Higher Maternal Protein Intake during Pregnancy Is Associated with Lower Cord Blood Concentrations of Insulin-like Growth Factor (IGF)-II, IGF Binding Protein 3, and Insulin, but Not IGF-I, in a Cohort of Women with High Protein Intake

Abstract

Background: Prenatal exposure to dietary protein may program growth-regulating hormones, consequently influencing early-life growth patterns and later risk of associated chronic diseases. The insulin-like growth factor (IGF) axis is of particular interest in this context given its influence on pre- and postnatal growth and its sensitivity to the early nutritional environment.

Objective: Our objective was to examine associations of maternal protein intake during pregnancy with cord blood concentrations of IGF-I, IGF-II, IGF binding protein-3 (IGFBP-3), and insulin.

Methods: We studied 938 mother-child pairs from early pregnancy through delivery in the Project Viva cohort. Using multivariable linear regression models adjusted for maternal race/ethnicity, education, income, smoking, parity, height, and gestational weight gain and for child sex, we examined associations of second-trimester maternal protein intake [grams per kilogram (weight before pregnancy) per day], as reported on a food frequency questionnaire, with IGF-I, IGF-II, IGFBP-3, and insulin concentrations in cord blood. We also examined how these associations may differ by child sex and parity.

Results: Mothers were predominantly white (71%), college-educated (64%), and nonsmokers (67%). Mean ± SD protein intake was 1.35 ± 0.35 g ⋅ kg⁻¹ ⋅ d⁻¹. Each 1-SD increment in second-trimester protein intake corresponded to a change of −0.50 ng/mL (95% CI: −2.26, 1.26 ng/mL) in IGF-I and −0.91 μU/mL (95% CI: −1.45, −0.37 μU/mL) in insulin. Child sex and parity modified associations of maternal protein intake with IGF-II and IGFBP-3: protein intake was inversely associated with IGF-II in girls (P-interaction = 0.04) and multiparous mothers (P-interaction = 0.05), and with IGFBP-3 in multiparous mothers (P-interaction = 0.04).

Conclusions: In a cohort of pregnant women with relatively high mean protein intakes, higher intake was associated with lower concentrations of growth-promoting hormones in cord blood, suggesting a pathway that may link higher protein intake to lower fetal growth. This trial was registered at clinicaltrials.gov as NCT02820402.

Introduction

Evidence from both animal and epidemiologic studies suggests that prenatal nutritional exposures may program growth-regulating hormones, growth patterns early in life, and, consequently, the risk of associated chronic diseases in later life. The insulin-like growth factor (IGF) axis is of particular interest in this context because of its major influence on growth during gestation (1–3) infancy (4, 5) and childhood (6), as well as its sensitivity to the early nutritional environment (5, 7–9). The 2 ligands of the IGF axis, IGF-I and IGF-II, are both expressed in the fetal bloodstream and tissues throughout gestation and stimulate fetal and placental growth in response to nutrient availability. IGF-I is directly correlated with birth weight, and concentrations are higher in large-for-gestational-age newborns (7, 10). IGF-II is the primary regulator of placental growth, nutrient transfer, and embryonic growth (7, 8).

In contrast to the later postnatal period, when IGF-I is largely regulated by growth hormone, fetal and infant IGF concentrations respond primarily to the nutritional environment. IGF-I decreases in utero when nutrient supply is restricted and increases in response to insulin, which signals adequate availability of nutrients (1, 7, 8). Postnatal protein intake in particular seems to play a major role in determining IGF-I and, as a consequence, patterns of linear growth in infancy and childhood (11, 12). Multiple studies have found that nonbreastfed infants, who consume higher amounts of protein than breastfed infants, have higher IGF-I concentrations during infancy (4, 5, 9, 12–15); in 2 separate randomized trials, infants receiving formula with a high protein content had higher IGF-I than infants fed a low-protein formula (13, 14). One of these trials found sex-specific effects, with stronger associations of a high-protein diet with IGF axis parameters among girls (16).

Associations of maternal protein intake with hormones that regulate growth in utero have received little attention. In addition, most of the literature examining postnatal nutritional influences on growth-promoting hormones has focused on IGF-I, but other hormones may be equally or more important in mediating the relation between maternal nutrition and fetal growth. In addition to IGF-II, the IGF binding proteins (IGFBPs), which limit the bioactivity of IGFs and possibly have independent effects on fetal growth (4, 9, 10), may also respond to nutritional status. Insulin, which is structurally similar to IGF-I, is an important independent regulator of fetal growth and stimulates release of IGF-I and IFG-II in response to the availability of glucose, a marker of maternal nutritional status (17, 18). The effect of maternal nutrition on fetal hormones may be modified by other factors such as child sex and parity. Our primary objective in this study was to examine associations of maternal protein intake during pregnancy with cord blood concentrations of IGF-I, IGF-II, IGFBP-3, and insulin in a well-characterized prospective cohort of mother-child pairs. We previously observed an inverse association of maternal protein intake and markers of fetal, infant, and child growth in this cohort (19). To better elucidate the role of maternal nutrition, we examined total intake of protein and of different sources of protein, specifically animal protein, plant protein, milk, and dairy products. Evaluation of the relations between protein intake during pregnancy and cord blood concentrations of IGF axis hormones will contribute to understanding the pathways by which maternal nutrition influences fetal growth.

Methods

Participants

We studied mother-child pairs enrolled in Project Viva, a prospective cohort study examining pre- and perinatal factors in relation to pregnancy and child health outcomes. We recruited women during 1999–2002 at their initial obstetric appointment from 8 offices of Atrius Health, a large, multispecialty group practice in eastern Massachusetts. Exclusion criteria included multiple gestation, inability to answer questions in English, gestational age >22 wk at the time of the initial obstetric appointment, and plans to move out of the local area before delivery. Women who agreed to participate (65% of those eligible) and provided informed consent completed the first study visit after their appointment. They ranged from 4.8 to 23.7 wk of gestation (mean: 10.5 wk) at this visit. Most women completed the second study visit at 26–28 wk of gestation, and they completed the third study visit at the hospital 1–3 d after delivery. Detailed recruitment and retention procedures have been described previously (20). Institutional review boards of participating institutions approved the study protocols, and participants provided written informed consent.

The Project Viva cohort consists of the 2128 women who delivered a live infant and their children. For this analysis, we included 938 mother-child pairs with available data from the second trimester of pregnancy and cord blood hormone concentrations. Cord blood was collected at only 1 of the 2 main delivery hospitals by different clinicians, whose primary focus was on clinical care, not research; cord blood was not collected from preterm or complicated deliveries when the infant went directly to the neonatal intensive care unit. Mean maternal protein intake during the second trimester was similar between the included and excluded participants (1.35 and 1.36 g ⋅ kg⁻¹ ⋅ d⁻¹, respectively). No differences were found in maternal educational level, parity, smoking history, prepregnancy BMI, age at enrollment, gestational weight gain, household income, or child sex between included and excluded participants. Maternal race/ethnicity was somewhat different between the 2 groups (P < 0.001), with a higher proportion of white mothers in the included group (71% compared with 62% in the excluded group). Mean gestational age at birth was slightly higher among included participants (39.7 compared with 39.3 wk; P < 0.0001).

Measurements

Assessment of maternal nutrient intake.

We derived data on maternal protein (total and from plant and animal sources), carbohydrate, fat, milk, and dairy intake from self-administered semiquantitative FFQs that mothers completed during the first and second study visits. The 166-item FFQ used in Project Viva was based on instruments validated in other cohorts, including the Nurses’ Health Study (21), and modified for use in pregnancy (22). The reference time period for the first FFQ, administered at enrollment, was the time since the last menstrual period. The reference period for the second FFQ, administered at 26–28 wk of gestation, was the past 3 mo (roughly corresponding to the second trimester). The FFQ included reference portion sizes (e.g., 8-oz glass) for each food item. Using the Harvard nutrient composition database, which includes food composition values from the USDA and is supplemented by other sources, we calculated mean daily protein intakes by multiplying a weight assigned to the reported frequency of intake of each protein-containing food by the protein content for the portion size specified in the FFQ (23). The same method was used to calculate carbohydrate and fat intakes. To address potential measurement error often attributed to the use of FFQs in assessing dietary intake, we adjusted individual nutrient estimates for total energy intake using the nutrient residual method (24). Adjustment for total energy intake has been shown to reduce the impact of measurement error inherent in applying FFQs specifically to assess protein intake (25).

The estimated average requirement, or median nutrient requirement for a given life stage and sex group, is based on individual prepregnancy weight. Therefore, we calculated first- and second-trimester protein intakes per kilogram of prepregnancy body weight using energy-adjusted protein intakes, and self-reported prepregnancy weights that have been previously validated in this cohort (26). Plots of the association of quantiles of protein intake (in grams per day and grams per kilogram of prepregnancy weight per day) with cord blood hormones revealed a more linear relation when protein intake was adjusted for prepregnancy weight. This weight-adjusted variable was used in all analyses.

Mothers reported skim, 1% or 2%, and whole-milk intake on the FFQ using 8 categories ranging from never or <1 time/mo to ≥4 glasses/d. We coded milk intake into an 8-category variable indicating daily servings of each type of milk; the value for each category was set to the mean within each FFQ category (e.g., 2–4 glasses/wk was coded as 3 glasses/wk, or 0.43 glasses/d). Because we were interested specifically in protein rather than other components of milk, and because the various types of milk have similar protein content (27), we added the values for the 3 types of milk (skim, 1% or 2%, and whole) to create a single variable for number of milk servings per day. For intake of dairy products, we added the number of daily milk servings to the number of servings of other dairy products (e.g., cheese, yogurt) to create a single variable indicating the number of daily servings of all dairy products, including milk.

Measurement of cord blood hormone concentrations.

Our primary outcomes were umbilical cord plasma concentrations of IGF-I, IGF-II, IGFBP-3, and insulin. Clinicians collected cord blood samples via syringe from the umbilical vein after delivery. Whole-blood samples were refrigerated for <24 h, centrifuged at 2000 × g at 4°C for 10 min, and then measured into aliquots for storage in liquid nitrogen (28). We measured insulin (competitive electrochemiluminescence immunoassay; Roche Diagnostics), IGF-I and IGFBP-3 (ELISA; R&D Systems), and IGF-II (ELISA; Alpco Diagnostics). Day-to-day variability for each of these assays was <10%.

Covariates.

At the initial prenatal visit, mothers self-reported their educational level, smoking history, height, race/ethnicity, parity, household income, age, prepregnancy weight, and height of the baby’s biological father (paternal height). We obtained the child’s sex from an interview conducted at the visit just after delivery. All data collection instruments used in Project Viva are publicly available (https://www.hms.harvard.edu/viva/). We extracted serial prenatal weights from medical records and used those weights to calculate gestational weight gain (GWG) ≤91 d of gestation (first-trimester GWG) and from 91 to 182 d of gestation (second-trimester GWG) (29). In our analyses, we did not adjust for third-trimester or total GWG because both might be affected by the second-trimester diet. We defined GWG categories according to 2009 Institute of Medicine recommendations (30). We defined categories of gestational glucose tolerance using results of routine prenatal screening (nonfasting 50-g oral glucose load) done at 26–28 wk of gestation, with a subsequent fasting 3-h, 100-g oral-glucose-tolerance test if the blood glucose exceeded 140 mg/dL (31).

Statistical analysis.

We examined associations of maternal protein intake and cord blood hormone concentrations with categories of maternal and infant characteristics using global (type 3) P value testing to identify potential confounders. We assessed correlations between concentrations of the 4 different cord blood hormones using Spearman correlation analysis. We analyzed data using sequential multivariable linear regression models adjusted for 1) maternal race/ethnicity, parity, height, and child sex; 2) model 1 covariates + maternal education, household income at enrollment, maternal smoking, and first- and second-trimester GWG; 3) model 2 covariates + IGFBP-3 (for models examining IGF-I or IGF-II as the outcome); and 4) model 2 covariates + total protein (when animal or plant protein was the main predictor). We adjusted models predicting IGF-I and IGF-II for IGFBP-3 in order to estimate associations with free (unbound) IGF-I and IGF-II concentrations, in addition to total concentrations. All models met standard assumptions for linear regression. Given interactions between other exposures and child sex when examining associations with cord blood hormones in our cohort (31), as well as evidence in the literature that associations between infant protein intake and growth-related hormones may differ by sex (16, 32), we decided a priori to look for protein-by-sex interactions and examine models for males and females separately, in addition to looking at overall effects. We also examined protein-by-parity interactions, based on evidence that concentrations of growth-promoting hormones in cord blood and the relations of these hormones with birth size may differ in nulliparous compared with multiparous pregnancies (2).

Statistical models containing only one macronutrient (protein, carbohydrate, or fat) and adjusted for total energy intake do not isolate the effect of the nutrient included in the model because an increase in one macronutrient is accompanied by a decrease in one or both of the other 2 macronutrients. We performed a sensitivity analysis using models adjusted for the other macronutrients in order to determine whether the associations of higher protein intake with lower cord blood hormone concentrations might be attributable to replacement of the other macronutrients with protein (33). In addition, we examined the fully adjusted models with additional adjustment for first-trimester protein intake (grams per kilograms prepregnancy body weight per day) to check for lingering effects of earlier protein intake.

We examined protein intakes during the first and second trimesters separately for our primary analysis of associations between maternal protein intake and cord blood hormones. Results were similar, but models using second-trimester protein had slightly stronger regression coefficients. Because fetal growth occurs primarily during the second and third trimesters, we present results of analyses considering second-trimester protein intake as the primary exposure. We categorized protein into quartiles, and because associations were reasonably incremental across quartiles, we examined protein intake as a continuous exposure and modeled the effect per 1-SD increment (0.35 g ⋅ kg⁻¹ ⋅ d⁻¹ for second-trimester intake).

We used multiple imputation to impute missing dietary and covariate data. We generated 50 imputed data sets using chained imputation (34) and combined estimates using Rubin's rules (35). We present results from the imputed analysis throughout the article unless otherwise indicated. All 2128 participants were included to generate the imputed data set, but only the 938 participant pairs (mother-infant) with second-trimester and cord blood hormone data were included in the analyses. We used hypothesis-driven models to look for trends and consistency of results across methods, and we interpreted all results in the context of our prespecified hypotheses. We performed all analyses using SAS version 9.3 (SAS Institute) and considered P values <0.05 to be statistically significant.

Results

Table 1 shows characteristics of the study population and bivariate associations of maternal and infant characteristics with second-trimester protein intake and cord blood hormone concentrations. The mothers included in this study were predominantly white (71%), college-educated (64%), and never smoked (67%); 54% were parous. Mean ± SD second-trimester protein intake was 1.35 ± 0.35 g ⋅ kg⁻¹ ⋅ d⁻¹. Protein intake was associated with sociodemographic factors (maternal education, race/ethnicity, and household income), maternal smoking history, prepregnancy BMI, GWG, parity and height, and infant birth weight z score. Female infants had higher cord blood concentrations of IGF-I, IGF-II, IGFBP-3, and insulin, and among both sexes, concentrations of all hormones were higher with higher birth weight–for–gestational age z score. IGF-I was directly associated with birth length and maternal height.

TABLE 1.

Second-trimester protein intake and cord blood hormone concentrations, by category of maternal and other characteristics, in 938 Project Viva participants¹

			Cord blood hormones
	Participants (n = 938), %	Second-trimester protein intake, g ⋅ kg−1 ⋅ d−1	IGF-I, ng/mL	IGF-II, ng/mL	IGFBP-3, ng/mL	Insulin, μU/mL
Overall	938	1.35 ± 0.35	56.5 ± 24.4	409 ± 92.8	1084 ± 318	6.62 ± 7.32
Maternal education
<4-y college degree	36	1.24 ± 0.37	57.5 ± 24.3	404 ± 93	1086 ± 325	6.69 ± 7.63
≥4-y college degree	64	1.41 ± 0.33	55.9 ± 24.5	412 ± 93	1084 ± 314	6.58 ± 7.15
P value		<0.0001	0.33	0.22	0.92	0.83
Maternal race/ethnicity
White	71	1.36 ± 0.33	56.9 ± 24.5	412 ± 94	1098 ± 321	6.08 ± 6.42
Asian	5	1.65 ± 0.40	55.8 ± 25.0	417 ± 90	1082 ± 324	9.38 ± 14.0
Black	14	1.22 ± 0.38	56.1 ± 25.1	398 ± 84	1056 ± 329	7.92 ± 7.71
Hispanic	6	1.26 ± 0.34	54.6 ± 22.6	411 ± 89	1049 ± 240	6.36 ± 5.06
Other	4	1.28 ± 0.37	53.8 ± 21.1	381 ± 106	983 ± 300	8.56 ± 9.85
P value		<0.0001	0.91	0.22	0.81	<0.01
Annual household income at enrollment
≤$70,000	42	1.28 ± 0.38	55.5 ± 25.1	413 ± 96	1086 ± 335	6.70 ± 6.28
>$70,000	58	1.40 ± 0.34	57.2 ± 24.5	406 ± 94	1083 ± 316	6.56 ± 8.14
P value		<0.0001	0.29	0.31	0.90	0.77
Maternal smoking history
Never	67	1.37 ± 0.36	57.6 ± 24.3	410 ± 92	1089 ± 310	6.42 ± 5.78
Before pregnancy	20	1.36 ± 0.31	56.8 ± 26.1	409 ± 101	1103 ± 362	6.91 ± 9.68
During pregnancy	13	1.23 ± 0.37	50.0 ± 21.2	405 ± 85	1033 ± 282	7.19 ± 10.2
P value		<0.001	<0.01	0.88	0.14	0.48
Maternal age at enrollment, y
<30	31	1.33 ± 0.37	53.5 ± 23.7	412 ± 92	1064 ± 326	6.04 ± 7.49
30 to <40	64	1.36 ± 0.35	58.2 ± 25.0	409 ± 93	1098 ± 317	6.96 ± 7.38
≥40	5	1.31 ± 0.27	52.7 ± 16.5	389 ± 92	1026 ± 252	5.70 ± 4.53
P value		0.32	0.02	0.34	0.15	0.15
Prepregnancy BMI, kg/m2
<18.5	4	1.74 ± 0.34	64.5 ± 27.3	427 ± 96	1186 ± 350	4.58 ± 7.89
18.5–24.9	59	1.49 ± 0.29	54.8 ± 22.4	399 ± 89	1051 ± 294	6.00 ± 6.88
25.0–29.9	22	1.18 ± 0.25	57.2 ± 26.4	425 ± 96	1117 ± 341	7.43 ± 4.16
≥30	15	0.94 ± 0.24	59.7 ± 27.3	419 ± 98	1140 ± 346	8.34 ± 7.53
P value		<0.0001	0.03	<0.01	<0.001	<0.001
Gestational weight gain
Inadequate	13	1.40 ± 0.38	57.8 ± 25.8	390 ± 82	1036 ± 291	5.67 ± 5.30
Adequate	29	1.42 ± 0.36	55.7 ± 23.2	403 ± 92	1070 ± 306	5.99 ± 5.11
Excessive	58	1.30 ± 0.34	56.6 ± 25.0	416 ± 96	1102 ± 332	7.14 ± 8.54
P value		<0.0001	0.73	0.01	0.08	0.03
Parity
0	46	1.39 ± 0.35	48.5 ± 21.6	407 ± 94	1030 ± 289	5.89 ± 5.78
≥1	54	1.32 ± 0.36	63.2 ± 24.6	410 ± 92	1130 ± 333	7.23 ± 8.36
P value		<0.01	<0.0001	0.60	<0.0001	<0.01
Maternal height, m
≤1.57	10	1.54 ± 0.41	51.0 ± 24.0	397 ± 89	1029 ± 278	7.52 ± 15.4
>1.57 to 1.68	61	1.39 ± 0.35	55.8 ± 24.0	408 ± 95	1080 ± 328	6.34 ± 5.62
>1.68	29	1.21 ± 0.30	59.7 ± 24.9	414 ± 88	1111 ± 307	6.92 ± 9.63
P value		<0.0001	<0.01	0.30	0.31	0.26
Maternal glucose tolerance
Normal	82	1.36 ± 0.35	55.9 ± 23.8	407 ± 92	1074 ± 312	6.15 ± 6.65
Isolated hyperglycemia	9	1.31 ± 0.36	56.3 ± 23.2	406 ± 95	1084 ± 307	8.06 ± 7.79
Impaired glucose tolerance	3	1.23 ± 0.30	60.9 ± 32.0	430 ± 118	1152 ± 391	6.64 ± 5.46
Gestational diabetes mellitus	6	1.29 ± 0.40	61.8 ± 29.5	432 ± 94	1178 ± 370	10.7 ± 13.3
P value		0.07	0.24	0.14	0.07	<0.0001
Child sex
Male	52	1.33 ± 0.35	52.4 ± 22.9	402 ± 92	1027 ± 300	6.29 ± 7.07
Female	48	1.37 ± 0.37	60.9 ± 25.2	416 ± 93	1147 ± 325	6.98 ± 7.57
P value		0.09	<0.0001	0.02	<0.0001	0.15
Gestational age at birth, wk
<37	5	1.28 ± 0.42	52.0 ± 22.1	359 ± 108	1055 ± 329	10.3 ± 11.0
37–42	93	1.35 ± 0.35	56.9 ± 24.5	412 ± 92	1087 ± 318	6.44 ± 7.05
>42 to 43	2	1.38 ± 0.34	48.1 ± 19.8	410 ± 74	1054 ± 278	5.28 ± 5.22
P value		0.37	0.17	<0.001	0.75	<0.01
Birth length, cm
Q1, 39.1–48.2	22	1.36 ± 0.46	51.5 ± 26.3	392 ± 99	1045 ± 362	6.96 ± 11.8
Q2, 48.3–49.7	25	1.37 ± 0.43	55.0 ± 27.5	413 ± 105	1085 ± 381	6.48 ± 7.88
Q3, 49.8–51.1	26	1.36 ± 0.37	57.7 ± 29.0	412 ± 102	1093 ± 338	6.44 ± 6.67
Q4, 51.2–56.1	27	1.31 ± 0.35	60.6 ± 27.5	416 ± 108	1106 ± 348	6.65 ± 6.96
P value		0.44	<0.01	0.08	0.34	0.92
Birth weight–for–gestational age z score2
Q1, −2.58 to −0.51	23	1.38 ± 0.41	43.3 ± 19.8	393 ± 85	967 ± 260	4.99 ± 7.95
Q2, −0.48 to 0.14	25	1.38 ± 0.37	53.8 ± 22.5	408 ± 93	1056 ± 306	5.93 ± 5.87
Q3, 0.14–0.86	26	1.36 ± 0.32	59.8 ± 22.6	406 ± 92	1110 ± 327	7.06 ± 8.29
Q4, 0.90–2.58	26	1.29 ± 0.33	67.1 ± 25.7	427 ± 97	1188 ± 328	8.24 ± 6.63
P value		0.01	<0.0001	<0.01	<0.0001	<0.0001

¹Data are means ± SDs unless otherwise indicated. P values are from global (type 3) P value testing for difference in outcome across categories of maternal and other characteristics. IGF, insulin-like growth factor; IGFBP-3, insulin-like growth factor binding protein-3; Q, quartile.

²Ranges are the actual observed values in each quartile.

IGF-II was higher with greater GWG. Insulin increased with higher maternal prepregnancy BMI and GWG, decreased with older gestational age at birth, and was higher in the cord blood of infants born to mothers of Asian ethnicity, black race, or other race or ethnicity, and to mothers with gestational diabetes. Infants born to parous mothers had higher concentrations of IGF-I, IGFBP-3, and insulin. IGF-I, IGF-II, and IGFBP-3 were lowest in infants born to mothers with a prepregnancy BMI in the normal range, and IGF-I was lowest in infants of mothers who smoked during pregnancy.

All cord blood hormones were positively correlated with each other, but the strongest correlations occurred between IGFBP-3 and both IGF-I (Spearman r = 0.65; P < 0.0001) and IGF-II (Spearman r = 0.65; P < 0.0001). IGF-I was moderately correlated with IGF-II (Spearman r = 0.28; P < 0.0001) and insulin (Spearman r = 0.39; P < 0.0001). Correlations between all hormones were reported previously in a similar subset of the Project Viva cohort (31).

In multivariable regression models adjusted for maternal race/ethnicity, parity, and height, and child sex, each 1-SD (0.35 g ⋅ kg⁻¹ ⋅ d⁻¹) increment in second-trimester total protein intake corresponded to changes of −9.99 ng/mL (95% CI: −16.7, −3.25 ng/mL) in IGF-II, −27.6 ng/mL (95% CI: −50.0, −5.11 ng/mL) in IGFBP-3, and −0.73 μU/mL (95% CI: −1.25, −0.21 μU/mL) in insulin concentrations in cord blood. These associations were strengthened after further adjustment for maternal education, household income at enrollment, maternal smoking, and GWG through the second trimester [β: −11.5 (95% CI: −18.6, −4.49) for IGF-II; β: −33.9 (95% CI: −57.4, −10.4) for IGFBP-3; and β: −0.91 (95% CI: −1.45, −0.37) for insulin]. When we further adjusted the model predicting IGF-II for IGFBP-3, the estimate was attenuated [β: −4.51 (95% CI: −9.66, 0.63)]. No evidence showed an association of second-trimester protein intake with IGF-I in any of the models (Supplemental Table 1, Figure 1). Associations of first-trimester total protein intake with cord blood hormones were very similar to associations with second-trimester intake (Supplemental Table 2).

FIGURE 1.

Adjusted β coefficients and 95% CIs for the association of second-trimester maternal protein intake with cord blood concentrations of IGF-I (A), IGF-II (B), IGFBP-3 (C), and insulin (D) in 938 mother-child pairs in Project Viva, and by child sex (450 girls and 488 boys) and parity (431 nulliparous, 507 multiparous). Points indicate β coefficients for the change in concentration of each hormone corresponding to a 1-SD (0.35 g ⋅ kg⁻¹ ⋅ d⁻¹) increment in second-trimester maternal protein intake; errors bars indicate 95% CIs. Data are from multivariable linear regression models adjusted for maternal race/ethnicity, height, education, smoking, gestational weight gain through the second trimester, household income at enrollment, and parity (except models stratified by parity), and child sex (except models stratified by sex). Protein intake was adjusted for total energy intake. P-interaction values are for protein-by-sex and protein-by-parity interactions. IGF, insulin-like growth factor; IGFBP-3, insulin-like growth factor binding protein-3.

We were interested in whether associations with cord blood hormones were derived from total protein or specific types of protein, that is, protein from animal or plant sources, dairy products, or milk. Therefore we repeated all analyses using second-trimester maternal intake of each of the following dietary components as the main predictor: animal protein (grams per kilogram per day), plant protein (grams per kilogram per day), milk (servings per day), and all dairy products (servings per day). Results for associations of animal and plant protein with cord blood hormones were similar to those for total protein (Supplemental Table 1), and when these models were also adjusted for total protein, the associations were not statistically significant. We did not find any evidence for associations of milk or total dairy product intake with any of the cord blood hormones studied (Supplemental Table 3).

Figure 1 displays the interactions of second-trimester protein intake with child sex and parity. In models adjusted for all covariates (maternal race/ethnicity, parity, height, education, smoking status, household income, and GWG through the second-trimester), each 1-SD (0.35 g ⋅ kg⁻¹ ⋅ d⁻¹) increment in second-trimester total protein intake corresponded to changes in cord blood IGF-II of −20.7 ng/mL (95% CI: −30.7, −10.8 ng/mL) in girls and −2.76 ng/mL (95% CI: −12.8, 7.31 ng/mL) in boys (P-interaction = 0.04). In the sex-stratified models, we found an inverse association between maternal protein intake and cord blood IGFBP-3 in girls only [β: −47.0 (95% CI: −81.2, −12.9) for girls; β: −20.2 (95% CI: −52.7, 12.3) for boys], but the interaction was not statistically significant (P = 0.23). In models adjusted for maternal race/ethnicity, height, education, smoking status, household income, GWG through the second trimester, and child sex, each 1-SD (0.35 g ⋅ kg⁻¹ ⋅ d⁻¹) increment in second-trimester total protein intake corresponded to changes in cord blood IGF-II of −5.41 ng/mL (95% CI: −15.6, 4.79 ng/mL) in nulliparous mothers and −16.8 ng/mL (95% CI: −26.8, −6.76 ng/mL) in multiparous mothers (P-interaction = 0.05), and to changes in cord blood IGFBP-3 of −7.77 ng/mL (95% CI: −38.7, 23.2 ng/mL) in nulliparous mothers and −52.7 ng/mL (−88.1, −17.3 ng/mL) in multiparous mothers (P-interaction = 0.04). In the stratified models, maternal protein intake was associated with cord blood insulin concentrations in boys and nulliparous mothers only, but neither the protein-by-sex nor the protein-by-parity interaction terms were statistically significant for insulin. There was no association with IGF-I in any of the subgroups.

When we included total second-trimester fat or carbohydrate intake (grams per day) in the models adjusted for all other covariates except IGFBP-3, the effect of increasing protein intake (while holding constant carbohydrate or fat, respectively) was similar to that observed when protein was examined without adjusting for other macronutrients. The association of higher total protein with lower insulin concentrations was stronger when also adjusted for carbohydrate intake; in all other models, the β coefficient changed by <10% (Table 2). When we examined the fully adjusted models with additional adjustment for first-trimester protein intake (grams per kilogram per day), the associations of second-trimester protein intake with cord blood hormones were in the same direction but attenuated, likely because of the high collinearity between first- and second-trimester protein intakes (Spearman r = 0.8; data not shown).

TABLE 2.

Associations of second-trimester maternal total protein intake with cord blood hormones, adjusted for other macronutrients, among 938 mother-child pairs in Project Viva¹

	Total protein3
Second-trimester maternal intake2	Model 14	Model 25	Model 36
IGF-I, ng/mL	−0.50 (−2.26, 1.26)	−0.46 (−2.22, 1.30)	−0.36 (−2.20, 1.47)
IGF-II, ng/mL	−11.5 (−18.6, −4.49)	−11.2 (−18.2, −4.09)	−10.4 (−17.8, −3.04)
IGFBP-3, ng/mL	−33.9 (−57.4, −10.4)	−33.2 (−56.7, −9.74)	−33.2 (−57.7, −8.69)
Insulin, μU/mL	−0.91 (−1.45, −0.37)	−0.93 (−1.48, −0.39)	−1.13 (−1.70, −0.56)

¹IGF, insulin-like growth factor; IGFBP-3, insulin-like growth factor binding protein-3.

²Intake was measured in grams per kilogram prepregnancy weight per day.

³The coefficients are for the change in outcome per 1-SD (0.35 g ⋅ kg⁻¹ ⋅ d⁻¹) increment in total protein. Data are β coefficients (95% CIs), based on multivariable linear regression models. Protein intake is adjusted for total energy intake.

⁴Adjusted for maternal race/ethnicity, parity, height, education, smoking, gestational weight gain through the second trimester, household income at enrollment, and child sex.

⁵Adjusted for covariates in model 1 plus second-trimester total fat intake (grams per day).

⁶Adjusted for covariates in model 1 plus second-trimester carbohydrate intake (grams per day).

Discussion

In 938 mother-child pairs, higher maternal protein intake during the second trimester of pregnancy was associated with lower umbilical cord blood concentrations of several key growth-regulating hormones. The relation with IGF-II is particularly interesting given its role in regulating placental growth and nutrient transfer, which in turn provides substrates for fetal growth. Associations were driven by total protein intake rather than protein from a specific source.

A direct association has been established between protein intake during infancy and IGF-I (4, 5, 9, 12, 16), yet our results suggest that maternal protein intake during pregnancy is inversely related to fetal concentrations of IGF axis hormones. Animal models of maternal nutrient restriction show that fetal production of these hormones responds to maternal nutritional status (36, 37). To our knowledge, maternal overnutrition has not been studied extensively, and most previous work has focused on excessive fetal nutrient supply in the context of maternal hyperglycemia. Ascertaining fetal nutrient exposure is complex because the maternal diet provides nutrients to the fetus indirectly via the placenta, and higher maternal intake does not necessarily translate to higher fetal exposure. One potential explanation for our findings is that higher protein intake by the mother actually results in lower availability of amino acids to the fetus, which responds to this nutrient restriction with reductions in growth-regulating hormones. The animal literature suggests that maternal diet composition affects placental nutrient transport capacity and that downregulation of placental amino acid transporters can impair fetal growth (38–40). In sheep, which are similar to humans in placental transport and metabolism of amino acids, simultaneous infusion of multiple amino acids resulted in competitive inhibition of umbilical uptake of some amino acids (41). The inverse association that we observed between protein intake and IGF-II, which is involved in placental nutrient transfer, supports this hypothesis.

We found that maternal protein intake was inversely associated with cord blood concentrations of IGF-II but was not associated with IGF-I. This is consistent with a study comparing cord blood hormones in mono- and dizygotic twins with those in gestational age–matched nonsibling singleton pairs, which found that cord blood IGF-I concentrations are predominantly regulated by genetics, whereas both environmental and genetic factors regulate IGF-II (42). Furthermore, IGF2 gene expression is more abundant than IGF1 expression in fetal tissues, fetal and cord blood concentrations of IGF-II are 3–10 times higher than those of IGF-I, and IGF-II overexpression leads to fetal overgrowth. This evidence suggests that IGF-II is the primary IGF involved in growth in utero, with a shift to growth regulation by IGF-I after birth (7, 8). We also observed inverse associations of maternal protein intake with cord blood IGFBP-3 and insulin. Although IGFBP-3 does not have a clear independent role in fetal growth, it binds IGF-I and IGF-II and extends their half-lives in circulation (43). Thus, high maternal protein intake may reduce total IGF-II both directly and indirectly via IGFBP-3 (7, 17). The association of maternal protein intake with IGF-II was attenuated after adjusting for IGFBP-3, possibly because of the high correlation between the 2 hormones. Fetal insulin stimulates fetal IGF-I and IGF-II production in response to nutrient availability (10, 17, 18), and the observed inverse association between maternal protein intake and insulin may reflect the downregulation of placental amino acid transporters with higher protein intake.

Given that the IGF axis is a key regulator of fetal growth and is sensitive to nutritional status during gestation, it is plausible that it mediates the associations between maternal protein intake and fetal growth observed in epidemiological studies (44, 45). Based on our observations, we would expect to see evidence of lower fetal growth in babies born to mothers with high protein intake during pregnancy. In fact, we recently reported an inverse association of maternal second-trimester protein intake with both birth weight and length for gestational age, as well as linear growth during infancy and childhood, in this cohort (19). The lower cord blood concentrations of growth-promoting hormones observed in the present study may reflect a pathway by which maternal protein intake and fetal amino acid supply influence fetal growth.

Studies have identified associations of specific protein sources, including animal protein, dairy products, and milk, with both hormones and growth (46, 47), yet our results were driven by total protein. Furthermore, adjusting the multivariable models examining second-trimester total protein intake as the main exposure for the other macronutrients did not appreciably alter the results, increasing our confidence that the observed estimates reflect associations with higher total protein intake, rather than with corresponding decreases in other macronutrients.

The association of maternal protein intake with cord blood IGF-II was stronger in girls. A few studies have demonstrated that offspring sex modifies associations of protein intake with growth parameters. In rats, one study demonstrated an association between a high-protein diet during pregnancy and higher body weight and fat mass in female offspring (48), and another showed that a high-protein diet had sex-specific associations with weight and other cardiometabolic parameters (49). In the EU Childhood Obesity Programme, high-protein infant formula was more strongly associated with higher IGF-I and lower IGFBP-2 in females than was low-protein formula (16). Also, in another Project Viva analysis that included all of the participants in the current sample, we observed sex-specific differences in associations of maternal gestational glycemia with cord blood concentrations of IGF-I, IGF-II, IGFBP-3, and insulin (31). Specifically, cord blood IGF-I was higher in girls born to mothers with gestational diabetes, but IGF-II, IGFBP-3, and insulin were higher in boys born to mothers with gestational diabetes, suggesting that males are more sensitive to hyperglycemia in utero. By contrast, our results indicate greater sensitivity to higher maternal protein intake in females and show that higher maternal protein intake is associated with lower cord blood hormone concentrations. Maternal glycemic status is a more direct measure of fetal overnutrition, as glucose is readily transported across the placenta for use as a fetal energy source (50). The discrepancy between the higher cord blood hormone concentrations in response to hyperglycemia, which is specifically driven by metabolism of carbohydrates and fats to glucose, and both the lower concentrations of the same hormones and lower measures of fetal growth in response to higher maternal protein intake (19), further supports the possibility that higher maternal protein intake translates to relative restriction of the availability of growth-promoting nutrients to the fetus. We observed stronger associations of maternal protein intake with IGF-II and IGFBP-3 in parous women. The explanation for this is unclear. One study found stronger associations of birth weight with cord blood IGF-I and IGF-II:IGF2R in multiparous pregnancies, and the authors suggested that maternal restraint of fetal growth in first pregnancies may override the contributions of other growth-promoting factors (2). Our results may reflect this proposed restriction of the fetal growth response to external factors, including nutrition, in nulliparous women.

Our study had a large sample size, prospectively assessed maternal diet at 2 times during pregnancy, and provided detailed measurements of many covariates. Limitations include the availability of cord blood from only about half of the cohort. Also, our sample comprised women of relatively high socioeconomic status, and results may not be generalizable to all populations of pregnant women. Protein intake in Project Viva mothers may be higher than that in other populations of women of childbearing age (51, 52). FFQs have some limitations in assessing nutritional exposures. However, we used several methods to mitigate measurement error when estimating nutrient intakes, including examining protein intake both as a continuous and a categorical variable, and adjusting all nutrient intake estimates for total energy intake.

In conclusion, in a cohort of pregnant women with relatively high protein intake, we observed an inverse association of maternal protein intake during the second trimester of pregnancy with cord blood concentrations of several hormones known to regulate fetal growth. Some associations were stronger in girls and babies born to parous mothers. Adequate nutrition and protein intake in particular are important for optimal fetal growth, yet our results suggest that high protein intake during pregnancy may downregulate growth-promoting hormones in utero.