up the child’s growth and development. Also, we had lead and cotinine levels in cord blood and adjusted these potential confounders in the models to control the effect of lead and cotinine exposure on child growth. Braun et al (2011) found that the urinary BPA concentrations were positively associated with serum cotinine levels among the pregnant mothers. Exposure to BPA from tobacco smoking because BPA comprises 25% of the weight of some cigarette filters (Jackson and Darnell, 1985). Therefore, Braun et al (2011) suggested that further studies should examine the joint effect of BPA and tobacco smoke exposure and adjust for on another.
Conclusion
According to the results of this study, we observed the consistent effect of
According to the results of this study, we observed the consistent effect of