Abbreviations: IQR = interquartile range; BMI = body mass index
1The “Other” race/ethnicity category is a condensation of larger categories with insufficient sample size for disaggregation (e.g., Hispanic/Latino ethnicity, South Asian, East Asian, Native American, multiracial).
Demographic characteristics for participants in this study are shown in Table 1. Overall, mothers with SGA, AGA, and LGA births were similar, as expected due to matching. Women in this study had a median age of approximately 33 years and a pre-pregnancy BMI in the normal/overweight range. They were predominantly white, with a college education and access to private health insurance. In addition, parity was similar between cases and controls. Few women reported using cigarettes or consuming alcohol during pregnancy. The distribution of fetal sex was also similar between cases and controls. Births classified as SGA had a median birth weight of 2.3 kg, while the median birth weights for AGA and LGA were 3.2 kg and 4.2 kg, respectively. Women selected into this study are similar to those in the parent LIFECODES cohort (Additional File 1: Supplemental Table 2).
Of the 33 analytes measured, 20 were detected in at least 50% of the samples at each study visit: two OPE metabolites, 11 phthalate metabolites, and seven phenolic compounds (Additional File 1: Supplemental Table 1). Within study visits, most participants had exposure biomarker measurements available at all three study visits, although some only provided two samples (n = 5) and one individual provided a single sample. Overall, samples were more likely to be missing at the last study visit (n = 89 at visit 1, n = 89 at visit 2, and n = 85 at visit 3) and were more likely to be missing in SGA (n = 4) cases compared to LGA (n = 2) or controls (n = 0). There were several differences in the average exposure biomarker concentrations between cases and controls in this study (Table 2). Urinary concentrations of mono-ethyl phthalate (MEP) and methyl paraben (MPB) were highest in mothers who went on to have AGA births and lowest in those who had LGA births. Similar trends were also noted for diphenyl phosphate (DPhP) and propyl paraben (PPB). Median concentrations and IQR of the uncorrected average exposure biomarkers and detection frequencies according to case status are reported in Additional File 1: Supplemental Table 3.
Table 2. Median (25th, 75th percentile) SG-adjusted average urinary concentrations of exposure biomarkers (ng/mL) in the study population and according to case status.
|
Chemical Class
|
Urinary Analyte (ng/mL)
|
|
Median (25th, 75th)
|
Overall
|
SGA
|
AGA
|
LGA
|
p1
|
(n = 90)
|
(n = 31)
|
(n = 31)
|
(n = 28)
|
OPEs
|
BDCPP
|
0.67 (0.40, 1.09)
|
0.84 (0.33, 1.35)
|
0.79 (0.44, 1.12)
|
0.51 (0.35, 0.78)
|
0.14
|
DPhP
|
0.74 (0.52, 1.22)
|
0.72 (0.47, 1.60)
|
0.82 (0.70, 1.34)
|
0.60 (0.45, 0.89)
|
0.07
|
Phthalates
|
MEP
|
42.9 (14.5, 128)
|
41.1 (8.48, 188)
|
65.1 (28.7, 133)
|
26.4 (11.7, 60.0)
|
0.03
|
MBP
|
9.51 (6.81, 12.9)
|
9.99 (6.27, 17.9)
|
10.5 (7.14, 16.8)
|
8.36 (5.92, 10.5)
|
0.16
|
MBzP
|
3.27 (1.86, 6.85)
|
3.57 (1.89, 7.55)
|
2.68 (1.55, 5.88)
|
3.39 (1.88, 6.78)
|
0.78
|
MiBP
|
5.69 (4.04, 8.84)
|
5.73 (4.26, 9.24)
|
6.18 (3.78, 9.95)
|
5.47 (4.03, 7.08)
|
0.30
|
MECPP
|
9.17 (5.81, 15.9)
|
11.1 (5.87, 18.5)
|
9.37 (5.81, 17.7)
|
8.52 (4.76, 15.5)
|
0.56
|
MEHHP
|
6.31 (4.45, 10.9)
|
7.60 (5.26, 10.9)
|
5.3 (4.29, 11.15)
|
5.98 (3.92, 10.4)
|
0.24
|
MEOHP
|
4.44 (3.09, 7.88)
|
5.32 (3.26, 7.88)
|
3.77 (3.21, 8.02)
|
4.18 (2.58, 7.22)
|
0.58
|
MEHP
|
1.89 (1.16, 2.90)
|
2.28 (1.34, 3.09)
|
1.76 (1.02, 3.26)
|
1.68 (1.08, 2.59)
|
0.38
|
MCPP
|
2.24 (1.28, 4.50)
|
2.13 (1.13, 5.97)
|
2.47 (1.64, 5.23)
|
2.04 (1.08, 3.43)
|
0.42
|
MCOP
|
2.02 (1.09, 5.66)
|
3.45 (1.09, 5.25)
|
2.03 (1.17, 6.79)
|
1.73 (0.71, 4.82)
|
0.42
|
MCNP
|
0.30 (0.21, 0.46)
|
0.31 (0.23, 0.50)
|
0.31 (0.23, 0.48)
|
0.27 (0.17, 0.40)
|
0.36
|
Phenols
|
2,4-DCP
|
0.31 (0.21, 0.56)
|
0.31 (0.21, 0.60)
|
0.28 (0.21, 0.51)
|
0.37 (0.23, 0.57)
|
0.60
|
2.5-DCP
|
0.62 (0.42, 1.36)
|
0.79 (0.42, 1.45)
|
0.59 (0.43, 1.47)
|
0.57 (0.32, 1.18)
|
0.54
|
BP3
|
34.7 (12.3, 95.0)
|
45.6 (9.20, 95.0)
|
26.7 (13.7, 89.4)
|
44.0 (13.2, 128)
|
0.81
|
BPA
|
0.63 (0.48, 0.92)
|
0.59 (0.48, 0.99)
|
0.62 (0.42, 0.90)
|
0.74 (0.50, 1.01)
|
0.54
|
MPB
|
105 (37.3, 188)
|
125 (56.8, 204)
|
125 (76.4, 219)
|
37.0 (24.9, 170)
|
0.01
|
PPB
|
19.0 (7.07, 40.0)
|
18.6 (7.07, 50.7)
|
23.8 (12.1, 52.5)
|
10.7 (4.89, 29.6)
|
0.05
|
TCS
|
6.30 (2.09, 33.3)
|
4.24 (1.44, 44.1)
|
5.57 (1.61, 15.7)
|
12.3 (3.32, 52.4)
|
0.13
|
1p-value: Kruskal-Wallis Test.
|
Exposure biomarker concentrations differed by several demographic characteristics (Additional File 1: Supplemental Table 4). For example, urinary concentrations of several phthalate metabolites and all measured phenols varied across maternal race categories, where most analytes were highest among Black women or women of other race/ethnicity, and lowest among white women. Levels of OPE metabolites did not vary across many demographic characteristics, although bis(1,3-dichloro-2-propyl) phosphate (BDCPP) was higher among participants reporting the use of public compared to private health insurance and was marginally higher among individuals with overweight or obese BMI compared to those with a normal/underweight BMI.
Overall, the exposure biomarkers were low-to-moderately correlated with one another (Figure 1). The strongest positive correlations were observed among phthalate metabolites, particularly among di-2-ethylhexyl phthalate (DEHP) metabolites (ranging from spearman’s rho [r] = 0.71 – 0.97). In addition, MPB and PPB were highly correlated (r = 0.84), likely reflecting similar usage in consumer products. In line with previous studies using repeated measures of these biomarkers, there was poor-to-fair stability for all analytes measured (0.25 ≤ ICC ≤ 0.74; Additional File 1: Supplemental Table 5) (10, 11, 54).
Table 3. Adjusted1 OR (95% CI) of SGA and LGA associated with an IQR-increase in average urinary exposure biomarker concentrations.
|
Chemical Class
|
Urinary Analyte
|
SGA
|
|
LGA
|
aOR1 (95% CI)
|
|
aOR1 (95% CI)
|
OPEs
|
BDCPP
|
1.12 (0.57, 2.18)
|
|
0.56 (0.27, 1.16)
|
DPhP
|
1.01 (0.55, 1.87)
|
|
0.40 (0.18, 0.87)*
|
Phthalates
|
MEP
|
0.56 (0.27, 1.19)
|
|
0.33 (0.14, 0.78)*
|
MBP
|
1.26 (0.78, 2.06)
|
|
0.71 (0.41, 1.25)
|
MBzP
|
1.37 (0.67, 2.79)
|
|
1.00 (0.46, 2.17)
|
MiBP
|
1.14 (0.60, 2.19)
|
|
0.54 (0.25, 1.16)
|
MECPP
|
1.08 (0.49, 2.36)
|
|
0.64 (0.28, 1.47)
|
MEHHP
|
1.41 (0.70, 2.86)
|
|
0.83 (0.39, 1.77)
|
MEOHP
|
1.28 (0.59, 2.78)
|
|
0.81 (0.36, 1.83)
|
MEHP
|
1.18 (0.60, 2.31)
|
|
0.74 (0.36, 1.52)
|
MCPP
|
0.83 (0.42, 1.64)
|
|
0.56 (0.27, 1.18)
|
MCOP
|
0.87 (0.38, 1.99)
|
|
0.57 (0.24, 1.36)
|
MCNP
|
0.89 (0.47, 1.68)
|
|
0.52 (0.25, 1.06)
|
Phenols
|
2,4-DCP
|
1.17 (0.54, 2.53)
|
|
1.27 (0.58, 2.81)
|
2,5-DCP
|
0.98 (0.50, 1.89)
|
|
0.70 (0.34, 1.42)
|
BP3
|
0.75 (0.36, 1.54)
|
|
0.79 (0.38, 1.65)
|
BPA
|
0.97 (0.55, 1.72)
|
|
1.42 (0.78, 2.60)
|
MPB
|
0.91 (0.42, 1.95)
|
|
0.25 (0.10, 0.63)*
|
PPB
|
0.80 (0.38, 1.71)
|
|
0.34 (0.14, 0.78)*
|
TCS
|
1.09 (0.44, 2.69)
|
|
1.95 (0.80, 4.79)
|
Asterisks indicate p < 0.05.
Abbreviations: IQR = interquartile range; aOR = adjusted odds ratio
1Models adjusted for age (years), pre-pregnancy BMI (kg/m2), maternal race (white/Black/other), maternal education (high school or less/some college or technical school/completed college or greater), and fetal sex (female/male).
|
After adjusting for age, pre-pregnancy BMI, maternal race, maternal education, and fetal sex, we observed inverse associations between several biomarkers and LGA. Specifically, an IQR-increase in average urinary concentrations of DPhP (OR: 0.40 [95% CI: 0.18, 0.87]), MEP (OR: 0.33 [95% CI: 0.14, 0.78]), MPB (OR: 0.25 [95% CI: 0.10, 0.63]), and PPB (OR: 0.34 [95% CI: 0.14, 0.78]) were associated with reduced odds of LGA (Table 3). Crude results were similar (Additional File 1: Supplemental Table 6). Associations between exposure biomarkers and SGA did not meaningfully deviate from the null.
Effect estimates were similar between strata of male and female infants, with the exception of several DEHP metabolites (i.e., mono-(2-ethyl-5-carboxypentyl) phthalate [MECPP], mono-(2-ethyl-5-hydroxyhexyl) phthalate [MEHHP], mono-(2-5-oxohexyl) phthalate [MEOHP], and mono-2-ethylhexyl phthalate [MEHP]), which exhibited suggestive sex-specific trends with respect to SGA (Additional File 1: Supplemental Table 7). Effect estimates, though imprecise, were above the null among female infants and below the null for male infants for associations between all DEHP metabolites and SGA. For example, an IQR-increase in MEHHP was associated with an OR of 3.10 (95% CI: 1.04, 9.24) among female infants and an OR of 0.63 (95% CI: 0.22, 1.80) among male infants (Wald p-value = 0.04 for interaction). There was little evidence for heterogeneity by sex for other exposure biomarkers with respect to SGA births. Similarly, few differences were observed by sex for LGA births. However, we note that stratum-specific estimates tended to be highly imprecise due to the small study sample.
Lastly, we conducted a sensitivity analysis to examine potential confounding by cigarette and alcohol use during pregnancy. Effect estimates were not meaningfully different from primary results when individuals reporting the use of cigarettes and alcohol were excluded from the analysis (Additional File 1: Supplemental Table 8).
Using quantile g-computation, we estimated the joint associations between mixtures of individual chemical classes (i.e. OPEs, phthalates, and phenols) and the odds of an SGA or LGA birth. The results from these multi-pollutant models were consistent with observations made from our single-pollutant models (Figure 2; Additional File 1: Supplemental Table 8). Namely, we observed that simultaneously increasing both OPE metabolites by one quartile was associated with reduced odds of LGA (OR: 0.49 [95% CI: 0.27, 0.89]). Weights from the quantile g-computation model indicate that DPhP contributed more strongly than BDCPP to this association (Additional File 1: Supplemental Table 9). Increasing all phthalate metabolites by one quartile was also associated with reduced odds of LGA (OR: 0.23 [95% CI: 0.07, 0.73]). Within this model, MEP was assigned the largest negative weight (i.e., the greatest contribution to the negative partial effect). Finally, a one quartile increase in all phenols was associated with lower odds of LGA (OR: 0.68 [95% CI: 0.40, 1.94]), although the association did not reach statistical significance. As in the single-pollutant approach, the associations between exposure biomarker mixtures and SGA did not differ from the null. |