Subclinical Changes in Maternal Thyroid Function Parameters in Pregnancy and Fetal Growth

Abstract

Context

Overt thyroid disease in pregnancy is a known risk factor for abnormal fetal growth and development. Data on the effects of milder forms of variation in maternal thyroid function on intrauterine growth are less well examined.

Objective

We explored these associations using repeated thyroid hormone and ultrasound measurements.

Design, Setting, and Participants

Data were obtained from 439 pregnant women without diagnosed thyroid disease who were participants in a case-control study of preterm birth nested within an ongoing prospective birth cohort in Boston, Massachusetts.

Main Outcome Measures

Ultrasound and delivery indices of fetal growth were standardized to those measured in a larger population.

Results

At median 10, 18, and 26 weeks of gestation, we observed significant inverse associations between free thyroxine (FT4) and birth weight z scores, with the strongest association detected at median 10 weeks, at which time a 10% increase in FT4 was associated with a 0.02 z score decrease (∼8.5 g) in birth weight (β = −0.41 for ln-transformed FT4; 95% confidence interval, −0.64 to −0.18). FT4 was also inversely associated with repeated measurements of estimated fetal weight, head circumference, and abdominal circumference. We observed weaker inverse associations for total T4 and a positive relationship between total triiodothyronine and birth weight z scores. We did not observe any associations for thyroid-stimulating hormone.

Conclusion

In pregnant women without overt thyroid disease, subclinical changes in thyroid function parameters may influence fetal growth.

Associations between subclinical changes in repeated measures of maternal thyroid function parameters and ultrasound and delivery indices of fetal growth were explored.

Impaired fetal growth is a major predictor of neonatal mortality and morbidity and may increase the risk of long-term health complications, such as diabetes and cardiovascular disease, in adulthood (1, 2). Thyroid hormones play a crucial physiological role in early placental development and in intrauterine growth and fetal tissue accretion and differentiation (3, 4). Overt thyroid dysfunction in pregnancy (hyperthyroidism and hypothyroidism) has been consistently linked to abnormal fetal growth and development, including low birth weight (LBW), preterm birth, decreased head circumference (HC), and neurodevelopmental delays (5–8).

Investigations of the effects of milder forms of thyroid dysfunction on fetal growth are less conclusive. Subclinical hypothyroidism [high thyroid-stimulating hormone (TSH) with normal free thyroxine (FT4)] and hypothyroxinemia (normal TSH with low FT4) have been associated with a smaller HC and birth length (8), intrauterine growth restriction (IUGR) (9), LBW (9, 10), and/or an increased risk for small for gestational age (SGA) neonates (8, 11). However, similar studies on these relationships have reported conflicting results (12–14) or null associations (15–18). The more consistent findings have been in generally euthyroid pregnant women in whom increases in FT4 have been associated with lower birth weight (14, 19–21) and an increased risk for SGA newborns (20).

Few longitudinal studies have been performed with repeated measures of maternal thyroid hormones in pregnant women (22–25). To our knowledge, only one study has used serial measurements of thyroid function parameters to explore potential associations with fetal growth (26). Specifically, Nishioka et al. (26) reported that an increase in maternal TSH concentrations between the first and third trimesters was associated with LBW (<2500 g). This study’s relatively small sample size, and thus its limited number of LBW babies (n = 10), precluded trimester-specific analyses and likely contributed to the lack of variation in free hormones between women who delivered LBW babies and controls.

Given the heterogeneous results in the available literature and the lack of longitudinally collected data across pregnancy, we investigated the extent to which thyroid function parameters, collected at up to four times points during pregnancy, were associated with birth weight and repeated measurements of fetal growth in 439 pregnant women without clinical thyroid disease.

Materials and Methods

Study population

The current study is a secondary analysis of data from a nested case-control study of preterm birth drawn from the LIFECODES cohort, which is a prospective birth cohort of pregnant women in Boston, MA. Women ≥18 years old were recruited early in pregnancy (<15 weeks of gestation) between 2006 and 2008 and were eligible for participation if they were carrying a singleton, nonanomalous fetus and planned to deliver at Brigham and Women’s Hospital. Details regarding recruitment and eligibility criteria are described in detail elsewhere (27). At the initial study visit (median = 10 weeks of gestation), women completed a medical questionnaire that collected sociodemographic and health-related information (e.g., personal and family health history). Participants were followed until delivery and were provided relevant health information and blood samples at three additional study visits: visit 2 (median = 18 weeks of gestation), visit 3 (median = 26 weeks of gestation), and visit 4 (median = 35 weeks of gestation). Gestational age at individual study visits and at delivery were calculated based on last menstrual period and confirmed by first-trimester ultrasound or were based entirely on first-trimester ultrasound. Birth weight was recorded at delivery.

From the prospective birth cohort, we selected 130 women who delivered preterm neonates (<37 weeks) and 352 randomly selected controls for inclusion in the nested case-control study. For the current analysis, we additionally excluded 41 women with self-reported pre-existing or gestational thyroid disease/conditions (e.g., hyperthyroidism or hypothyroidism, Graves disease, or thyroid cancer) based on answers to medical questionnaires administered at each of the study visits. Two women who did not provide plasma samples at any study visit were also excluded. Our final study population included 116 cases of preterm birth and 323 controls. The study protocols were approved by the ethics review board at Brigham and Women’s Hospital (Partners Health Research Committee), and all study participants gave written informed consent.

Ultrasound measurements

The American College of Obstetricians and Gynecologists’ guidelines for perinatal care recommend that all women undergo evaluations for aneuploidy in the first trimester and again in the second trimester to examine fetal anatomy (28). Thus, in our study population all women had ultrasound scans that provided crown rump length at visit 1 and fetal morphology at visit 2. Ultrasound scans at visits 3 and 4 were not routine but were obtained frequently for a variety of indications, including maternal gestational diabetes or suspected restricted fetal growth (29). However, thyroid hormone concentrations did not statistically significantly differ between women without ultrasound scans at visits 3 and 4 (n = 146) compared with women with scans during at least one of these visits (n = 293) (Supplemental Table 1). Thus, the potential for differential bias resulting from the inclusion of ultrasound metrics obtained after visit 2 (i.e., those obtained for clinical indications and not uniformly measured for all subjects) is limited. All ultrasound measurements were conducted by experienced faculty sonologists with active Society of Maternal-Fetal Medicine certification.

In the current analysis, we used the following ultrasound measurements from visits 2 through 4: HC, abdominal circumference (AC), and femur length (FL). We calculated estimated fetal weight (EFW) at each of the study visits using the Hadlock formula, which combines biparietal diameter, AC, and FL (30). To combine and compare fetal growth measurements across different time points, we standardized raw ultrasound measurements, EFW, and birth weight using z scores based on the gestational age–specific means and standard deviations of the corresponding measurements available for 18,904 nonanomalous singleton pregnancies delivered between 2006 and 2012 at Brigham and Women’s Hospital (29, 31).

Thyroid hormone measurements

A total of 1443 plasma samples from 439 pregnant women were assayed for thyroid function parameters at the University of Michigan Clinical Ligand Assay Service Satellite Laboratory (Ann Arbor, MI). These samples were collected at up to four study visits. We measured TSH as well as total triiodothyronine (T3) and thyroxine (T4) using an automated chemiluminescence immunoassay (Bayer ADVIA Centaur; Siemens Health Care Diagnostics, Inc., Tarrytown, NY). FT4 was measured using direct equilibrium dialysis followed by radioimmunoassay (IVD Technologies, Santa Ana, CA).

The manufacturer only provided a nonpregnant normal range of 0.35 to 5.50 μIU/mL for TSH. In the absence of trimester-specific reference ranges, the American Thyroid Association recommended in 2011 the following TSH reference ranges: first trimester, 0.1 to 2.5 μIU/mL; second trimester, 0.2 to 3.0 μIU/mL; and third trimester, 0.3 to 3.0 μIU/mL (32). However, the American Thyroid Association recently recommended an upper limit of ∼4.0 μIU/mL when TSH pregnancy references are not available (33). The laboratory pregnancy reference ranges for FT4 were: first trimester, 0.7 to 2.0 ng/dL; second trimester, 0.5 to 1.6 ng/dL; and third trimester, 0.5 to 1.6 ng/dL. The limits of detection were 0.01 μIU/mL for TSH, 0.1 ng/mL for T3, 0.3 μg/dL for T4, and 0.1 ng/dL for FT4. The interassay coefficients of variation for all hormones ranged from 2.3% (for total T3) to 10.4% (for FT4), and the intra-assay coefficients of variation ranged from 1.2% (for total T3) to 12.3% (for FT4) (25).

Statistical analyses

All analyses were performed using R version 3.3.2. Fetal growth indicators included in our study are secondary outcomes in the original nested case-control study of preterm birth. Thus, to correct for potential over-representation of preterm birth cases and to make the current study more representative of the base LIFECODES cohort, we applied to all analyses inverse probability weights representing the inverse sampling fraction for selection of cases (90.1%) and controls (33.9%) from the base population (34, 35).

To assess the bivariate associations between population characteristics and birth weight z scores, we used linear regression models adjusted for gestational age at delivery. We examined the distributions of raw ultrasound measurements and thyroid hormone parameters by calculating selected percentiles at each study visit of sample collection as well as at delivery (for birth weight). The empirical distributions of total T3 and T4 approximated normality. The distributions of TSH and FT4 were positively skewed and were natural log transformed (ln) for all statistical analyses. We calculated percentiles and tested the differences in mean thyroid hormone concentrations between women who underwent ultrasound scans at visit 3 and/or visit 4 vs those who did not undergo scans at either visit using linear mixed models (LMMs) with subject-specific random intercepts.

In cross-sectional analysis, we explored the relationships between continuous thyroid hormone concentrations measured at each of the four study visits and birth weight z scores using stratified (by study visit) multivariable linear regression models. We included maternal age, race/ethnicity, body mass index (BMI) at initial study visit, and fetal sex a priori. Additional covariates, such as health insurance provider, educational attainment, parity, and smoking and alcohol use in pregnancy, were retained in models if their inclusion resulted in ≥10% change in the main β estimates using a manual forward stepwise selection procedure. Crude models were adjusted for gestational age at time of sample collection. Final models were adjusted for maternal age (continuous), race/ethnicity (white/black/other regardless of Hispanic origin), BMI at initial study visit (continuous), fetal sex (male/female), health insurance provider (private/public), and parity (no previous pregnancy/one previous pregnancy/more than one previous pregnancy). Given previously published data showing the potential for fetal sex to modify the associations between maternal thyroid function and fetal growth (14), we investigated differences in these relationships by fetal sex by including an interaction term in each cross-sectional model.

Our second analysis explored the relationships between repeated measures of thyroid hormones and fetal growth z scores using LMMs with one growth indicator regressed on one thyroid hormone per model. Given that ultrasound measurements at visit 2 tended to be more homogeneous regarding EFW and to remain consistent with previous analyses of fetal growth indicators within this cohort (29), we presented results from repeated measures models using thyroid hormone and ultrasound measurements from visits 3 and 4. We also reported results from LMMs that included visit 2 through delivery for comparison. For models with fetal weight as the outcome, we created vectors for each subject with repeated measures of EFW at visits 2 (or 3) through 4 and birth weight at delivery. Because maternal blood samples were not collected at delivery, we imputed these concentrations by using the last observation carried forward method. That is, hormone concentrations from visit 4 were imputed for missing values at delivery. Final models included a subject-specific random intercept and slope for gestational age, chosen based on the Akaike information criterion, and were adjusted for the same covariates as those in our cross-sectional analysis. All LMMs were repeated with the addition of an interaction term for fetal sex to investigate potential sex differences in these relationships.

Because thyroid hormones directly influence fetal tissue metabolism (4) and are regulated to remain within a narrow range (36), we hypothesized that the degree of variation in thyroid hormones within individual pregnant women may influence fetal growth. To explore this hypothesis, we calculated the ratios of intraindividual variation to interindividual variation (= within-person variance/between-person variance) for samples collected at visits 1 through 4. We regressed natural log-transformed variability ratios for each hormone on repeated measures of fetal growth z scores using fully adjusted LMMs with the same random effects and covariates as those used in the main repeated measures analysis. Results from models using repeated measurements from visits 2 through delivery as well as visits 3 for delivery were presented. All associations were considered statistically significant at the 5% level.

Results

Our study participants were predominantly white (56%) and of a healthy BMI (<25 kg/m²; 53%) and held private health insurance (80%); many were college educated (38%) (Table 1). In our bivariate analyses presented in Supplemental Table 2, we observed statistically significantly lower birth weight z scores among babies born to black women compared with white women and to women who reported smoking in pregnancy vs those who reported no smoking. Statistically significantly greater birth weight z scores were detected among babies born to overweight (25 to 30 kg/m²) and obese (>30 kg/m²) women compared with women with a healthy BMI, among babies born to women with at least one previous pregnancy compared with women with no previous pregnancies, and in female vs male babies.

Table 1.

Population Characteristics of 439 Pregnant Women

Population Characteristicsa	n (%)b
Age, y
18–24	54 (13)
25–29	92 (21)
30–34	176 (40)
35+	117 (27)
Race/ethnicity
White	247 (56)
Black	75 (17)
Other	117 (27)
Education level
High school	67 (15)
Technical school	76 (17)
Junior college or some college	127 (30)
College graduate	159 (38)
Health insurance provider
Private	344 (80)
Public	83 (20)
BMI at initial visit, kg/m2
<25	223 (53)
25–30	113 (26)
>30	99 (21)
In vitro fertilization
No	414 (95)
Yes	25 (6)
Fetal sex
Male	198 (46)
Female	241 (54)
Parity
No previous pregnancies	197 (45)
One previous pregnancy	144 (34)
More than one previous pregnancy	98 (21)
Tobacco use
No smoking in pregnancy	402 (93)
Smoked in pregnancy	31 (7)
Alcohol use
No alcohol use during pregnancy	412 (95)
Alcohol use during pregnancy	18 (5)

^aMissing observations: n = 10 for education level, n = 12 for insurance provider, n = 4 for BMI, n = 9 for alcohol use, n = 6 for tobacco use, and n = 3 for subclinical hypothyroidism.

^bProportions weighted by preterm birth case-control sampling probabilities to represent the general sampling population.

The distributions of raw fetal growth indicators (not z scored) and thyroid hormone concentrations by study visit of sample collection are shown in Table 2. The weighted median (25th to 75th percentile) gestational age at delivery was 38.9 weeks (37.9 to 40.0 weeks) and for birth weight was 3345 g (2945 to 3660 g). Thyroid function parameters were highly detected in this study population (percent detected for total T4 and T3 = 100, TSH = 99.5, and FT4 = 98). The weighted median (25th to 75th percentile) concentrations of the four parameters in the overall study population were: TSH, 1.26 µIU/mL (0.85 to 1.79 µIU/mL); FT4, 1.08 ng/dL (0.85 to 1.32 ng/dL); T4, 10.2 µg/dL (9.1 to 11.5 µg/dL); and T3, 1.56 ng/mL (1.31 to 1.89 ng/mL). More women had ultrasound scans at visit 2 (n = 389 for HC and AC; n = 390 for FL; n = 324 for EFW) compared with visit 3 (n = 201 for all growth indicators) and visit 4 (n = 221 for HC; n = 223 for AC, FL, and EFW). All women had birth weight measurements (n = 439).

Table 2.

Weighted Distributions of Thyroid Hormone and Raw Ultrasound Measurements by Study Visit of Sample Collection

	n	Selected Percentiles
		25th	50th	75th	90th	95th	Max
Visit 1 (median 10 wk)
Thyroid hormonesa
TSH, µIU/mL	317	0.54	0.91	1.50	1.97	2.36	7.23
FT4, ng/dL	374	1.16	1.39	1.64	2.07	2.57	12.5
T4, µg/dL	364	8.70	10.0	11.3	12.6	13.7	18.0
T3, ng/mL	304	1.11	1.31	1.61	1.86	2.11	2.91
Visit 2 (median 18 wk)
Ultrasound measurements
HC, mm	389	141	148	159	166	173	202
AC, mm	389	119	127	137	148	154	177
FL, mm	390	25	27	29	31	32	40
EFW, g	324	221	253	290	334	361	528
Thyroid hormonesa
TSH, µIU/mL	317	0.98	1.36	1.91	2.52	3.15	6.08
FT4, ng/dL	368	0.92	1.13	1.30	1.54	1.77	8.96
T4, µg/dL	357	9.50	10.6	11.8	12.7	13.6	16.4
T3, ng/mL	301	1.36	1.60	1.90	2.16	2.28	2.86
Visit 3 (median 26 wk)
Ultrasound measurements
HC, mm	201	231	248	263	280	288	307
AC, mm	201	207	221	239	255	263	280
FL, mm	201	46	50	54	57	60	68
EFW, g	201	810	978	1200	1442	1619	2032
Thyroid hormonesa
TSH, µIU/mL	306	0.90	1.29	1.70	2.31	2.86	8.73
FT4, ng/dL	349	0.81	1.00	1.18	1.44	1.97	12.5
T4, µg/dL	333	9.20	10.3	11.4	12.7	13.5	21.7
T3, ng/mL	286	1.36	1.62	1.90	2.15	2.37	2.79
Visit 4 (median 35 wk)
Ultrasound measurements
HC, mm	221	308	317	326	333	340	366
AC, mm	223	302	317	333	349	361	393
FL, mm	223	65	68	71	74	75	77
EFW, g	223	2323	2671	3034	3478	3643	4384
Thyroid hormonesa
TSH, µIU/mL	270	0.97	1.35	1.92	2.44	2.71	6.22
FT4, ng/dL	344	0.77	0.96	1.17	1.43	1.58	6.19
T4, µg/dL	337	8.90	10.0	11.4	12.7	13.6	20.2
T3, ng/mL	239	1.40	1.66	2.00	2.21	2.4	3.22
Birth (median 38 wk)
Birth weight, g	439	2946	3345	3660	3909	4139	4720

Analyses weighted by preterm birth case-control sampling probabilities.

^aNumber of samples per hormone varied due to limitations in sample volume.

Table 3 presents results from our fully adjusted cross-sectional analysis in which we explored the associations between thyroid hormone parameters sampled at each study visit and birth weight z scores. In measurements collected at visits 1 through 3, we observed significant inverse associations between FT4 and birth weight, with the greatest association detected at visit 1. At this visit, a 10% increase in FT4 was associated with almost a 0.02 z score decrease in birth weight [β = −0.41 for ln-transformed FT4; 95% confidence interval (CI), −0.64 to −0.18], or ∼8.5 g based on the birth weight standard deviation at 40 weeks estimated for the reference population (31). The inverse associations for total T4 were weaker than those observed for FT4, although a significant association was observed at visit 3 (β = −0.08; 95% CI, −0.13 to −0.02). In contrast to the inverse associations observed for FT4 and total T4, total T3 was generally positively associated with birth weight z scores. We detected a significant association for T3 at visit 3, at which point a unit increase in T3 was associated with a 0.35 z score (95% CI, 0.01–0.69) or an ∼149 g increase in birth weight. No significant associations were detected for TSH at any of the four study visits of sample collection. There were no significant interactions with fetal sex (data not shown).

Table 3.

Weighted Multivariate Cross-Sectional Associations Between Thyroid Function Parameters and Birth Weight z Scores by Study Visit of Sample Collection (n = 439 Pregnant Women)

Hormones	Visit 1 (Median 10 wk)			Visit 2 (Median 18 wk)			Visit 3 (Median 26 wk)			Visit 4 (Median 35 wk)
	n	β (95% CI)	P Value	n	β (95% CI)	P Value	n	β (95% CI)	P Value	n	β (95% CI)	P Value
TSHa	303	−0.02 (−0.13 to 0.10)	0.79	308	0.02 (−0.17 to 0.20)	0.86	297	−0.14 (−0.36 to 0.09)	0.23	259	−0.16 (−0.38 to 0.05)	0.13
FT4a	359	−0.41 (−0.64 to −0.18)	<0.001	357	−0.26 (−0.46 to −0.05)	0.01	339	−0.24 (−0.44 to −0.05)	0.02	330	−0.10 (−0.27 to 0.07)	0.25
T4	349	−0.01 (−0.06 to 0.05)	0.79	346	−0.06 (−0.12 to 0.00)	0.06	324	−0.08 (−0.13 to −0.02)	0.01	323	−0.05 (−0.11 to 0.00)	0.06
T3	290	0.25 (−0.09 to 0.59)	0.15	292	0.25 (−0.09 to 0.59)	0.16	277	0.35 (0.01 to 0.69)	0.04	228	0.18 (−0.13 to 0.50)	0.26

All models adjusted weighted by preterm birth case-control sampling probabilities and adjusted for gestational age at time of sample collection, maternal age, race/ethnicity, BMI at initial study visit, insurance provider, parity, and fetal sex.

^aln-transformed prior to analysis.

^bBold text indicates P < 0.05.

Table 4 shows the results from repeated measures analysis of thyroid function parameters and z scored fetal growth indicators (HC, AC, and EFW) from visits 3 and 4 and delivery (birth weight). Similar to cross-sectional associations, FT4 was inversely associated with fetal growth indicators, with significant estimates found for EFW (β = −0.14; 95% CI, −0.26 to −0.02), HC (β = −0.17; 95% CI, −0.31 to −0.02), and AC (β = −0.16; 95% CI, −0.29 to −0.02). We also observed significant inverse associations between T4 and HC (β = −0.05; 95% CI, −0.09 to 0.00) and AC (β = −0.05; 95% CI, −0.10 to 0.00). Although repeated measures associations were generally inverse for TSH and positive for total T3, the β estimates for these hormones were not significant. We found no significant associations between thyroid function parameters and FL in our study population and no significant interactions with fetal sex (data not shown). For comparison, results from our repeated measures analysis using samples and ultrasound measurements from visits 2 through 4 and delivery are shown in Table 5. The only significant association in that analysis was observed between FT4 and HC (β = −0.12; 95% CI, −0.23 to −0.01).

Table 4.

Weighted Multivariate Repeated Measures Associations Between Thyroid Function Parameters and Fetal Growth z Scores (n = 439 Pregnant Women; Measurements From Visits 3 Through Delivery)

Hormones	EFW (V3–Delivery)			HC (V3–V4)			AC (V3–V4)			FL (V3–V4)
	N a	β (95% CI)	P Value	N a	β (95% CI)	P Value	N a	β (95% CI)	P Value	N a	β (95% CI)	P Value
TSHb	555	−0.12 (−0.30 to 0.05)	0.17	294	−0.06 (−0.26 to 0.14)	0.55	296	−0.17 (−0.38 to 0.05)	0.13	296	−0.06 (−0.29 to 0.17)	0.61
FT4b	693	−0.14 (−0.26 to −0.02)	0.02	361	−0.17 (−0.31 to −0.02)	0.02	363	−0.16 (−0.29 to 0.02)	0.03	363	−0.10 (−0.26 to 0.05)	0.18
T4	671	−0.04 (−0.08 to 0.01)	0.09	346	−0.05 (−0.09 to 0.00)	0.048	348	−0.05 (−0.10 to 0.00)	0.04	348	−0.03 (−0.08 to 0.02)	0.23
T3	493	0.21 (−0.05 to 0.48)	0.11	263	0.15 (−0.13 to 0.42)	0.29	265	0.16 (−0.15 to 0.47)	0.31	265	0.06 (−0.27 to 0.39)	0.71

All analyses weighted by preterm birth sampling probabilities. LMMs include subject-specific random intercept and slope and were adjusted for gestational age at time of sample collection/delivery, maternal age, race/ethnicity, BMI at initial study visit, insurance provider, parity, and fetal sex.

^aN = total number of samples/measurements.

^bln-transformed prior to analysis.

^cBold text indicates P < 0.05.

Table 5.

Weighted multivariate Repeated Measures Associations Between Thyroid Function Parameters and Fetal Growth z Scores (Measurements From Visits 2 Through Delivery)

Hormones	EFW (V2–Delivery)			HC (V2–V4)			AC (V2–V4)			FL (V2–V4)
	N a	β (95% CI)	P Value	N a	β (95% CI)	P Value	N a	β (95% CI)	P Value	N a	β (95% CI)	P Value
TSHb	834	0.01 (−0.12 to 0.13)	0.89	573	0.00 (−0.13 to 0.13)	0.95	575	0.01 (−0.13 to 0.15)	0.92	576	0.03 (−0.10 to 0.17)	0.63
FT4b	1017	−0.07 (−0.17 to 0.02)	0.14	685	−0.12 (−0.23 to −0.01)	0.03	687	−0.07 (−0.19 to 0.05)	0.25	688	−0.05 (−0.16 to 0.07)	0.41
T4	985	−0.02 (−0.05 to 0.02)	0.35	660	−0.02 (−0.06 to 0.02)	0.32	662	−0.01 (−0.06 to 0.03)	0.47	663	−0.02 (−0.06 to 0.02)	0.41
T3	757	0.11 (−0.09 to 0.32)	0.28	527	0.15 (−0.06 to 0.37)	0.16	529	0.21 (−0.02 to 0.45)	0.08	530	−0.13 (−0.37 to 0.10)	0.26

All analyses weighted by preterm birth sampling probabilities. LMMs include subject-specific random intercept and slope and were adjusted for gestational age at time of sample collection, maternal age, race/ethnicity, BMI at initial study visit, insurance provider, parity, and fetal sex.

^aN = total number of samples/measurements.

^bln-transformed prior to analysis.

^cBold text indicates P < 0.05.

In our variability analysis, we found higher interindividual variation compared with intraindividual variation for all hormones, with total T4 displaying the greatest variability overall (Supplemental Table 3). Upon regressing the variability ratio (an index of an individual’s hormone variation) on repeated measures of fetal growth z scores, we observed that a unit increase in the variability ratio for total T4 was statistically significantly associated with reduced EFW (β = −0.29; 95% CI, −0.58 to −0.01), HC (β = −0.41; 95% CI, −0.74 to −0.08), and FL (β = −0.43; 95% CI, −0.77 to −0.08) in models using repeated measurements from visits 3 through delivery (Supplemental Table 4). Similar results were observed for models including visit 2 ultrasound measurements (Supplemental Table 5). We did not detect significant associations for any of the other thyroid hormones in this secondary analysis.

Discussion

In this prospective study of pregnant women without overt thyroid disease, we observed consistent inverse associations between FT4 and fetal growth indicators. Specifically, FT4 was inversely associated with birth weight in cross-sectional models stratified by study visit and with EFW in repeated measures models. Not surprisingly, given that HC and AC are the major constituents of EFW, the models also found these metrics to be similarly inversely associated. We observed weaker inverse associations for total T4 and a positive relationship between total T3 at median 26 weeks of gestation and birth weight. No significant associations were observed for TSH. Finally, we did not detect any different effects by fetal sex, although the lack of an interaction may be due to sample size limitations.

Maternal thyroid hormones can indirectly and/or directly influence fetal growth and development. The placenta is a thyroid hormone–responsive organ, evident in part by its high nuclear binding capacity for T3, expression of thyroid hormone transporters and binding proteins, and concentration of deiodinase enzymes that together regulate the transplacental passage of maternal thyroid hormones to the developing fetus (37–39). Prior to the formation of the fetal thyroid gland in the second trimester, the fetus relies solely on the transplacental supply of maternal thyroid hormones (38), which are integral in the maintenance and function of the human placenta in early pregnancy (3). In vitro studies have shown that maternal T3 may play a physiological role in placentation by regulating trophoblast proliferation and presumed invasion and decidual remodeling (37, 40). Inadequate trophoblast invasion leads to restricted perfusion, which would be a risk factor for IUGR among other gestational abnormalities (37, 41).

Studies investigating the effects of subclinical thyroid dysfunction in pregnancy on fetal growth have produced variable results. Subclinical hypothyroidism (elevated TSH and euthyroid FT4) in early pregnancy (<20 weeks of gestation) has been associated with smaller HC and birth length in 1107 mother-child pairs (8) and in the third trimester has been linked to higher rates of IUGR (EFW <10th percentile for gestational age) and LBW (<2500 g) in another large study (9). High maternal TSH, defined by study-specific cutoffs (i.e., exceeding reference ranges or 90th percentile), in the first trimester has also been associated with LBW (10) and SGA (birth weight <10th percentile for gestational age) infants (11). In contrast to these findings, a recent study found that subclinical hypothyroidism in early pregnancy was associated with a greater risk for large for gestational age (birth weight >90th percentile for gestational age) in male newborns (14).

Results have also been conflicting for maternal hypothyroxinemia (normal TSH with low FT4). One study reported a higher risk for SGA (8), whereas others have found larger birth weights among these babies compared with those born to euthyroid mothers (12, 13). Despite these results, studies conducted within other large birth cohorts have reported no associations between maternal subclinical hypothyroidism or hypothyroxinemia and similar growth metrics at delivery (e.g., birth length, HC, LBW, or SGA) (15–18). In the current study, although estimates were generally inverse, we did not detect any associations between TSH in pregnancy and birth weight or repeated growth measurements. These dissimilar results may be due to differences in several components of study design and/or laboratory methods, such as study sample size, laboratory assays for measurement of free hormones (e.g., chemiluminescence immunoassay vs radioimmunoassay), time of sample collection in pregnancy, regional iodine status, or different cutoff points used to define subclinical thyroid disease (e.g., different percentile and/or concentrations thresholds).

The more consistent findings from these studies have been among generally euthyroid women in whom increases in FT4 have been associated with smaller birth weight (14, 19–21). In pregnant women without overt thyroid disease, Shields et al. (21) observed that maternal FT4 at 28 weeks of gestation was inversely associated with birth weight in a dose-dependent manner. Maternal FT4 within the normal range in early pregnancy (median ∼13 weeks) was also inversely associated with birth weight in the Generation R cohort (20) and in a recent study among a community-based cohort of pregnant women living in Amsterdam (14). Additionally, Haddow et al. (19) found that babies born to euthyroid pregnant women with second-trimester FT4 concentrations in the highest quintile had the lowest birth weight. Our results from the current study support these findings. In pregnant women without diagnosed thyroid disease, we detected inverse associations between maternal FT4 and birth weight z scores at three time points in pregnancy (at median 10, 18, and 26 weeks of gestation). We also found consistent inverse associations between FT4 and estimated fetal growth as well as HC and AC in repeated measures analysis.

During the second half of gestation, thyroid hormones influence intrauterine growth by stimulating fetal metabolism (i.e., consumption of oxygen and glucose), by affecting fetal bioavailability of growth-related hormones (e.g., growth hormones and prostaglandins) and growth factors (e.g., insulinlike growth factors), and by indirectly regulating fetal tissue accretion and differentiation near term (4). The direct effects of thyroid hormones on skeletal growth and tissue differentiation have been shown in animal studies in which the lack of thyroid hormones (specifically T3) in mice resulted in skeletal abnormalities (42), and in fetal sheep thyroidectomy showed delayed bone maturation and altered bone strength and density (4, 43). Although free T3 is the more biologically active thyroid hormone (44), plasma sample volume constraints precluded its measurement in our study population. Furthermore, free T3 is generally not routinely assessed due to its low sensitivity and specificity for diagnosing hypothyroidism (45, 46). For total T3, we observed a positive association with birth weight in samples collected at 26 weeks. We are not aware of published data on the relationship between total T3 and fetal growth. However, our findings contrast with the null associations reported between unbound T3 and the risk for LBW and SGA neonates in a population-based cohort of pregnant women (10).

Our results from repeated measures analysis showing inverse associations between FT4 and estimated fetal growth as well as HC and AC suggest that maternal FT4 plays a role in fetal soft tissue accretion and possibly bone growth. The stronger associations observed in measurements collected after visit 2 (median 18 weeks of gestation) may indicate that the majority of the effects of maternal thyroid function on fetal growth occurs in the latter third of pregnancy. The underlying physiology behind these observed inverse associations between FT4 and total T4 and the fetal growth indicators is unclear. In our cross-sectional analyses, we observed that associations between FT4 and birth weight z scores were strongest in the first trimester in terms of magnitude and significance. A unit increase in ln-transformed FT4 measured at median 10 weeks of gestation was associated with a decrease in birth weight of ∼175 g. Although most fetal weight gain occurs after ∼24 weeks of gestation (31, 47), our results suggest that maternal FT4 concentrations in early pregnancy may have a lasting impact on fetal growth across gestation. These effects may be explained by altered placental function, which has an indirect effect on fetal growth by modifying nutritional transport and metabolism (48). A crude marker of placental function is placental weight, which through surface area influences the capacity of nutrient transfer to the fetus (48) and correlates closely with birth weight (49, 50). In a cohort of 321 euthyroid pregnant women, Bassols et al. (51) observed an inverse association between maternal FT4 and placental weight, suggesting a potential role of maternal thyroid hormones in regulating placental growth and, more indirectly, fetal growth. Higher maternal FT4 in early pregnancy has also been found to influence placental function later in pregnancy (3). Specifically, among pregnant women without thyroid disease, a higher maternal FT4 at median 13 weeks of gestation was associated with increased vascular resistance in the latter half of pregnancy and thus may be a potential risk factor for impaired placentation and/or vascularization (3). Maternal FT4 may also influence fetal growth via the effects of deiodinase activity in the placenta. Type 3 deiodinase inactivates T4 and T3 and reduces the concentrations of T4 and T3 in fetal circulation (39, 52). It is possible that suppressed type 3 deiodinase activity in pregnancy, perhaps due to a lack of an estrogenic response (52), leads to higher T4 and/or T3 exposure to the fetus and thus to altered fetal growth and development. Our findings for FT4 may be also explained by other factors determining thyroid hormone bioavailability in utero (e.g., transporter or receptor proteins) or additional thyroid hormone–dependent processes involved in fetal growth that are not captured by the biochemical markers of thyroid function available in our study. Further research is required to characterize the underlying mechanisms by which FT4 may influence fetal growth in additional cohorts of generally euthyroid pregnant women.

Results from our variability analysis support the findings from a previous longitudinal investigation among 132 pregnant women showing markedly higher interindividual variation compared with corresponding intraindividual variation for all hormones assayed (22). Similar to our study, Boas et al. (22) observed the greatest overall variability (both within and between) for total T4 and the largest variability ratio for FT4. Our repeated measures analysis revealed that a unit increase in within-subject variance in total T4 was associated with decreased HC, FL, and EFW z scores. These results suggest that increased variability of total T4 within the normal range—perhaps a marker of estrogen fluctuations in pregnancy that indirectly influence total T4 concentrations and fetal bone development (53, 54)—may affect fetal bone maturation. Similar associations were not observed for fetal AC, which predominantly captures liver size (55). Together, our findings from this secondary analysis have implications for trimester-specific reference intervals. That is, current reference ranges based on population data may not be sensitive markers of thyroid dysfunction at the individual level because they greatly exceed the ranges of intraindividual variation (22, 56). Although additional plasma measurements per woman may reduce exposure misclassification, our variability findings suggest that these effects may be marginal and that study funds may be better used by recruiting more study participants rather than collecting additional measurements per subject. Additional studies are required to determine the public health and clinical impact of these findings and to explore whether accounting for the hormonal trajectories within each woman in reference intervals (i.e., via creating individual-level reference intervals) may help mitigate the potential adverse birth outcomes associated with subclinical thyroid dysfunction.

Our study was limited by sample size for the evaluation of subclinical thyroid disease (e.g., subclinical hypothyroidism and hypothyroxinemia) and by the fewer ultrasound measurements available at later time points in pregnancy. As previously reported, ultrasound scans at visits 3 (median 26 weeks) and 4 (median 35 weeks) were not routine and were requested if pregnancy complications were suspected. However, receiving an ultrasound scan later in pregnancy was not associated with thyroid hormone concentrations in our study population, thereby reducing the possibility of differential bias in our results. Additionally, we did not measure maternal antithyroid antibodies, the presence of which may increase the risk of adverse maternal and birth outcomes independent of thyroid function (33).

Strengths of our study include the availability of up to four repeated measures of thyroid hormone concentrations, which allowed us to investigate potential windows of susceptibility in pregnancy to thyroid hormone disruption. Repeated ultrasound measurements, standardized to those from a large population of pregnant women in Boston, allowed us to explore the influence of subclinical changes in thyroid function in gestation on intrauterine growth. Finally, our assay method of measuring FT4 using direct equilibrium dialysis followed by radioimmunoassay is preferred over traditional immunoassays because it is not influenced by serum binding proteins, which increase dramatically in pregnancy (57, 58).

Conclusions

Among pregnant women without clinical thyroid disease, we observed consistent inverse associations between FT4 and fetal growth indicators, including birth weight. Future animal and human health studies are needed to elucidate the biological mechanisms underlying the relationship between FT4 and fetal growth in generally euthyroid pregnancies.

Abbreviations:

AC

abdominal circumference

BMI

body mass index

CI

confidence interval

EFW

estimated fetal weight

FL

femur length

FT4

free thyroxine

HC

head circumference

IUGR

intrauterine growth restriction

LBW

low birth weight

LMM

linear mixed model

SGA

small for gestational age

T3

triiodothyronine

T4

thyroxine

TSH

thyroid-stimulating hormone.