Demographics of the study population are shown in Table 2. The majority of mothers were under the age of 30 (67.1%), had at least some college education (79%), were employed (63%), had an annual household income under $30,000 (63.1%), were married (53.1%), had never smoked (86%) or been exposed to environmental tobacco smoke (88.7%), did not drink alcohol during pregnancy (93.6%), had given birth to less than 2 previous children (86.9%), and had a pre-pregnancy BMI of less than 25 (56.1%).
Table 2
Maternal demographic characteristics of the study population (N = 976)
| N (%) |
Maternal Age (years) | |
18–24 | 354 (36.3%) |
25–29 | 301 (30.8%) |
30–34 | 206 (21.1%) |
35–41 | 115 (11.8%) |
Maternal Education | |
GED or less | 203 (21%) |
Some College | 331 (34.2%) |
Bachelors or Higher | 433 (44.8%) |
Employment Status | |
No | 357 (37%) |
Yes | 608 (63%) |
Annual Household Income | |
< 10k | 269 (31.6%) |
10k-<30k | 268 (31.5%) |
30k-<50k | 203 (23.8%) |
>=50k | 112 (13.1%) |
Marital Status | |
Single | 197 (20.4%) |
Married | 521 (53.9%) |
Cohabitating | 249 (25.7%) |
Smoking Status | |
Never | 833 (86%) |
Ever | 121 (12.5%) |
Current | 15 (1.55%) |
Daily Environmental Tobacco Smoke Exposure | |
Never | 808 (88.7%) |
1 Hour or less | 40 (4.39%) |
> 1 Hour | 63 (6.92%) |
Alcohol Use | |
Never | 504 (52.2%) |
Yes, before Pregnancy | 400 (41.4%) |
Yes, currently | 62 (6.42%) |
Number of Previous Children | |
0 | 355 (42.7%) |
1 | 367 (44.2%) |
2 to 5 | 109 (13.1%) |
Pre-Pregnancy BMI | |
[0,25] | 520 (56.1%) |
(25, 30] | 240 (25.9%) |
Above 30 | 167 (18%) |
Fetal Sex | |
Female | 464 (48%) |
Male | 502 (52%) |
Distributions of hormone concentrations are shown in Supplementary Table S1. Most hormone concentrations were significantly different at 18 and 26 weeks’ gestation, with notable increases occurring with estriol (median 15.1 and 38.2 ng/mL at 18 and 26 weeks, respectively) and progesterone (median 39.3 and 73.5 ng/mL at 18 and 26 weeks, respectively). ICCs for all other hormones ranged from 0.647 (T4) to 0.856 (testosterone).
Distributions of birth outcomes are shown in Table 3. PTB and spontaneous PTB occurred in 9.9% and 5.8% of the study population, respectively. Preeclampsia and GDM were less prevalent (2.9% and 1.9%, respectively). Occurrences of SGA and LGA births were similar (8.9% and 9.6%, respectively). Median gestational age of the study population was 39.1 weeks (IQR: 38.1–40).
Table 3
Distributions of continuous and binary birth outcomes
| Min | 10th | 25th | 50th | 75th | 90th | Max |
Gestational Age (wks) | 20.3 | 36.7 | 38.1 | 39.1 | 40 | 40.7 | 42.7 |
Birthweight Z-Score (ounces) | -5.34 (19.0) | -1.19 (91.0) | -0.571 (102) | -0.00005 (113) | 0.707 (123) | 1.25 (133) | 9.70 (224) |
| N (%) | | | | | | |
Preterm Birth | | | | | | | |
No | 867 (90.1%) | | | | | | |
Yes | 95 (9.88%) | | | | | | |
Spontaneous Preterm Birth | | | | | | | |
No | 883 (94.2%) | | | | | | |
Yes | 54 (5.76%) | | | | | | |
Preeclampsia | | | | | | | |
No | 947 (97.1%) | | | | | | |
Yes | 28 (2.87%) | | | | | | |
Gestational Diabetes | | | | | | | |
No | 900 (98.1%) | | | | | | |
Yes | 17 (1.85%) | | | | | | |
Small for Gestational Age | | | | | | | |
No | 842 (91.1%) | | | | | | |
Yes | 82 (8.87%) | | | | | | |
Large for Gestational Age | | | | | | | |
No | 835 (90.4%) | | | | | | |
Yes | 89 (9.63%) | | | | | | |
Figure 1 shows the associations between hormone concentrations and birth outcomes at each study visit (all effect estimates and p-values are shown in Supplementary Table S2). There were greater odds of spontaneous PTB with increasing progesterone concentrations at 26 weeks (OR: 2.12, 95% CI: 1.29, 3.47) and fT4 concentrations at both study visits (18wk OR: 1.60, 95% CI: 1.07, 2.39; 26wk OR: 1.73, 95% CI: 1.04, 2.86). The risk of spontaneous PTB was significantly different between study visits with an IQR increase in Prog/E3 (interaction p = 0.026), a null association observed at 18 weeks and increased odds observed at 26 weeks (OR: 1.63, 95% CI: 1.05, 2.54). Reductions in gestational age at birth were observed with increased concentrations of progesterone (β: -3.56 days, 95% CI: -6.02, -1.10), fT4 (β: -2.22 days, 95% CI: -3.84, -0.61), and T4 (β: -1.87 days, 95% CI: -3.62, -0.11) around 18 weeks, and with prog/e3 at both study visits (18wk β: -1.77 days, 95% CI: -3.36, -0.19; 26wk β: -1.98 days, 95% CI: -3.58, -0.37). Notably, the effect of progesterone was significantly different between study visits (interaction p = 0.044).
Results at 18 weeks suggested that elevated progesterone and reduced estriol are associated with increased risk of having an SGA infant (E3 OR: 0.66, 95% CI: 0.45, 0.97; progesterone OR: 1.53, 95% CI: 1.09, 2.17; prog/E3 OR: 1.77, 95% CI: 1.29, 2.44). This trend remained at 26 weeks for only prog/E3 (OR: 1.53, 95% CI: 1.07, 2.17). Similarly, prog/E3 at 18 weeks was inversely associated with birthweight z-score (β: -0.12, 95% CI: -0.23, -0.02) and estriol at 26 weeks was positively associated with birthweight z-score (β: 0.21, 95% CI: 0.01, 0.41).
A protective effect against preeclampsia was observed with increases in SHBG at 18 weeks (OR: 0.55, 95% CI: 0.30, 0.99) and estriol (OR: 0.42, 95% CI: 0.17, 0.99) and SHBG (OR: 0.46, 95% CI: 0.25, 0.83) at 26 weeks. Conversely, elevated risk of preeclampsia was observed with an increase in TSH at 26 weeks (OR: 2.18, 95% CI: 1.19, 3.99). The odds of GDM increased with an IQR increase in TSH (OR: 1.67, 95% CI: 1.02, 2.72), T3 (OR: 2.83, 95% CI: 1.04, 7.68), and T3/T4 (OR: 2.97, 95% CI: 1.20, 7.35) at 18 weeks, and increased with higher estriol at 18 weeks (OR: 5.95, 95% CI: 1.27, 27.8). None of the associations with preeclampsia or GDM were significantly different between study visits.
Sensitivity analyses revealed that many associations were significantly different between male and female pregnancies (Fig. 2; all effect estimates and p-values are shown in Supplementary Table S3). The most compelling effect modification by fetal sex was observed for preterm birth; the interaction term between hormone concentration and fetal sex indicator was significant among 7 out of 11 hormones and hormone ratios assessed. SHBG was protective against PTB at 26 weeks among female (OR: 0.60, 95% CI: 0.37, 0.96), but not male, pregnancies (interaction p = 0.032). Higher testosterone at both study visits was associated with increased odds of PTB among female pregnancies and reduced odds of PTB among male pregnancies (interaction p < 0.001). Notably, increased odds of PTB were observed among only male pregnancies with elevated concentrations of CRH (OR: 1.82, 95% CI: 1.09, 3.05; interaction p = 0.002), estriol (OR: 1.81, 95% CI: 1.07, 3.06; interaction p = 0.022), progesterone (OR: 1.88, 95% CI: 1.16, 3.04; interaction p = 0.011), and fT4 (OR: 1.63, 95% CI: 1.06, 2.51; interaction p = 0.115) at 18 weeks. Assessment of gestational age as a continuous variable did not provide such compelling results, but it did provide additional evidence of fetal sex modifying the association with progesterone at 18 weeks (male pregnancy β: -4.9 days, 95% CI: -2.73, -7.07 days; interaction p = 0.015).
The spontaneous subtype of PTB also showed several cases of effect modification by fetal sex. An IQR increase in CRH at 18 weeks was associated with greater odds of spontaneous PTB among only male pregnancies (OR: 2.73, 95% CI: 1.38, 5.43; interaction p = 0.003). Increases in testosterone at both visits were protective against spontaneous PTB among only male pregnancies (interaction p = 0.001). Increases in T3 and fT4 at both study visits were associated with increased odds of spontaneous PTB among only male pregnancies, but effect modification was significant only for T3 (interaction p = 0.013). Finally, higher progesterone at 26 weeks was associated with increased off of spontaneous PTB among only male pregnancies (OR: 2.34, 95% CI: 1.36, 4.03).
Fetal sex modified the association between SGA and only the ratio prog/E3 (interaction p = 0.022), which was positive among only male pregnancies at both 18 weeks (OR: 2.39, 95% CI: 1.59, 3.60) and 26 weeks (OR: 1.98, 95% CI: 1.29, 3.05). Accordingly, increased estriol resulted in increases in birthweight z-score at both 18 weeks (β: 0.19, 95% CI: 0.02, 0.36) and 26 weeks (β: 0.31, 95% CI: 0.08, 0.53) among only male pregnancies (interaction p = 0.030). Fetal sex did not modify any associations between hormones and odds of LGA.
Though there was no evidence of effect modification by fetal sex on associations between hormones and preeclampsia, significant effects were observed only among female pregnancies with increases in SHBG (OR: 0.34, 95% CI: 0.14, 0.81), TSH (OR: 2.41, 95% CI: 1.11, 5.23), and fT4 (OR: 0.40, 95% CI: 0.17, 0.92) at 26 weeks. Conversely, there was significant evidence of effect modification by fetal sex on the association between various hormones and odds of GDM. Elevated thyroid hormones were observed to be protective against GDM among female pregnancies [(fT4 at 18wks OR: 0.29, 95% CI: 0.10, 0.85; interaction p = 0.001), (T4 at 18wks OR: 0.32, 95% CI: 0.11, 0.90; interaction p = 0.002)], but positively associated with GDM among male pregnancies [(T3 at 18wks OR: 6.04, 95% CI: 1.72, 21.3; interaction p = 0.028), (fT4 at 26wks OR: 4.87, 95% CI: 1.53, 15.5), (T4 at 26wks OR: 3.05, 95% CI: 1.02, 9.13)]. A similar trend was observed for the ratio of prog/E3; there was a protective effect at 18 weeks among female pregnancies (OR: 0.25, 95% CI: 0.09, 0.71) and a positive association at 26 weeks among male pregnancies (OR: 2.93, 95% CI: 0.99, 8.69; interaction p = 0.004).
PTB and Gestational Age
We observed greater odds of PTB and spontaneous PTB with increasing progesterone concentrations (when fetal sex was male), but other studies demonstrating similar significant associations are lacking. One study observed progesterone concentrations measured between 28 and 32 weeks’ gestation to be higher among women who delivered preterm compared to full term22. We observed higher progesterone concentrations among PTB cases when fetal sex was male, but only around 18 weeks’ gestation. We also observed higher progesterone concentrations around 26 weeks among women who spontaneously delivered preterm compared to women who carried to term.
Previous work has shown that a ratio favoring estriol in mid-pregnancy21 and at delivery35 is associated with earlier time of labor. Progesterone concentrations rise steadily during pregnancy, contributing to uterine quiescence, downregulation of prostaglandin production, and immune tolerance of the fetus36,37. At the onset of human labor, progesterone concentrations do not notably decrease; rather, the body’s response to progesterone is dampened. It is not clear exactly how this occurs, but possibilities include reduction in progesterone receptor expression, changes in receptor isoforms, and local progesterone metabolism38. As term approaches, the ratio of progesterone to estriol shifts to favor estrogens, with the functional decrease in progesterone driving initiation of labor39. The new dominance of estrogens promotes an increase in prostaglandin and oxytocin receptors and enzymes responsible for muscle contractions, which work together to help promote labor40. We observed a positive association between odds of PTB and estriol concentrations (when fetal sex was male), but we also unexpectedly observed later gestational age at birth with higher concentrations of estriol at 26 weeks’ gestation when the fetus was female. In contrast with previous studies, we observed that higher prog/E3 was associated with reduced gestational age and increased odds of SGA. Interestingly, among women who delivered preterm, a previous study observed lower prog/E3 among only those without premature rupture of membranes41, possibly implicating different endocrine pathways in the occurrence of PTB with and without premature rupture of membranes.
Decreased odds of PTB have been shown with increased concentrations of fT4 in the second23 and third42 trimesters, which contradicts our finding that fT4 was inversely associated with gestational age at birth (among the whole study population and when the fetus was male), and increased odds of PTB (when the fetus was male) and spontaneous PTB. One prior study also found increased odds of PTB with greater T3 concentrations at 10 and 26 weeks gestation42. Similarly, we found that T3 was associated with spontaneous PTB when the fetus was male. Mechanisms of the association between thyroid hormones and PTB are poorly understood, but previous research has suggested that altered thyroid hormone concentrations may be involved in other disease states or exposures for which we have evidence of associations with PTB such as oxidative stress and inflammation43–45, or environmental exposures such as phthalates46–48.
Several previous studies have observed that male fetal sex is associated with a greater risk of delivering preterm. Proposed biological explanations for this observation include a pro-inflammatory environment generated by a male fetus27 and larger size at birth for males relative to females28. Increased risk of PTB when the fetus was male among only Caucasian women has also been observed, suggesting a potential interaction between race and fetal sex29. We observed significant associations with PTB unique to women carrying a male fetus for CRH, estriol, progesterone, and fT4, providing further evidence that the effect of fetal sex on the occurrence of PTB is complex, possibly involving diverse endocrine pathways.
Preeclampsia
Among all pregnancies, we observed reduced odds of preeclampsia with an increase in estriol at 26 weeks. In accordance with our findings, another study showed that estriol concentrations in the second trimester49 were lower among women with preeclampsia than women with normal pregnancies. Previous studies have also found increased odds of preeclampsia with higher second trimester fT4 concentrations23,50, and lower third trimester fT4 concentrations51. All associations we observed between fT4 and preeclampsia were inverse, and the inverse association at 26 weeks among female pregnancies was significant. The association between fT4 and preeclampsia has been shown to be modified by human chorionic gonadotropin (hCG) concentrations, with high fT4 positively associated with preeclampsia only when hCG is low52. This effect modification may be due to the known angiogenic role of hCG during early pregnancy53.
Hormonal involvement in the etiology of preeclampsia is complex due to the angiogenic dysfunction of the affected uterus. In preeclampsia cases, proper remodeling and infiltration of blood vessels by placental extravillous trophoblasts does not occur, and this can be observed before the onset of clinical symptoms54,55. It is unclear whether endocrine disruption plays a causal role in initiation of uterine dysfunction, or if uterine dysfunction triggers a maternal endocrine response in an attempt to adapt to the hypoxic state56.
Gestational Diabetes Mellitus
A previous epidemiology study has demonstrated associations between high second trimester estriol concentrations and greater odds of GDM25. We also observed increased odds of GDM with estriol at 26 weeks. Testosterone concentrations were inversely associated with odds of GDM among male pregnancies in our study, which differs from previous research that showed higher testosterone concentrations among women with GDM26 and with greater insulin resistance57 compared to women with normal pregnancies.
Previous work has suggested that fT4 concentrations early in pregnancy are inversely associated with odds of GDM51,58. In accordance with those findings, the ratio of fT3 to fT4 has been observed to be positively associated with odds of GDM59, suggesting that increased conversion of T4 to biologically active T3 may play a role in the onset of GDM. In alignment with those findings, we observed greater odds of GDM among all pregnancies with increased T3 concentrations at 18 weeks, and greater odds of GDM among male pregnancies with increased T3 at both study visits. We also observed an inverse association between fT4 and odds of GDM at 18 weeks among female pregnancies, while that association was positive among male pregnancies at 26 weeks. Previous work has shown that women with GDM have higher circulating concentrations of inflammatory cytokines such as IL-6 and TNF-alpha60, which have been observed to be inversely associated with T3 concentrations61. These inflammatory markers may increase insulin resistance during pregnancy and, mediated by alterations in thyroid hormone concentrations, contribute to higher circulating glucose levels and increased odds of GDM. Several previous studies have observed greater risks for GDM among women carrying a male fetus62–64, possibly due to poorer beta-cell function among male fetuses65.
Birth Size
We observed that decreased birthweight among females was marginally associated with elevated T4 at 26 weeks. Previous work found similar inverse associations, but with fT4 instead of total T466,67. Thyroid hormones are critical for fetal growth, possibly via their influences on fetal insulin-like growth factor, leptin, or the placenta’s abilities to transfer nutrients68. Even in the case of nearly identical patterns of thyroid hormone concentrations throughout gestation between mothers, differences in expression of hormone transporters in the placenta and intracellular receptors in fetal tissues can result in different thyroid hormone exposure profiles for the fetus and, consequently, varying effects on fetal growth and development69. Assessment of thyroid hormone effects on birth outcomes in the second half of gestation is even more complex as the fetal thyroid gland begins to produce hormones and the fetus relies less on maternal supply of T469. Conflicting results on the relationship between thyroid hormones and birthweight between studies may be due in part to unmeasured differences in fetal thyroid function.
Strengths and Limitations
The present study was subject to several limitations. We were not able to measure hCG or assess thyroid autoantibody status. Thus some of our results could be biased due to unmeasured confounding variables. Some critical changes in the maternal endocrine environment occur later in gestation than we were able to measure, such as the exponential increase in CRH right before the onset of labor. Although the goal of this study was to determine whether mid-pregnancy hormone levels were indicative of increased risk of adverse pregnancy outcomes, measurements at later time points could shed additional light on the various endocrine pathways implicated in adverse birth outcomes. We observed low rates of preeclampsia and GDM, which reduces the reliability of effect estimates. However, these lower rates were observed because we excluded women with preexisting conditions from our cohort to allow more precise examination of associations between hormone concentrations and birth outcomes, since preexisting conditions can influence hormone concentrations and susceptibility to adverse birth outcomes. Furthermore, excluding women with preexisting conditions may limit the generalizability of our findings. Finally, some results assessing preeclampsia and GDM may be subject to reverse causation bias if the disease state, before clinical observation, resulted in the hormonal changes that we observed.
Despite the aforementioned limitations, this study was also strong in various ways. This is one of few studies to assess a broad panel of hormone concentrations at more than one time point during gestation to investigate relationships with various birth outcomes and different windows of susceptibility. Many epidemiological studies limit their analytical panel to either thyroid or steroid hormones, or do not assess the spontaneous subtype of PTB. We are also one of few groups to assess interactions between gestational hormone concentrations and fetal sex. Finally, our study was strengthened by a higher sample size of mothers than was seen in most previously published cohorts, which is particularly important when studying rare outcomes occurring in less than 5% of the population.