- Top of page
- Supporting Information
Embryonic and fetal growth and placentation are determined by several pathways and molecular biological processes, in which multiple genetic and environmental factors interact. The first trimester is a sensitive period of pregnancy. Despite associations between adverse exposures during early pregnancy and the occurrence of pregnancy complications and adverse birth outcomes, this period is largely neglected in antenatal care.[1-4] Ultrasound measurements in the first trimester are mainly used to estimate crown–rump length (CRL) for an accurate determination of gestational age. Discrepancies between menstrual and ultrasound estimations usually lead to gestational age adjustment as it is interpreted as earlier/delayed ovulation or implantation. These discrepancies can also be the result of differences in embryonic growth. Chromosomal abnormalities can affect embryonic growth and it has been shown that maternal characteristics, such as age, smoking and the use of a folic acid supplement may also affect embryonic growth.
Another important exposure during pregnancy is nutrition, interest in which has increased in the last decade. The Mediterranean dietary pattern (high intakes of vegetables, fruits, nuts, fish, olive oil and moderate intakes of alcohol) has been associated with better embryo quality, increased fetal growth and reduced risk of adverse birth outcomes.[8-11] Important characteristics of this pattern are its high content of methyl donors for the one-carbon pathway. This pathway plays a vital role in biological processes implicated in growth and programming, especially periconceptional. So far no information is available on the influence of the quality and the quantity of specific periconceptional maternal dietary patterns on first-trimester growth, and on the biomarkers of the one-carbon pathway. Therefore, in a population-based prospective birth cohort, we aimed to study (1) first-trimester maternal dietary patterns; (2) the influence of the dietary patterns on the biomarkers of the one-carbon pathway; and (3) associations with growth in each trimester of pregnancy and birthweight.
- Top of page
- Supporting Information
This study was embedded in the Generation R Study, a population-based prospective cohort from early pregnancy onwards (Rotterdam, the Netherlands) and approved by the Medical Ethics Committee of the Erasmus Medical Centre, Rotterdam, the Netherlands. Written informed consent was obtained from all mothers for both maternal and child data. Eligible women were those who were resident in the study area and who delivered between April 2002 and January 2006.
As dietary habits are culturally determined, only Dutch women were selected for this study. Furthermore, the Food Frequency questionnaire that was used was not validated in non-Caucasian women. A total of 2243 prenatally enrolled Dutch women with a singleton pregnancy and a CRL measurement were included (Figure 1). Only women with a CRL measurement within the recommended gestational age range (10+0 weeks to 13+6 weeks) were selected. Mothers with an unknown first day of the last menstrual period (LMP) or mothers with no regular menstrual cycle of 28 ± 4 days were excluded (n = 1179) to minimise confounding of gestational age. The LMP was obtained from the referring letter and confirmed at enrolment. Additional information on regularity and cycle duration was obtained. Of the remaining 1064 women, 961 filled out a food frequency questionnaire (FFQ). If mothers participated in Generation R with two or more pregnancies, one sibling was randomly selected to avoid bias due to paired data (n = 87). After exclusion of women who underwent any form of fertility treatment (n = 10) or who reported drug abuse (n = 17), 847 women were eligible for present study.
Standard ultrasound planes and intraclass correlation coefficients for first-trimester measurements were used as described previously.[14, 15] Fetal growth measurements were performed using standardised ultrasound procedures in second trimester (median 20.3 weeks, 90% range 19.1–22.0) and third trimester (median 30.1 weeks, 90% range 29.0–31.9). Estimated fetal weight (EFW) was calculated using Hadlock's formula. Ultrasound examination was performed using an Aloka model SSD-1700 (Tokyo, Japan) or the ATL-Philips Model HDI 5000 (Seattle, WA, USA). Gestational age-adjusted standard deviation (SD) scores, based on reference growth curves from the entire study population, were constructed for CRL and EFW.[1, 14] Information on date of birth; birth anthropometrics and gender of the child were obtained from community midwife and hospital registries. Gestational age and sex-adjusted SD scores for birthweight were constructed using standards from Usher and McLean.
At enrolment (median 12.3 weeks, 90% range 11.0–13.7), nutritional intake of the previous 3 months was assessed using a modified, validated semiquantitative FFQ, consisting of 293 food items and structured according to meal patterns. Questions included consumption frequency, portion size, preparation method and food additions. Portion sizes were estimated using household measures and photographs. To calculate average daily nutritional values the Dutch food composition table 2006 was used. The original food items were reduced to 20 predefined food groups (listed in Table 1) based on origin, culinary usage and nutrient profiles. This was followed by principal component factor analysis (PCA) applied for energy intake unadjusted food groups of the women to construct overall dietary patterns. The PCA aims to explain the largest proportion of variation in food groups in terms of a few linear functions called principal components. A dietary pattern consists of a selection of the initial food groups, each with its own coefficient, defining the observed correlation of the food group with the dietary pattern. The three most prevalent dietary patterns with the highest explained variance were selected. A woman was given scores for each of the dietary patterns. According to their personal score the women were stratified into three equally sized groups and labelled as having a low; intermediate or high adherence to the dietary pattern. The dietary patterns were analysed in a continuous and categorical measure. This approach was chosen to explore both linearity and nonlinearity of the association.
Table 1. Relationships between food groups and the identified dietary patterns
|Food group||Medianb||90% range||Mediterranean dietary pattern (12.3%)||Energy-rich dietary pattern (9.9%)||Western dietary pattern (7.6%)|
|Bread, breakfast cereals||145.4||52.8–242.9||0.06a||0.80a||0.09a|
|Condiments, starches, sauces||29.5||9.8–63.7||−0.01||0.00||−0.04a|
|Nonsweetened, nonalcoholic beverages||1107.1||259.3–2142.9||0.26a||0.28a||0.04a|
|Sweetened, nonalcoholic beverages||322.5||31.9–1215.8||−0.01||−0.24a||0.09a|
Venous blood serum and plasma samples were drawn and stored at −80°C and at the same time the CRL was measured and the FFQ was handed out. Biomarkers of the one-carbon pathway, folate and total homocysteine in plasma and vitamin B12 in serum, were assessed. Between-run coefficients of variation for the biomarkers are listed in Table S1. All three biomarkers were available in 81.3% of the women (n = 689/847). No significant difference in nutritional intake was observed between women with and without all three biomarker concentrations available (P = 0.54).
The analyses were performed using multiple linear regression models with the identified dietary pattern as regressor (as a continuous and categorical measure) and CRL (SD score) as response of outcome. This model was compared with a different approach to extract dietary patterns, namely multiple linear regression analyses where each food group was included separately as a continuous measure to predict CRL. In additional analyses, multiple linear regression analyses were performed to assess associations between the dietary pattern and SD scores of EFW in second/third trimester and birthweight. The CRL SD score was added to the latter model to assess whether the effect on CRL mediates the effect on EFW and birthweight.
The final multiple regression analyses take into account potential confounders, which were selected on previous literature and determined a priori. From self-administered questionnaires, data were available on maternal age, education, marital status, parity, smoking, folic acid supplement use, nausea/vomiting/fever during first trimester, comorbidity (chronic hypertension and/or heart disease and/or diabetes and/or high cholesterol and/or thyroid disease and/or systemic lupus erythematosus and/or multiple sclerosis), repeated miscarriages (two or more miscarriages), sexually transmitted diseases and the season in which the FFQ was filled out. Education was assessed by the highest completed education and classified as (1) low (none/primary); (2) medium (secondary); (3) high (college/university). Parity was classified as (1) nulliparous and (2) multiparous. Maternal smoking was assessed in each trimester. Women who reported any or no smoking during pregnancy were respectively classified as ‘smokers’ and ‘non-smokers’. Folic acid supplement use was categorised into (1) folic acid supplement use (preconceptional or postconceptional start); (2) no folic acid supplement use. At enrolment maternal weight and height were measured to calculate body mass index (BMI, kg/m2). Information on fertility treatment was obtained from community midwives and obstetricians. Blood pressure was measured using the validated Omron 907 automated digital oscillometric sphygmomanometer (OMRON Healthcare Europe B.V., Hoofddorp, the Netherlands). The mean value of two blood pressure readings over a 60-second interval was documented. Covariates were selected as confounders by testing whether the potential confounder changed the dietary pattern as a continuous measure with at least 10% in the exploratory analyses. By using this approach, maternal age, maternal BMI, maternal folic acid supplement use, duration of the last menstrual cycle and fetal gender were included in the final analyses. Maternal and paternal height, paternal BMI, maternal educational level, parity, maternal smoking, mean diastolic blood pressure and mean systolic blood pressure at intake were included by default. All variables were added to the model simultaneously. We have tested for potential interactions with the dietary pattern. No significant interaction term was observed in our study population. Therefore, no interaction terms were added to the model.
Missing data on maternal BMI (0.4%), paternal height (9.3%), paternal BMI (9.7%), maternal educational level (0.5%), parity (0.1%), maternal smoking (7.0%), the use of folic acid supplement (14.9%) and maternal diastolic (0.7%) and systolic (0.7%) blood pressure were completed on the Dutch population with a CRL measurement (n = 2243) using the Markov-Chain–Monte-Carlo multiple imputation technique. Details of the multiple imputation model are provided in the Table S2. For all analyses, results including imputed missing data are presented.
All analyses were performed using SPSS software, version 17.0 (SPSS Inc., Chicago, IL, USA).
- Top of page
- Supporting Information
Three major factor were identified in 847 women of which the correlation coefficients are shown in Table 1. These three principal components explained the highest percentage of variance of all the principal components and together, they explained 29.8% of the variance. The first component was labelled the Mediterranean dietary pattern, explaining 12.3% of the variance of dietary intake of the total study group. It comprised high intakes of vegetables, legumes, pasta/rice, dairy, fish/shellfish, vegetable oils, alcohol, nonsweetened nonalcoholic beverages and low intakes of processed meat (all r ≥ 0.20 and P < 0.05). The second component was labelled the energy-rich dietary pattern, explaining 9.9% of the total variance and comprising high intakes of bread/breakfast cereals, margarine, nuts, snacks/sweets and nonsweetened nonalcoholic beverages and low intakes of sweetened, nonalcoholic beverages (all r ≥ 0.20 and P < 0.05). The third component was labelled the Western dietary pattern, explaining 7.6% of the total variance and comprising high intakes of potatoes, pasta/rice, dairy, fresh meat, processed meat, margarine and alcohol and low intakes of nuts, fish/shellfish (all r ≥ 0.20 and P < 0.05). The Mediterranean and Western dietary patterns showed no significant association with CRL. For that reason we continue by describing the analysis of the energy-rich dietary pattern.
Maternal and fetal characteristics are shown in Table 2. Women with high adherence to the energy-rich dietary pattern were older. Furthermore, differences were observed between the adherence categories and BMI, household income, smoking habits and the use of a folic acid supplement. Nonresponse analyses showed that excluded mothers were younger, lower educated, had a lower income, were less frequently folic acid supplement users and gave birth earlier. Maternal and fetal characteristics for the Mediterranean and Western dietary patterns are shown in Table S3.
Table 2. Maternal and fetal characteristics
| ||Excluded Dutch women with CRL measurement||Study group||P valuea||Energy-rich dietary pattern||P valueb|
|Low adherence||Intermediate adherence||High adherence|
|n = 1396||n = 847||n = 283||n = 282||n = 282|
| Maternal characteristics |
|Age (years), mean (SD)||30.8 (4.2)||31.7 (4.0)||<0.001||31.1 (4.1)||31.8 (4.1)||32.2 (3.7)||0.004|
|Maternal BMI, median (90% range)||23.5 (19.0–30.7)||23.1 (19.6–32.3)||NS||23.3 (19.7–32.5)||23.5 (19.3–32.2)||22.8 (19.5–28.1)||NS|
|20–24.9 (%)||58.5||58.9|| ||61.2||52.7||63.5|| |
|25–29.9 (%)||20.9||23.5|| ||21.4||28.5||20.9|| |
|30–34.9 (%)||6.2||5.9|| ||8.5||4.3||5.0|| |
|>35 (%)||3.7||2.8|| ||2.1||3.2||2.1|| |
|Education (%)||0.001|| || || ||NS|
|Low||4.1||2.3|| ||3.2||1.1||2.5|| |
|Household income (%)||0.016|| || || ||0.045|
|<1200 €||5.3||3.3|| ||5.0||2.7||2.4|| |
|Parity (%)||NS|| || || ||NS|
|0||59.3||62.1|| ||66.1||60.1||59.9|| |
|Smoking (%)||NS|| || || ||0.001|
|Yes||16.2||12.6|| ||18.7||8.7||10.3|| |
|Until pregnancy recognition||10.4||10.4||13.0||8.0||10.2|
|Folic acid supplement use (%)||0.017|| || || ||0.04|
|Adequate||57.8||63.4|| ||56.5||68.8||64.6|| |
|Daily nausea (%)||32.0||28.2||NS||26.7||30.0||27.8||NS|
|Daily vomiting (%)||5.3||3.6||NS||4.2||3.4||3.1||NS|
| Fetal characteristics |
|Male gender (%)||50.3||49.6||NS||50.4||48.2||50.4||NS|
|Gestational age at birth (weeks), median (90% range)||39.9 (36.3–42.1)||40.1 (36.7–42.0)||0.031||40.0 (35.8–42.1)||40.3 (37.3–42.0)||40.1 (36.9–42.1)||NS|
|Small for gestational age (<5th birthcentile) (%)||NA||4.5||NA||6.2||3.8||3.5||NS|
The biomarker concentrations and nutrient intakes are shown in Table 3. Lower homocysteine levels were observed in women with a high adherence to the energy-rich dietary pattern. High adherence to the energy-rich dietary pattern was associated with a higher energy intake. These women showed a relatively high total fat intake, but also a high protein and carbohydrate intake. Most energy was derived from fats and less from carbohydrates and proteins. Furthermore, higher intakes of saturated fats, monounsaturated and polyunsaturated fats, vegetable proteins, linoleic acid and fibres were observed. The biomarker concentrations and nutrient intakes for the Mediterranean and Western dietary patterns are shown in Table S4.
Table 3. Biomarker concentrations and nutrient intakes
| ||Total (n = 847)||Energy-rich dietary pattern||Linear trend analyses|
|Low adherence (n = 283)||Intermediate adherence (n = 282)||High adherence (n = 282)|
|ρ||Median (90% range)||Median (90% range)||Median (90% range)||Median (90% range)|
| Biomarker concentrations |
|Folate (nmol/l), plasma||0.07||20.5 (7.65–35.2)||20.7 (6.7–33.5)||20.3 (8.7–35.2)||21.0 (8.3–36.5)||NS|
|Folic acid supplement use||0.09||24.2 (11.1–37.0)||24.6 (8.5–37.4)||23.0 (11.1–35.3)||24.6 (12.5–37.4)||NS|
|No supplement use||−0.06||16.3 (6.1–30.7)||17.0 (5.9–32.7)||16.4 (5.9–29.1)||15.0 (6.5–30.0)||NS|
|Total homocysteine (μmol/l), plasma||−0.15||7.0 (5.0–10.2)||7.2 (4.9–10.5)||7.0 (5.0–10.6)||6.7 (5.0–9.7)||0.006|
|Folic acid supplement use||−0.13||7.3 (5.0–9.4)||7.0 (4.9–10.0)||6.8 (4.9–9.6)||6.6 (5.0–9.0)||NS|
|No supplement use||−0.08||6.8 (5.4–12.1)||7.4 (4.9–12.1)||7.4 (5.4–13.4)||7.1 (5.3–11.6)||NS|
|Vitamin B12 (pmol/l), serum||−0.04||174.0 (90.0–368.0)||191.0 (90.1–349.7)||171.0 (89.1–334.9)||174.5 (90.5–401.8)||NS|
|Folic acid supplement use||0.04||175.0 (90.2–379.8)||189.0 (86.5–327.6)||171.0 (89.0–427.4)||187.0 (100.2–396.0)||NS|
|No supplement use||−0.22||173.0 (88.6–380.2)||206.0 (91.1–435.7)||171.0 (88.4–358.6)||161.0 (78.8–441.2)||NS|
| Energy/macronutrients |
|Energy (KJ/day)||0.58||8847 (5564–12726)||7171 (4772–10978)||8771 (6370–11685)||10446 (7373–13473)||<0.001|
|Fat (% of energy)||0.26||36.3 (27.3–44.3)||34.6 (25.3–43.6)||35.7 (27.7–42.8)||38.3 (29.4–46.1)||<0.001|
|Total fat (g/day)a||0.65||83.3 (51.0–126.9)||35.7 (39.9–99.7)||82.8 (55.4–116.4)||104.8 (70.0–139.3)||<0.001|
|Saturated fats (g/day)a||0.50||30.6 (18.4–48.5)||25.1 (15.0–41.8)||30.3 (19.9–45.1)||37.1 (23.8–55.1)||NS|
|Monounsaturated fats (g/day)a||0.58||30.2 (17.5–46.6)||24.5 (14.4–37.5)||29.6 (19.4–43.4)||37.1 (25.6–50.7)||NS|
|Polyunsaturated fats (g/day)a||0.72||19.1 (9.6–32.7)||13.6 (8.6–21.9)||19.0 (11.5–27.6)||26.6 (17.6–39.5)||<0.001|
|Linoleic acid (g/day)a||0.73||15.5 (7.2–27.0)||10.7 (6.3–18.0)||15.3 (8.6–22.5)||22.1 (13.9–33.2)||<0.001|
|Cholesterol (mg/day)a||0.17||170.9 (91.6–279.1)||159.2 (84.1–273.7)||171.2 (99.7–284.8)||181.1 (103.4–272.9)||<0.001|
|Protein (% of energy)||−0.15||14.9 (11.4–19.3)||15.2 (11.3–20.2)||15.0 (11.7–19.1)||14.4 (11.3–17.6)||<0.001|
|Total protein (g/day)a||0.47||79.1 (47.9–112.0)||66.8 (40.7–96.3)||79.1 (57.0–106.1)||88.4 (63.3–119.2)||NS|
|Vegetable protein (g/day)a||0.76||30.4 (16.5–46.6)||22.7 (14.5–32.5)||30.7 (22.2–40.3)||38.7 (28.6–52.6)||<0.001|
|Animal protein (g/day) a||0.17||47.4 (25.9–73.2)||44.5 (23.7–68.4)||47.9 (28.4–72.4)||49.4 (28.6–78.2)||<0.001|
|Carbohydrate (% of energy)||−0.16||48.3 (39.2–59.1)||49.7 (38.5–61.3)||48.5 (40.7–58.1)||47.0 (38.5–57.7)||0.001|
|Total carbohydrate (g/day)a||0.42||254.7 (149.6–397.6)||212.5 (115.4–359.9)||251.1 (170.3–381.0)||293.0 (192.6–412.8)||<0.001|
|Carbohydrates mono- and disaccharides (g/day)a||0.19||139.8 (69.6–242.7)||124.7 (60.6–238.8)||137.6 (70.2–244.0)||154.2 (81.6–250.2)||<0.001|
|Carbohydrate polymers (g/day)a||0.63||113.5 (63.2–172.4)||84.3 (50.6–133.5)||116.6 (82.6–154.7)||135.2 (98.7–195.3)||<0.001|
|Fibres (g/day)a||0.62||23.3 (13.3–35.6)||17.8 (10.4–28.1)||23.5 (15.7–32.5)||28.3 (21.0–40.8)||<0.001|
The associations between maternal adherence to the energy-rich dietary pattern and CRL are presented in Table 4. In the univariate analyses, positive associations between CRL and high adherence (difference, mm: 2.15, 95% confidence interval [95% CI] 0.79–3.50 and SD 0.28, 95% CI 0.10–0.47) to the energy-rich dietary pattern were observed. In the multivariable analyses these associations remained significant among women with high adherence (difference, mm: 1.62, 95% CI 0.52–2.72 and SD 0.23, 95% CI 0.08–0.38) to the energy-rich dietary pattern. Additional adjustment for energy resulted in significant associations among women with high adherence (difference, mm 1.84, 95% CI 0.49–3.18 and SD 0.25, 95% CI 0.06–0.43) to the energy-rich dietary pattern. These associations remained after multiple testing adjustments (three independent factors). Thereafter, we restricted the analyses to women with a gestational age based on LMP within 7 days of the gestational age based on CRL (n = 785). The results attenuated, but a significant association between the high adherence group and CRL remained in the univariate (difference in mm: 1.48, 95% CI 0.33–2.63 and SD 0.21, 95% CI 0.05–0.38) and multivariate analyses (difference in mm: 1.04, 95% CI 0.10–1.99 and SD 0.15, 95% CI 0.01–0.29). Finally, we compared the analyses, performed with PCA, with a multiple linear regression model where the food groups were included continuously to predict CRL. This model showed an association between CRL and the intake of fresh meat (difference in mm: 0.023, 95% CI 0.002–0.044 and difference in SD 0.003, 95% CI 0.000–0.006).
Table 4. Associations between the degree of adherence to the energy-rich dietary pattern and CRL
|Energy-rich dietary pattern||CRL in mm, effect size (95% CI)||CRL in SD score, effect size (95% CI)|
|Intermediate adherence||1.17 (−0.0 to 2.34)||0.87 (−0.23 to 1.96)||0.13 (−0.03 to 0.29)||0.10 (−0.05 to 0.25)|
|High adherence||2.15 (0.79 to 3.50)b||1.62 (0.52 to 2.72)b||0.28 (0.10 to 0.47)a||0.23 (0.08 to 0.38)b|
|Linear trend analysesc||0.64 (0.18 to 1.09)b||0.56 (0.11 to 1.01)a||0.09 (0.02 to 0.15)a||0.08 (0.01 to 0.14)a|
The associations between maternal adherence to the energy-rich dietary pattern and EFW and birthweight are shown in Table 5. The energy-rich dietary pattern was not associated with EFW in second or third trimester or with birthweight. After addition of the SD score for CRL to the multivariable model, the effect estimates for the SD scores for EFW in the second and third trimester and birthweight were attenuated.
Table 5. Associations between the degree of adherence to the energy-rich dietary pattern and EFW and birthweight
|Energy-rich dietary pattern||EFW second trimester||EFW third trimester||Birthweight|
|Effect size (SD score), 95% CI||Effect size (SD score), 95% CI||Effect size (SD score), 95% CI|
| Model 1 |
|Intermediate adherence||0.04 (−0.12 to 0.20)||0.15 (−0.02 to 0.32)||0.05 (−0.13 to 0.23)|
|High adherence||0.14 (−0.02 to 0.30)||0.14 (−0.03 to 0.31)||0.15 (−0.03 to 0.33)|
|Linear trend analyses||0.05 (−0.02 to 0.12)||0.06 (−0.01 to 0.12)||0.04 (−0.04 to 0.11)|
| Model 2 |
|Intermediate adherence||−0.01 (−0.15 to 0.12)||0.11 (−0.04 to 0.26)||0.04 (−0.14 to 0.22)|
|High adherence||0.01 (−0.12 to 0.15)||0.03 (−0.12 to 0.18)||0.10 (−0.08 to 0.28)|
|Linear trend analyses||0.01 (−0.05 to 0.06)||0.02 (−0.04 to 0.08)||0.02 (−0.05 to 0.09)|
- Top of page
- Supporting Information
In 847 Dutch women derived from a multiethnic population-based birth cohort, increasing maternal adherence to an energy-rich dietary pattern in the periconceptional period is significantly associated with CRL. These associations are independent of several covariates.
The identified dietary pattern is determined by higher intakes of fats (especially unsaturated fats), but also by higher intakes of carbohydrates and proteins and is therefore labelled ‘energy-rich’. It consists of higher intakes of bread and nuts, containing carbohydrates and vitamin B6. Nuts also contain high amounts of methionine. In the one-carbon pathway methionine is an important substrate. Vitamin B6 is a cofactor. In addition, methionine, choline, folate and vitamin B12 are essential one-carbon donors for this pathway. Deficiencies of cofactors and substrates result in hyperhomocysteinaemia and excessive release of reactive oxidative species. In 50%, homocysteine is trans-sulphurated, a vitamin B6-dependent conversion, which is stimulated by high methionine intake to prevent hyperhomocysteinaemia. Therefore, the observed lower plasma concentrations of total homocysteine in women with high adherence to the energy-rich dietary pattern is in accordance with high intakes of methionine and vitamin B6.
A study in the same cohort comprising 1631 multiethnic women showed that maternal age, diastolic blood pressure, higher haematocrit levels, smoking and folic acid supplement use were associated with CRL at the end of the first trimester. In agreement with our study, they found no effects of maternal weight, height or BMI on CRL in this study. The effects of nutrition found in the present study are small, but comparable in effect size to these associations. The association between the energy-rich dietary pattern and CRL may be attributable to the first-trimester nourishment of the embryo by mainly energy-rich carbohydrate secretions from the endometrial glands. Another explanation might be that fetal growth is being programmed in the first trimester. Moreover, up to 11 weeks of gestation the embryo develops in a stable nutritional environment. This may explain why the effects are only found on CRL and seem to disappear in the second trimester. Unfortunately, we did not observe an effect of adherence to the energy-rich dietary pattern and fetal growth from the second trimester onwards. Therefore, the clinical implications remain unclear. However, first-trimester growth restriction has been linked to an increased risk of adverse birth outcomes and growth acceleration in early child hood in the same cohort. The lack of association of nutrition with growth from the second trimester onwards may be a result of our small study group. This is substantiated by the association of the dietary pattern with lower homocysteine levels. It has to be elucidated whether increased growth in the first trimester is a beneficial predictor for subsequent pregnancy course and outcome.
Although maternal nutrition in pregnancy has been studied extensively, this is the first study showing associations between maternal nutrition and early growth. Godfrey et al. showed an inverse association between the energy intake in early pregnancy and placental weight (difference in grams: 38, 95% CI 5–72) and birthweight (difference in grams: 134, 95% CI 11–256). Maternal use of a Traditional dietary pattern in early pregnancy (high intakes of potatoes, meat, vegetables) reduced the risk of having a small-for-gestational-age (SGA) child by 14% (95% CI 0.75–0.99). In contrast, Knudsen et al. showed that maternal adherence to a Western dietary pattern, based on red and processed meat and high-fat dairy, was associated with an increased risk of SGA (odds ratio 0.74, 95% CI 0.64–0.86 for women in the Health Conscious class compared with women in the Western Diet class). The Dutch Famine study demonstrated that exposure to famine during early pregnancy resulted in a risk for cardiovascular disease in later life, but independent of birthweight. Whether our finding is beneficial for pregnancy course and outcome warrants further investigation.
Previous research within Generation R revealed that low adherence to a Mediterranean dietary pattern was associated with an increased risk of SGA (relative risk 2.78, 95% CI 1.64–4.76). The Mediterranean dietary pattern identified in this study, however, was not associated with CRL. This difference between our study and that by Timmermans et al. can have multiple causes. First, we used a different method to define the dietary patterns. Second, we have a smaller study population. The analyses were performed on this smaller group and therefore our finding might be explained by this selection. Third, we defined 20 food groups, whereas Timmermans et al. defined 21 food groups. Last, we were also interested in the quantity of the used food groups. Therefore, we did not adjust the food groups for energy before we defined the dietary patterns. Inadequate dietary intake can indicate overnutrition and undernutrition as well as vitamin and mineral deficiency. Our study implies that both the quality and quantity of the dietary pattern matters in the first trimester of pregnancy. It must be noted that the Mediterranean diet is currently considered to constitute a healthier diet.
Some strengths and limitations have to be addressed. This study was embedded in a large cohort from whom a selection of Dutch women was studied. Detailed information on food consumption and potential confounders was prospectively collected. Selective participation in this study did occur, because mothers of lower socio-economic status were less represented in our study. This selection towards a healthy study population may have been harmful to the internal validity of our study, especially when the associations of the dietary pattern with CRL differ between the study population and the excluded mothers. This is difficult to ascertain because we do not know these associations, but they have to be taken into consideration.
The timing of pregnancy to assess gestational age at the moment of CRL measurement is very important in the analysis and interpretation of the results. We were not able to assess in this population-based cohort the exact timing of ovulation. However, to minimise misclassification of gestational age, we have restricted the analyses to women with a reliable first day of the LMP and a regular period of 28 ± 4 days. Moreover, we also adjusted the analyses for the duration of the menstrual cycle. We are aware however that residual confounding by gestational age cannot be excluded completely. The duration of the menstrual cycle can be confounded by recall bias and other maternal characteristics, such as age and smoking. Therefore, we repeated the analyses with a selection of women with a gestational age based on LMP with a 7-day range of the gestational age based on CRL (785/847) shown in Table S5. The association of women with a high adherence to the energy-rich dietary pattern and CRL remained significant. However, the effect attenuated and the linear trends were no longer significant.
A PCA was used for the dietary pattern analyses, which does not take into account previous knowledge. This approach has the advantage over a hypothesis-oriented approach that it takes into account the correlation structure of the food groups and does not focus on selected aspects of a diet. The amount of variance explained by the dietary patterns is small but is comparable with previous studies in pregnant women.[33, 34] Furthermore, the variance explained by dietary patterns depends heavily on the number of food groups used in the PCA analyses. We used 20 predefined food groups, which allows more variance in the model than if we had used fewer food groups. However, it should be noted that PCA assumes approximate normality of the data, but can still produce a good projection of the data when it is not normally distributed. Balder et al. examined the influence of analytical decisions on the stability of the dietary pattern in four European cohort studies. Sensitivity analyses were conducted with dichotomisation of extremely skewed variables, but this transformation did not affect the food groups with significant factor loadings, the magnitude of the factor loadings or the explained variance, and so the order of the extracted patterns. Accordingly we also dichotomised variables with a high percentage (>75%) of nonusers and found no difference in results. This supports the maintenance of the PCA using the continuous data.
Nutritional studies are prone to some bias, including imprecise measurement. Maternal dietary intake was assessed once at intake using a modified version of the validated semiquantitative FFQ of Klipstein-Grobusch et al. This FFQ was validated in an older white population. Several studies compared the results of dietary pattern analyses with the use of an FFQ, prospective diaries and weighted dietary records and observed no differences.[33, 35] Determination of dietary intake in our study was carried out at the end of the periconceptional period, covering the previous 3 months. For EFW and birthweight, it could be argued that diet can change throughout pregnancy. This may also explain the lack of association of our dietary pattern and growth from the second trimester onwards. Previous research, however, revealed no differences in nutritional intake during different time periods.[37, 38] Furthermore, underreporting of nutritional intake could occur in women with high BMI. We observed differences between the adherence categories and the different BMI groups. We tested underreporting by estimating the mean basal metabolic rate using the new Oxford equation for women aged 30–60 years: basal metabolic rate (mJoule/day) = 0.0407*weight (kg) + 2.9. The physical activity level was then calculated: mean reported energy intake/mean basal metabolic rate. To evaluate underreporting, a cut-off of 1.35 was used. In the total study group, the PAL was 1.56. In women with a BMI > 25, the physical activity level was 1.34, suggesting underreporting. Restricting analyses to women with normal BMI did not change the effect estimates. Therefore analyses on the entire study population are displayed.
Finally, to minimise multiple testing, we applied PCA to reduce the information on nutrition in principal components leading to fewer parameters in our regression model. By taking all nutrients into account, we limited the possibility of selective testing. Moreover, associations between the dietary pattern and biomarker concentrations of the one-carbon pathway were examined, to validate the dietary patterns and minimise the possibility of chance-finding. In addition, the associations of women with a high adherence to the energy-rich dietary pattern with CRL remain after multiple testing adjustment.