Maternal protein intake is not associated with infant blood pressure

Abstract

Background

Animal data show that low protein intake in pregnancy programs higher offspring blood pressure, but similar data in humans are limited. We examined the associations of first and second trimester maternal protein intake with offspring blood pressure (BP) at the age of six months.

Methods

In a prospective US cohort study, called Project Viva, pregnant women completed validated semi-quantitative food-frequency questionnaires (FFQ) to measure gestational protein intake. Among 947 mother-offspring pairs with first trimester dietary data and 910 pairs with second trimester data, we measured systolic blood pressure (SBP) up to five times with an automated device in the offspring at the age of six months. Controlling for blood pressure measurement conditions, maternal and infant characteristics, we examined the effect of energy-adjusted maternal protein intake on infant SBP using multivariable mixed effects models.

Results

Mean daily second trimester maternal protein intake was 17.6% of energy (mean 2111 kcal/day). First trimester nutrient intakes were similar. Mean SBP at age 6 months was 90.0 mm Hg (SD 12.9). Consistent with prior reports, adjusted SBP was 1.94 mm Hg lower [95% confidence interval (CI) −3.45 to −0.42] for each kg increase in birth weight. However, we did not find an association between maternal protein intake and infant SBP. After adjusting for covariates, the effect estimates were 0.14 mm Hg (95% CI −0.12 to −0.40) for a 1% increase in energy from protein during the second trimester, and −0.01 mm Hg (95% CI −0.24 to −0.23) for a 1% increase in energy from protein in the first trimester.

Conclusions

Variation in maternal total protein intake during pregnancy does not appear to program offspring blood pressure.

Hypertension poses an enormous global medical and economic burden, with up to one billion people affected worldwide.^1–3 Despite the availability of effective antihypertensive therapy, suboptimal blood pressure (BP) control has been identified as responsible for 62% of cerebrovascular disease and 49% of ischemic heart disease.³ Effective population preventive measures for hypertension are urgently needed.

The observation that lower birth weight is associated with higher BP throughout childhood and adulthood^4,5 suggests a potential opportunity for prenatal prevention of hypertension. Barker^6,7 proposed that lower birth weight may be an indicator of fetal undernutrition during the second half of gestation, programming lifelong hypertension. However, simple energy restriction alone is unlikely to explain the inverse birth weight–BP association. Prenatal exposure to severe caloric restriction during the Dutch famine has not been associated with BP.⁸ Furthermore, the inverse birth weight–BP relationship has been demonstrated in developed nations, where calorie restriction is uncommon.^4,5,9,10

In a well-characterized rat model, a relatively mild restriction in maternal protein intake during pregnancy (~75% of the requirement in pregnancy) leads to both lower birth weight and higher offspring BP.^11,12 In humans, however few studies have examined whether protein intake during pregnancy affects offspring BP. Most studies are retrospective, and their results have been inconsistent.^13–15 These inconsistent results may reflect in part the inherent limitations of retrospective studies, including reliance on historical dietary records, incomplete data regarding confounding variables, and large attrition rates. Only one study used prospectively collected exposure and outcome data. In that study, a 1% increase in energy from protein during pregnancy was associated with a −0.2 mm Hg lower systolic blood pressure (SBP) in adolescent Filipino boys.¹⁶ This finding, if confirmed in other populations, could have important ramifications for nutritional counselling during pregnancy.

Our study aims were to examine the associations of first and second trimester maternal protein intake with offspring BP during infancy, using a prospective cohort study of pregnant women and their offspring (Project Viva).

Methods

Recruitment for Project Viva occurred between April 22, 1999 and July 31, 2002 at eight obstetric offices of Harvard Vanguard Medical Associates (HVMA), a multi-specialty, managed-care group practice in the Boston, MA area.¹⁷ Women with singleton pregnancies were eligible for the study if they entered prenatal care within the first 22 weeks of gestation, intended to continue their obstetric care at HVMA, and were able to answer questionnaires in English. Study protocols were approved by the Human Subjects Committees of Harvard Pilgrim Health Care, Brigham and Women’s Hospital, and Beth Israel Deaconess Medical Center, and all the participants gave informed consent.

A trained research assistant met each participant after her initial clinical prenatal visit and at 26–28 weeks gestation. At each visit, an assistant interviewed the participant, and gave her a series of questionnaires to complete and mail to the study office. Within 72 h after delivery, an assistant interviewed the mother at the hospital, and recorded the newborn infant’s measurements. When the infant reached 6 months of age, an assistant interviewed the mother, recorded maternal and infant measurements, and collected previously mailed self-administered questionnaires. Data was also collected from hospital and clinic medical records. The data collected included demographics (infant gender, maternal marital status, and race), socioeconomic status (education, household income), and reproductive history (maternal gravidity, pre-pregnancy height and body mass index, pregnancy weight gain, third trimester blood pressure, and smoking).

Assessment of nutrient intake

We assessed maternal diet using semi-quantitative food-frequency questionnaires (FFQ) validated for use during pregnancy.¹⁸ The 166-item FFQ was self-administered near the end of the first and second trimesters. We considered food-frequency questionnaires invalid if daily energy intake was <600 or >6000 kcal. We estimated nutrient intakes using the Harvard nutrient database, which contains food composition values from the US Department of Agriculture, supplemented by other data sources.¹⁹ We measured daily dietary protein, carbohydrate and fat both as absolute intakes (g/day) and as nutrient densities (% of daily energy). We used protein density (protein as % of daily energy) as our primary measure of protein intake because of its relevance to public health recommendations,²⁰ improved validity of measurement in food frequency questionnaires,²¹ and previous data suggesting that gestational dietary composition, rather than absolute nutrient intakes, influences offspring BP.^13,14,16 Prior studies in humans have suggested that maternal dietary protein intake in the second half of gestation may exert a larger effect on offspring BP than dietary protein intake in the first half of gestation.^13,14 Therefore, we defined our primary exposure variable as the % of daily energy from protein during the second trimester.

Blood pressure measurements

For each infant at the 6-month visit, one of eight trained research assistants measured BP up to five times, at 1-min. intervals, using a Dinamap 8100, Pro 100 or Pro 200 (Critikon Inc., Tampa, FL, USA) automated BP monitor. At the time of the measurements, we recorded the conditions of measurement; room temperature, activity state of the infant (crying, quiet sleep, active sleep, or quiet awake), cuff size (infant or newborn, child, or small adult), appendage used for BP measurement (left arm, right arm, or calf), and infant position (on the examination table, seated, or held). We defined our primary endpoint to be infant systolic BP (SBP) at 6 months of age. Systolic rather than diastolic BP was used because of the validity its measurement using an automated device and its superior predictive ability of future BP.²²

Anthropometric measures

Infant birth weight data were obtained from hospital records. At the 6-month visit, a research assistant weighed each infant and measured its length to the nearest 0.1 cm using a research standard Shorr infant length board (Shorr Productions, Olney, MD, USA). Gestational age at birth was computed as the number of days between the first day of the last menstrual period and the delivery date and confirmed by ultrasound fetal measurements at 16–20 weeks. For date discrepancies of more than 10 days, the ultrasound-based gestational age was used.

Study population

The study population has been described previously.¹⁷ Of the 2128 Viva participants who delivered a live-born infant, 70 women disenrolled and 38 infants were less than 6 months old as on the data analysis date (March 15, 2003), leaving 2020 mothers with offspring at least 6 months of age eligible for this analysis. Of the 2020 eligible mothers, 1701 women consented to enrol their infants for BP measurement at the age of 6 months. 1169 (69% of 1701) mothers and infants returned for BP measurement when the infant was 6 months old. In 96 infants, we were unable to obtain measurements due to infant fussiness (n = 17), research assistant or machine error (n = 22), parental refusal (n = 19) or other reasons (n = 38). In one infant, the SBP was measured but biologically implausible. Thus, we obtained BP measurements in 1072 infants. Of these infants, valid FFQs had been completed by 982 mothers during the first trimester, and by 940 mothers during the second trimester. For this analysis, we also excluded infants lacking length, weight, or birth weight data. Therefore, we included in our analyses 947 mother-offspring pairs with first trimester nutrient intakes and 910 mother-offspring pairs with second trimester nutrient intakes.

Statistical methods

To assess multivariable associations between protein intake and offspring BP, we used mixed effect regression models, incorporating each of the up to 5 BP measurements per infant as repeated outcome measures.²³ In contrast to the standard ordinary least squares analysis, mixed effects models weight subjects based on the number of measurements and their variability, and thus result in more appropriate standard errors.

To control for possible confounding by energy intake, we adjusted all models for daily energy intake using the multivariable nutrient density model.²⁰ Adjustment for energy intake also attenuates potential random misclassification in nutrient intake caused by differences in body size, physical activity and metabolic inefficiency.²⁰ Other potential confounders that did not change the estimated effect of protein by more than 10% were omitted from subsequent models. To reduce measurement error in infant BP, we included BP measurement conditions in all of our models, including the baseline model, Model 1. In Model 2, we added biological variables pertaining to the infant including age, gender, length and weight at 6 months, gestational age, and birth weight for gestational age, z-score.²⁴ For comparability with other reports, we also performed separate analyses using birth weight in place of its component variables (birth weight for gestational age and gestational age). In Model 3, we added to Model 2 adjustment for maternal biological and socioeconomic variables. Potential confounders we examined and excluded from our analyses included maternal age, education level, marital status, pre-pregnancy body mass index, maternal height, gravidity, pregnancy weight gain, smoking status prior to pregnancy, maternal third trimester BP, and room temperature.

Prior studies have suggested possible interactions between the effects of protein and other macronutrient intakes on offspring BP.^13,14 To assess the BP effect of substituting protein for gestational carbohydrate intake (while keeping energy constant), we created a multivariable substitution model, by entering the terms protein density, fat density, and energy into Model 3.²⁰ In this model, the regression coefficient for the protein density term represents the effect of using energy from protein to replace an equal percentage of energy from carbohydrate.²⁰ Likewise, we also assessed the effect of substituting protein energy for fat energy by entering protein density, carbohydrate density, and energy into Model 3. We conducted all data analyses using SAS version 8.2 (SAS Institute Inc., Cary, NC, USA).

Results

Participant nutrient intakes are shown in Table 1. The first trimester FFQ was completed at a mean gestational age of 11.7 weeks (SD 3.1), and the second at a mean of 29.0 weeks (SD 2.2). In the second trimester (our primary exposure), women derived between 11.7% and 25.3% (1st–99th percentile) of their energy from protein, with a mean of 17.6% energy from protein (92.2 g/day). Second trimester caloric intake ranged from 1002 to 3849 kcal/day (1st–99th percentile). The pattern of protein and other nutrient intakes in the first and second trimesters were similar.

Table 1.

Participant nutrient intakes during pregnancy

	First trimester	Second trimester
Nutrient	Mean (SD) (n = 947)	Mean (SD) (n = 910)
Protein (% of energy)	17.4 (2.9)	17.6 (2.8)
Protein (g/day)	88.8 (29.6)	92.2 (28.4)
Animal protein (g/day)	60.8 (23.9)	64.3 (22.9)
Animal protein (% of energy)	12.0 (3.3)	12.3 (3.1)
Carbohydrates (% of energy)	55.5 (7.0)	54.5 (6.8)
Fat (% of energy)	28.2 (5.3)	29.7 (5.3)
Total calories (kcal/day)	2044 (626)	2112 (614)

Compared with the cohort of 2020 eligible mothers, mothers included in our analyses were less likely to be of the non-white race/ethnicity (25% vs 34%) and less likely to have a low education level (26% vs 34% completed less than a college degree). Mean birth weight among infants included in our analyses was slightly higher than among eligible infants (3506 g vs 3463 g). Among eligible women with second trimester dietary data (n = 1611), mean protein intake was 17.5% of energy (93.7 g/day), similar to the 17.6% of energy (92.2 g/day) among women included in our analyses. Gestational weight gain, number of prior pregnancies, and maternal BMI were similar among included and eligible mothers.

There were no apparent patterns in birth weight or blood pressure by quartile of protein intake (% of energy) (Table 2). Results of multivariable models examining the effect of second trimester maternal protein intake on 6-month infant SBP are shown in Table 3. Controlling for energy and BP measurement conditions (Model 1), protein intake was not associated with SBP; the effect estimate was 0.11 mm Hg [95% confidence internal (CI) −0.16–0.38] per 1% increase in energy from protein. Adjustment for infant biological variables, maternal race, and income variables (Models 2 and 3) made no material difference to the effect estimate. Not only was there no apparent association between protein intake and infant BP, but the lower bound of the confidence interval excluded any important inverse association between maternal protein intake and infant BP, including the inverse association shown by Adair et al.¹⁶

Table 2.

Characteristics of 910 participants by quartile of second trimester protein intake (as % of energy)

	Quartile of protein intake (% of energy)
Characteristics	1	2	3	4
Second trimester nutrient intakes	Mean
Protein (% of energy)	14.2	16.6	18.3	21.1
Protein (g/day)	78.1	89.8	96.5	104.2
Animal protein (g/day)	49.1	60.9	68.2	78.7
Total calories (kcal/day)	2197	2159	2111	1979
Maternal characteristics
Age (yr)	31.2	32.9	32.8	32.8
Race (%)
White	68.0	80.0	80.7	70.5
Black	16.4	7.8	7.5	11.5
Hispanic	6.7	4.4	4.0	4.4
Asian	4.4	4.4	4.0	8.8
Other	4.4	3.5	4.0	4.9
Pre-pregnancy height (m)	1.7	1.7	1.7	1.6
Pre-pregnancy BMI (kg/m2)	24.5	24.5	24.2	24.3
Married (%)	79.1	90.4	90.4	89.0
Education (%)
High school	8.4	3.9	5.7	5.7
<4 yr of college	25.3	22.6	15.4	18.5
4 yr of college	35.6	40.0	33.8	37.9
Graduate degree	30.7	33.5	45.2	37.9
Income (%)
<$20 000	2.7	1.3	1.8	3.1
$20–<$40 000	11.6	7.4	4.8	7.1
$40–$70 000	25.3	20.0	19.3	22.0
>$70 000	51.1	67.0	70.6	63.0
Missing	9.3	4.4	3.5	4.9
Smoked 3 months prior to pregnancy (%)	14.2	10.1	7.2	8.2
Pregnancy weight gain (kg)	15.4	16.1	15.6	15.5
Third trimester SBP (mm Hg)	110.5	111.9	111.0	111.3
Infant characteristics at birth
Male (%)	49.8	48.7	48.7	52.4
Birth weight (g)	3461	3521	3559	3481
Birth weight for gestational age z-score (units)	0.13	0.21	0.30	0.18
Gestational age (weeks)	39.5	39.7	39.7	39.6
Infant characteristics at 6-month visit
Age (months)	6.4	6.5	6.5	6.6
Weight (kg)	8.1	8.0	8.1	8.1
Length (cm)	66.8	66.4	66.7	66.7
SBP (mm Hg)	89.5	89.5	90.6	90.2
Diastolic BP (mm Hg)	56.4	56.7	56.7	57.7

Table 3.

Multivariable mixed effects models showing increment in 6-month infant SBP (mm Hg) per 1% increase in energy from protein in second trimester (n = 910 mothers and their infants)

Modelsa	Change in 6-month infant SBP (mm Hg) per each 1% of energy from protein in the second trimester	SE	95% CI
Model 1: Protein (% of energy) + energy	0.11	0.14	−0.16–0.38
Model 2: Model 1 + birth weight for gestational age z-score, gestational age, infant age, gender, and 6-month length and weight	0.12	0.13	−0.14–0.38
Model 3: Model 2 + maternal race/ethnicity, income	0.14	0.13	−0.12–0.40

^aAll models were adjusted for blood pressure measurement order, cuff size, appendage, position, state, and machine model.

To compare our findings with other reports, we examined birth weight itself as a predictor of offspring BP. In Model 3, SBP was lower by 1.94 mm Hg (95% CI −3.45 to −0.42; P = 0.01) per 1 kg increment in birth weight.

Using second trimester absolute maternal protein intake as the exposure variable in Model 3, absolute protein intake was not associated with SBP (0.04 mm Hg per 1 g increase in daily protein intake, 95% CI −0.01–0.09). In multivariable substitution models, substituting protein for carbohydrate intake did not affect offspring SBP. The effect of replacing 1% of energy from carbohydrate with 1% of energy from protein was 0.15 mm Hg (95% CI −0.12–0.41). Similarly, substituting protein energy for energy from fat did not affect SBP (0.18 mm Hg per % energy substituted, 95% CI −0.14–0.50).

We found no difference between boys and girls in the association between second trimester protein intake and offspring SBP (data not shown). We also found no association between animal protein intake and SBP. Among women with intakes of less than 50 g animal protein per day (n = 246), the effect estimate was 0.32 mm Hg (95% CI −0.29–0.92) per 1% increase in energy from animal protein. Among women with intakes of at least 50 g of animal protein, the effect estimate was 0.15 mm Hg (95% CI −0.19–0.48) per 1% increase in energy from animal protein.

Results for diastolic BP were very similar to those for SBP. Model 3 estimated a 0.13 mm Hg increase in diastolic BP for each 1% increase in energy from protein (95% CI −0.07–0.32) during the second trimester.

We found no association between first trimester protein intake and infant BP (Table 4). Adjusted for other covariates, the effect estimates for systolic and diastolic BP were respectively −0.01 mm Hg (95% CI −0.24–0.23) and −0.07 mm Hg (95% CI −0.25–0.12) for each 1% increase in protein energy during the first trimester.

Table 4.

Multivariable mixed effects models showing increment in 6-month infant SBP (mm Hg) per 1% increase in energy from protein in first trimester (n = 947 mothers and their infants)

Modelsa	Change in 6-month infant SBP (mm Hg) per each 1% of energy from protein in the first trimester	SE	95% CI
Model 1: Protein (% of energy) + energy	−0.07	0.12	−0.31–0.17
Model 2: Model 1 + birth weight for gestational age z-score, gestational age, infant age, gender, and 6-month length and weight	−0.04	0.12	−0.28–0.20
Model 3: Model 2 + maternal race/ethnicity, income	−0.01	0.12	−0.24–0.23

^aAll models were adjusted for blood pressure measurement order, cuff size, appendage, position, state, and machine model.

Discussion

A link between higher adult BP and lower birth weights is now supported by epidemiologic data from the developed and developing world.^4,5,9,10 In many studies, birth weight has been interpreted as a proxy for maternal nutrition during pregnancy. However, birth weight is influenced by many factors in addition to maternal diet.²⁵ Rather than relying on birth weight as an indirect indicator, we assessed the direct impact of maternal diet composition on offspring BP.

To our knowledge, this is the first prospective study in a developed nation to investigate the effect of maternal diet during pregnancy on offspring BP. We found no association between maternal protein intake during pregnancy and offspring BP, but birth weight was inversely related to SBP. The magnitude of the effect, −1.94 mm Hg decrease in SBP per 1 kg increment in birth weight, was consistent with the summary effect estimate from a large systematic review.⁴ Together, these two findings suggest that the inverse birth weight—BP relationship cannot be explained by the variation in protein intake during pregnancy, at least among well-nourished women like those in our cohort.

Our findings diverge from previous studies in humans suggesting that dietary protein composition during pregnancy affects offspring BP. Shiell et al.¹⁵ reported that higher absolute gestational protein intake was associated with higher offspring BP. Campbell et al.¹³ reported a complex relationship: in women with mean animal protein intakes of less than 50 g per day, an increasing % of energy from animal protein was associated with lower offspring BP. However, in women who consumed more than 50 g animal protein per day, the trend was reversed.¹³ We did not find any association between animal protein density and offspring BP. Roseboom et al.¹⁴ reported that in adults exposed in utero to the Dutch famine, BP was inversely related to maternal protein–carbohydrate ratio. We used multivariable substitution models to assess the effect of replacing energy from carbohydrates with energy from protein, while keeping total energy constant. Substituting protein energy for energy from gestational carbohydrate intake had no effect on offspring BP. Because our models were adjusted for daily energy intake, they simulated isocaloric animal models that vary only dietary composition.

Our results are consistent with some but not all findings in the only other prospective study. Adair et al.¹⁶ found no effect of maternal protein intake on SBP in adolescent girls. However, in adolescent boys, Adair et al. found that for each 1% increment in energy from maternal protein intake, SBP was lower by −0.2 mm Hg. The lower limit of our effect estimate CI (−0.12 mm Hg) excluded this value. Thus, although we cannot completely rule out an inverse relationship between maternal protein intake and offspring BP, our results suggest that the magnitude of any such effect is unlikely to be clinically important.

One explanation of our null findings could be that an association with BP may be detectable only at levels of protein intake below a certain threshold. Within our cohort, mean maternal protein intake exceeded the minimum protein intake recommended by the World Health Organization (51 g/day for a 60 kg pregnant woman).²⁶ In both trimesters, protein intake was similar to gestational protein intakes in the US and Finland, as measured by 24-hour recall²⁷ and by other validated FFQ.^28,29 However, our mean second trimester protein intake of 92 g/day (18% of energy) was higher than in all studies that found an inverse relationship between maternal protein intake and offspring BP.^13,14,16 Campbell et al. reported a mean maternal protein intake of 73 g (12.2% of energy).¹³ In the study by Adair et al.,¹⁶ mean maternal protein intake was 48 g, with mean caloric intake of 1506 kcal. (Personal communication, L. Adair, 2003). Nevertheless, even among women who consumed <50 g of animal protein per day, the direction of the association was opposite to the inverse association hypothesized.

We found that neither the first nor the second trimester protein density was associated with offspring BP; it remains possible that programming effects of dietary protein depend upon timing of reduced protein intake during the third trimester. In rats, a low protein diet can program offspring hypertension regardless of when the diet is administered during pregnancy.^12,30 In humans, associations between maternal protein intake and offspring BP have been noted only for protein intake during late pregnancy (>20 weeks).^13,14,16 Our study did not measure third trimester protein intake. However, 45% of second trimester questionnaires were administered after 29 weeks of gestation was completed, incorporating protein intake during the early third trimester as well as the second trimester.

A major strength of this study was our careful assessment of BP and measurement conditions, resulting in increased precision of effect estimates. Other study strengths included the use of an FFQ validated in pregnancy, the detailed, prospectively recorded covariate data, and the relatively large cohort size. One of the limitations of this study was that our analyses included only half of the original 2020 eligible mother-infant pairs. Patterns of maternal protein intakes were similar among eligible and analyzed groups, and mean infant blood pressures were consistent with previous reports.^31,32 These findings suggest that participants excluded from our analyses were not systematically different in maternal protein intake or infant BP, rendering selection bias less likely. Generalizability of our study may be limited, since the women included in the study were predominantly of white race, with a relatively high level of education and income.

We cannot exclude the possibility that our outcome assessment occurred too early to detect the impact of gestational protein intake. BP levels from the age of six months have been shown to be associated with BP in later life, with reported tracking associations ranging from 0.2 to 0.7,^33–36 although not all studies are in agreement.³² In the few studies of infants <1 yr, an inverse relationship between birth weight and SBP has been found in a few, but not all studies.^37–41 We did demonstrate the expected inverse birth weight–SBP relationship, with a magnitude of effect consistent with prior reports in children and adults.⁴

In conclusion, we were unable to detect any association between maternal protein intake and offspring BP at the age of six months, even in the face of the expected inverse relationship between birth weight and SBP. Our work does not rule out nutritional programming effects of the diet, or even protein, on BP. Indeed, these relationships are probably much more complex than those suggested by simple protein-restriction animal models. Recently, investigators have suggested that the amino acid composition of the diet, rather than the absolute amounts of total protein, is responsible for programming effects on offspring BP.^42,43 Certain non-essential amino acids, such as glycine, may be conditionally required for normal cardiovascular development.⁴⁴ Further studies are needed to understand how the pattern of maternal amino acid consumption might interact with fetal and placental metabolic pathways to permanently alter offspring blood pressure.^25,45

KEY MESSAGES.

• This is the first prospective study in a developed nation to investigate the association of maternal protein intake during pregnancy with offspring blood pressure.

• Consistent with prior reports, birth weight was inversely related to SBP.

• Maternal protein intake during the first and second trimesters of pregnancy was not associated with infant blood pressure at the age of 6 months.