Dietary vitamin A intake below the recommended daily intake during pregnancy and the risk of congenital diaphragmatic hernia in the offspring §

Authors

  • Leonardus W. J. E. Beurskens,

    1. Department of Pediatric Surgery, Sophia Children's Hospital
    Current affiliation:
    1. Department of General Practice, Maastricht University Medical Center, Maastricht, The Netherlands
    Search for more papers by this author
  • Lieske H. Schrijver,

    1. Department of Obstetrics and Gynecology, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    Search for more papers by this author
  • Dick Tibboel,

    1. Department of Pediatric Surgery, Sophia Children's Hospital
    Search for more papers by this author
  • Mark F. Wildhagen,

    1. Department of Obstetrics and Gynecology, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    2. Department of Urology, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    Search for more papers by this author
  • Maarten F. C. M. Knapen,

    1. Department of Obstetrics and Gynecology, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    Search for more papers by this author
  • Jan Lindemans,

    1. Department of Clinical Chemistry, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    Search for more papers by this author
  • Jeanne de Vries,

    1. Department of Nutrition, Wageningen University, The Netherlands
    Search for more papers by this author
  • Régine P. M. Steegers-Theunissen

    Corresponding author
    1. Department of Obstetrics and Gynecology, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    2. Department of Clinical Genetics, Erasmus MC - University Medical Center Rotterdam, The Netherlands
    • Erasmus Medical Centre – University Medical Center Rotterdam, Department of Obstetrics and Gynecology, Building Ee Room 22-71a, PO Box 2040, 3000 CA Rotterdam, The Netherlands
    Search for more papers by this author

  • Disclosure of interest: The authors declare that they have no conflicts of interest.

  • Contribution to authorship: L.W.J.E. Beurskens: inclusion of patients, data input and analysis, drafting manuscript; L.H. Schrijver: data input and analysis, drafting manuscript; D. Tibboel: primary researcher, inclusion of patients, manuscript revision; M.F. Wildhagen: database management and data extraction; M.F.C. Knapen: inclusion of patients, manuscript revision; J. Lindemans: sample analysis, manuscript revision; J. de Vries: design questionnaires and data analysis, manuscript revision; R.P.M. Steegers-Theunissen, primary researcher, data analysis, manuscript revisions.

  • §

    Details of ethics approval: Permission to conduct the study was granted by the Central Committee of Research in Human in The Hague, The Netherlands, and by the Medical Ethical Committee of Erasmus University Rotterdam (MEC-2006-158; July 20, 2006).

  • Supported by a grant from the Mother and Child Center of the Erasmus Medical Centre – Sophia's Children's Hospital.

Abstract

BACKGROUND

Vitamin A has been related to the etiology of congenital diaphragmatic hernia (CDH). We performed a case-control study to investigate whether maternal dietary vitamin A intake is related to CDH in the offspring.

METHODS

Thirty-one pregnancies diagnosed with CDH and 46 control pregnancies were included during the study. After CDH diagnosis and inclusion of controls by risk set sampling, maternal vitamin A intake was investigated with a food frequency questionnaire. Serum retinol and retinol-binding protein were determined. Univariable and multivariable logistic regression models were used to calculate risk estimates with adjustment for potential confounders.

RESULTS

We found no significant differences in the overall nutrient and vitamin A intake between case and control mothers. After stratification in body mass index (BMI) categories, case mothers with normal weight showed a lower energy adjusted vitamin A intake (685 vs. 843 μg retinol activity equivalents [RAEs] / day; p = 0.04) and a slightly lower serum retinol (1.58 vs. 1.67 μmol/L; p = 0.08) than control mothers. Vitamin A intake <800 μg retinol activity equivalents (recommended daily intake) in normal weight mothers was associated with a significantly increased CDH risk (odds ratio [OR], 7.2; 95% confidence interval [CI], 1.5–34.4; p = 0.01). Associations were not significantly different in underweight and overweight mothers.

CONCLUSIONS

In normal-weight mothers, dietary vitamin A intake during pregnancy below the recommended daily intake is significantly associated with an increased risk of a child with CDH. This finding supports the retinoid hypothesis in human CDH, but warrants further investigation in larger study populations. Birth Defects Research (Part A), 2013. © 2013 Wiley Periodicals, Inc.

INTRODUCTION

Vitamin A has several crucial roles in fetal development. Nutritional experiments in animals have shown that dams fed a vitamin A deficient diet give rise to offspring with multiple congenital anomalies (See et al., 2008). On the other hand, an excess of dietary vitamin A is teratogenic in both animals and humans (Geelen, 1979; Mastroiacovo et al., 1999). In humans, the effect of high vitamin A exposure by drugs, such as the retinoic acid isomer isotretinoin and synthetic vitamin A derivative, has been described as the retinoic acid embryopathy (Lammer et al., 1985; Shaw et al., 1995). To prevent the teratogenic effect of retinoids, pregnant women are recommended to keep preformed vitamin A intake within the recommended daily allowance (RDA) for pregnancy.

Disturbances in vitamin A homeostasis have been associated with the development of congenital diaphragmatic hernias (CDHs). The birth prevalence rate of this severe developmental anomaly is approximately 1 in 3000 (Torfs et al., 1992). In CDH, the diaphragm fails to close during early embryogenesis resulting in a hole in the diaphragm. The underlying pulmonary hypoplasia and gastrointestinal sequelae lead to a considerable mortality and morbidity in patients with CDH (American Academy of Pediatrics Section on Surgery, 2008). In the vast majority, the cause of the failure of the diaphragm to close is unknown. Genetic defects and candidate genes have been identified in animal models, but only in a minority of patients with CDH (Holder et al., 2007). The phenotypic and genetic heterogeneity in patients with CDH may be explained by a multifactorial etiology in which genetic variations and harmful periconceptional exposures disrupt early diaphragm development. Of these exposures, maternal nutritional factors have been increasingly associated with CDH (Beurskens et al., 2009).

In animal models, a CDH-like phenotype has been linked to a deficient maternal intake of vitamin A (Anderson, 1941), teratogenic disturbances of retinoid homeostasis (Mey et al., 2003), genetic defects in retinoid signaling (Mendelsohn et al., 1994), and the expression of retinoid-related genes in the primordial diaphragm (Clugston et al., 2008). Noteworthy, administration of vitamin A reduced the incidence of CDH in animal models (Thébaud et al., 1999). In humans, CDH has been linked to frequently deleted or duplicated chromosomal regions that contain genes related to the vitamin A pathway (Klaassens et al., 2005; Holder et al., 2007). Newborns with CDH show reduced concentrations of retinol and retinol binding protein (RBP) in cord blood compared with healthy newborns (Major et al., 1998; Beurskens et al., 2010). However, the retinol and RBP concentrations at birth seem not to be deranged in the mothers of these children (Major et al., 1998; Yang et al., 2008; Beurskens et al., 2010).

From this background, we suggest that CDH is associated with maternal dietary intake of vitamin A during pregnancy. Because of the relatively low birth prevalence of CDH, a prospective cohort study from preconception onward is not feasible and, therefore, a case-control study was being conducted. To reduce confounding by gestational age, we performed risk-set sampling of the controls at the moment of CDH diagnosis during pregnancy.

METHODS

Study Population

The Congenital Diaphragmatic Hernia, Environment, Retinoids, Nutrition, Inheritance, and other Associations (HERNIA) study is a hospital-based case-control study conducted at Erasmus Medical Centre, University Medical Center, Rotterdam, The Netherlands. The study protocol was approved by the Central Committee of Research in Humans in The Hague, The Netherlands, and by the Medical Ethical Committee of Erasmus University Rotterdam. Written informed consent was obtained from both parents and on behalf of their child. The cohort investigated in the current study largely overlaps the cohort we described in an earlier article, albeit that this cohort was studied at a different time point (Beurskens et al., 2010).

The case mothers were enrolled immediately after the diagnosis of CDH by a prenatal ultrasound scan. Controls were recruited from the same tertiary center by risk-set sampling at the moment of diagnosis of the child with CDH. The moment of inclusion was defined as the gestational age at inclusion calculated from the last menstrual period. To increase homogeneity, we selected controls with a comparable maternal age (± 1 year), nationality, and mode of conception. The main exclusion criterion for controls was a fetus with a congenital anomaly. Immediately after inclusion, mothers completed a general questionnaire on health, family and obstetrical history, life-style factors, exposures, socio-economic status, and vitamin supplement use. Furthermore, they completed a food frequency questionnaire (FFQ) covering the preceding 4 weeks. The periconception period was defined as 4 weeks before until 8 weeks after conception. The data presented here are derived from the case and control inclusions between 2006 and 2009, and for the current analysis, the maternal data only are evaluated.

Fifty pregnancies of a fetus with CDH were included during this study. Nineteen of these were excluded from the analysis: 11 mothers did not complete the questionnaires and 4 were not able to do so because of language problems; 1 case family did not give informed consent; and 3 questionnaires were not returned for unknown reasons. All included control families completed the questionnaires and were included in the analysis.

Data Collection

To estimate the dietary intake of vitamin A, energy, and macronutrients, we used a validated semiquantitative FFQ that covered the maternal intake of the 4 weeks preceding the moment of enrollment (i.e., during pregnancy [Feunekes et al., 1993]). Nutritional habits do not tend to change during pregnancy (Devine et al., 2000). However, we took into account the covariates nausea, vomiting, and the use of a special diet during the first trimester. Therefore, the nutritional intake at the time of enrollment (2nd trimester) largely reflects the intake during the first trimester (van Driel et al., 2009; Willett, 1998). The FFQ has been updated and validated twice and has been shown reliable for the estimation of dietary intake of energy, fats, folate, and vitamin B12 in women of reproductive age (Feunekes et al., 1993; Feunekes et al., 1995; Verkleij-Hagoort et al., 2007). The FFQ consists of 121 items and covers 90% of the daily intake of foods or nutrients. Information of preparation methods, portion sizes, additions, and frequency of foods (per day, per week, per month, or not at all) are collected. The average daily nutrient intake is calculated by multiplying the frequency of consumption of food items by portion size and nutrient content per gram based on the 2006 Dutch Food Composition Table. The vitamin A intake was estimated in retinol activity equivalents (RAEs) using the conversion formula 1 RAE = 1 μg retinol = 12 μg β-carotene = 24 μg α-carotene and β-cryptoxanthin (Food and Nutrition Board, Institute of Medicine, 2001). We used “RAE intake” or “vitamin A” intake.

At inclusion, non-fasting venous blood samples were collected in lithium-heparin tubes, protected from light by wrapping in aluminum foil and kept refrigerated (4°C) until further processing. As soon as possible after collection, plasma was processed by centrifugation (15-minutes at 1800 × g), aliquotted into amber tubes and stored at −80°C. The serum retinol concentration was measured with reversed phase HPLC and the serum RBP concentration was measured with an ELISA, as described previously (Beurskens et al., 2010). The intra-assay and inter-assay variabilities of retinol were 3.9% and 5.1%, respectively; those of RBP 5.0% and 9.8%, respectively.

Maternal body mass index (BMI) was calculated from height and weight before pregnancy using the standard formula weight (kg) / height2 (m2). We defined three categories: underweight (BMI <18.5 kg/m2), normal weight (BMI 18.5–24.9 kg/m2), and overweight (BMI ≥25 kg/m2). The overweight category includes obese participants (BMI > 30 kg/m2) because of limited numbers. Education was defined according to Statistics Netherlands (2008). High education was defined as higher vocational or university education (Statistics Netherlands, 2008). Gestational age was calculated from the last menstrual period or based on ultrasound examination and calculated from the date the FFQ was completed.

Data Analysis

Questionnaires on nutrition may be subject to underreporting (Goldberg et al., 1991). To evaluate underreporting of energy intake, we used the physical activity level (PAL) (Goldberg et al., 1991), which we adjusted for pregnant women. PAL is calculated by the ratio of the reported energy intake divided by the basal metabolic rate (EI/BMR) (Goldberg et al., 1991). The BMR was calculated with the Schofield equations (Schofield, 1985), and we added 0.5 megajoules to the individual BMR because pregnant women have a higher BMR of 0.5 megajoules /day on average (Butte and King, 2005; Lof et al., 2005). Finally, the PAL cutoff point was decreased 10% to 1.305 because the PAL of pregnant women is 10% lower than that of nonpregnant women (Butte and King, 2005; Lof et al., 2005).

We compared the general characteristics between cases and controls with the two-sample t test for normal distributed continuous variables, and the chi-square test for categorical variables. The intake of various nutrients (medians) was compared using the Mann-Whitney U test. Patients were stratified in BMI categories underweight, normal weight, and overweight, as described above. The inclusion of the study populations was matched on group level. Therefore, we performed unmatched analyses with univariable and multivariable logistic regression models to estimate odds ratios (ORs) for the association between CDH and nutrient intake and to correct for the potential confounders maternal age, EI, BMI, and the use of vitamin supplements. Because of the relationship between BMI and nutrient intake in a sensitivity analysis, we performed a stratified analysis for BMI. We estimated the OR using the RDA per nutrient as the cutoff point. All analyses were performed with statistical software (SPSS 15.0, IBM, Chicago, IL).

RESULTS

Table 1 displays the general characteristics of the study population. Maternal age, parity, and the level of education were significantly higher in controls than in cases. We found no significant differences in the periconceptional use of a multivitamin or folic acid supplement. Factors that could disturb maternal dietary intake in the first trimester, such as nausea, vomiting, and a special diet, were not different between case and control mothers (Table 1). Mean PAL for the total study population was 1.51 and comparable between cases and controls (1.47 vs. 1.54; p = 0.60). Because the value of PAL was higher than the cutoff value (1.305), there is no evidence for differential underreporting in case and control mothers. However, we found a significant trend for a decrease in PAL with an increase in BMI. In normal weight mothers (case vs. control: 1.55 vs. 1.67; p = 0.52), the PAL was higher than in overweight mothers (case vs. control: 1.25 vs. 1.30; p = 0.74). This suggests underreporting of energy intake in overweight case and control mothers, which is a regular finding in nutritional studies.

Table 1. General Characteristics of Case and Control Mothers
 CharacteristicsCases (n = 31)Controls (n = 46) p value 
  • General characteristics of case and control mothers at the moment of congenital diaphragmatic hernia diagnosis and the inclusion moment. Values are means ± SD or numbers (%).

  • a

    Gestational age is presented in median (interquartile range); the statistical difference was tested using the Mann-Whitney U test.

  • b

    Education is defined according to Statistics Netherlands (2008). Categorized as low (primary/lower vocational/intermediate secondary education, higher secondary/intermediate vocational education), or high (higher vocational/university education) (Statistics Netherlands, 2008).

  • c

    Physical activity level is defined as the ratio of energy intake and basal metabolic rate.

  • d

    The periconceptional period was defined as 4 weeks before until 8 weeks after conception.

  • e

    Separately or in multivitamin supplement.

  • f

    Mode of conception was defined as “artificial” if the mother used hormones or if in vitro fertilization took place. Mode of conception missing in one case and one control.

  • gAdjusted for energy intake. Energy intake was adjusted for by defining the nutrient intake in energy percentage (En%) and using this value in the statistical tests.

  • h

    Retinol 25 cases and 42 controls. RBP 25 cases and 42 controls.

  • Differences between cases and controls were tested with the Mann-Whitney U test. Energy intake was adjusted for by adjusting the variable for energy intake (in energy percentage [En%]) and test the adjusted variable between groups. Both maternal age and parity did not significantly differ between the included and excluded cases (data not shown).

  • *

    Significant at p < 0.05.

  • **

    Significant at p < 0.01.

  • BMI, body mass index; En%, energy percentage; MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; RBP, retinol binding protein.

Constitutional (determined at inclusion)    
 Age (years)31.6 ± 5.034.7 ± 4.30.004** 
 BMI category (kg/m2)    
  Underweight (<18.5 kg/m2)1 (3.2)0 (0)0.44 
  Normal weight (18.5–24.9 kg/m2)17 (54.8)26 (56.5)  
  Overweight (>25.0 kg/m2)8 (25.8)14 (30.4)  
  Unknown5 (16.1)6 (13.0)  
 Parity  0.001** 
  021 (67.7)14 (30.4)  
  ≥110 (32.3)32 (69.6)  
 Gestational age (weeks)a Median (interquartile range)30.0 (24.0–37.0)30.0 (19.0–35.0)0.85 
 Nationality  0.55 
  Dutch27 (87.1)42 (91.3)  
  Non-Dutch4 (12.9)4 (8.7)  
 Educationb  0.03* 
  Low22 (71.0)21 (45.7)  
  High9 (29.0)25 (54.3)  
 Use of special diet2 (6.5)4 (8.6)0.72 
 Basal metabolic rate (kJ/day)6776 ± 8646690 ± 7680.68 
 Physical activity levelc1.47 ± 0.431.54 ± 0.580.60 
Periconceptionald    
 Folic acid supplement usee  0.37 
  Yes23 (74.2)38 (82.6)  
  No8 (25.8)8 (17.4)  
 Multivitamin supplement use  0.54 
  Yes21 (67.7)28 (60.9)  
  Underweight (<18.5 kg/m2)10  
  Normal weight (18.5–24.9 kg/m2)1118  
  Overweight (>25.0 kg/m2)59  
  Unknown41  
  No10 (32.3)18 (39.1)  
  Underweight (<18.5 kg/m2)00  
  Normal weight (18.5–24.9 kg/m2)68  
  Overweight (>25.0 kg/m2)35  
  Unknown15  
 Mode of conceptionf  0.84 
  Spontaneous28 (93.3)39 (86.7)  
  Artificial2 (6.7)6 (13.3)  
 Smoking  0.15 
  Yes8 (25.8)6 (13.0)  
  No23 (74.2)40 (87.0)  
First trimester    
 Nausea21 (67.7)30 (65.2)0.82 
 Vomiting8 (25.8)11 (23.9)0.85 
Nutrient intakes at inclusion Median (interquartile range)   p value p value (adjusted)g
 Retinol activity equivalents (μg/day)695 (366)805 (388)0.440.39
 Energy (kJ/day)9636 (2458)9188 (4008) 0.63
 Total fat (En%)35.5 (4.4)35.5 (5.9) 0.88
 Saturated fat (En%)13.2 (2.74)13.1 (2.4) 0.95
 MUFA (En%)11.1 (2.2)11.3 (2.2) 0.72
 PUFA (En%)6.5 (1.8)6.9 (2.8) 0.53
 Protein (En%)15.2 (2.9)15.3 (2.6) 0.88
 Carbohydrates (En%)49.8 (4.75)49.5 (7.9) 0.74
Biomarkersh at inclusion Mean (SD)    
 Retinol (μmol/L)1.62 (0.26)1.65 (0.34)0.75 
 RBP (mg/L)10.79 (4.07)10.84 (4.89)0.97 

Table 1 shows the nutrient intake and biomarker concentrations of the case and control mothers. We found no significant differences between case and control mothers in crude and energy adjusted nutrient intakes, vitamin A intake, and the proportion of carotenoids of the RAE (0.23 vs. 0.24; p = 0.20). Adjustment for parity did not change this outcome. The serum concentrations of retinol and RBP were comparable between case and control mothers. The correlation between dietary vitamin A intake and retinol or RBP concentrations was low. Spearman's rho for energy adjusted RAE and retinol was −0.108 (p = 0.39) and for RBP −0.156 (p = 0.21).

In addition, we stratified nutrient intake into three BMI categories. In case mothers with normal weight, energy adjusted vitamin A intake was significantly lower (685 vs. 843 μg/day; p = 0.04) (Table 2). CDH risk was significantly increased in normal weight mothers with a vitamin A intake below the Dutch daily recommended intake of 800 μg per day (OR, 5.4; 95% confidence interval [CI], 1.4–20.5; p = 0.01) energy adjusted (OR, 7.2; 95% CI, 1.5–34.4; p = 0.01; Table 3). After adjustment for maternal age and education, the risk remained significantly increased (OR, 1.6; 95% CI, 1.14–20.79; p = 0.03). In normal weight mothers, the level of retinol was lower in the case mothers, albeit not significantly (Table 2).

Table 2. Maternal Dietary Vitamin A Intake Calculated in RAEs and Serum Concentrations of the Maternal Vitamin A Biomarkers Retinol and RBP Stratified in BMI Categories
 RAE intake (μg/day) Median (interquartile range) Serum retinol (μmol/L) Mean (SD)Serum RBP (mg/L) Mean (SD)
 No.CaseNo.Control p value p value adjustedaCase/ controlCaseControl p valueCaseControl p value
  • Maternal dietary vitamin A intake calculated in RAEs and serum concentrations of the maternal vitamin A biomarkers retinol and RBP stratified in BMI categories. Analysis was performed for each category. RAE intake was compared using the Mann-Whitney U test. Retinol and RBP concentrations were compared using the t test.

  • a

    Adjusted for energy intake.

  • *

    Significant at p < 0.05.

  • Serum retinol and RBP levels during pregnancy in case and control mothers, stratified by BMI category. Adjustment for parity, age, or education did not change significance.

  • RAEs, retinol activity equivalents; RBP, retinol binding protein; BMI, body mass index.

Total population26695 (366)40805 (388)0.440.3924/391.62 (0.26)1.65 (0.34)0.7510.79 (4.07)10.84 (4.89)0.97
Stratified according to BMI category             
 Underweight16980---1/02.05- 13.90- 
 Normal weight17685 (343)26843 (441)0.0530.04*16/251.58 (0.26)1.67 (0.37)0.0810.10 (4.53)10.79 (4.06)0.31
 Overweight8734 (438)14613 (368)0.2380.247/141.64 (0.24)1.64 (0.31)0.9912.11 (3.13)11.41 (6.06)0.78
Table 3. Maternal Risk of CDH Offspring According to Maternal Vitamin A Intake Calculated as RAE Intake <800 μg/Day Stratified in BMI Categories
 Maternal RAE intake <800 μg/day
 Case/controlOR adjusteda (95% CI) p value
  • The maternal risk of CDH offspring according to maternal vitamin A intake calculated as RAE intake <800 μg/day stratified in BMI categories. The risk was calculated by using univariate and multivariable logistic regression models.

  • a

    Adjusted for energy intake.

  • *

    Significant at p < 0.05.

  • Logistic regression analysis was not possible in the underweight category because of low numbers.

  • CDH, congenital diaphragmatic hernia; RAE, retinol activity equivalent; BMI, body mass index; OR, odds ratio; CI, confidence interval.

Total population20/242.3 (0.8–6.4)0.122
UnderweightNANANA
Normal weight12/87.2 (1.5–34.4)0.013*
Overweight4/40.35 (0.05–2.42)0.29

DISCUSSION

In this case-control study, we demonstrate that in normal weight mothers, a lower dietary vitamin A intake is associated with an increased risk of a child with CDH. In line with our previous finding in the same study population (Beurskens et al., 2010), serum retinol and RBP in case mothers with normal weight were lower than in control mothers, albeit not significantly. The relatively small sample size in the current study warrants validation in larger study populations.

The association between a maternal dietary vitamin A intake below the recommended daily intake and CDH risk is in line with data from animal models in which vitamin A deficiency induces the development of CDH (Anderson, 1941; See et al., 2008). In humans and other mammals, the diet is the only source of vitamin A and, although vitamin A deficiency is an endemic problem throughout the developing world (West, 2003), there are no reports of increased incidences of CDH in developing countries. However, a detailed epidemiologic study has not been performed and most of these countries lack registries of birth defects. In addition, it is well known that vitamin A and its metabolite retinoic acid have a crucial role in animal lung development (Chen et al., 2010), which is disturbed in patients with CDH. Further, maternal vitamin A repletion in a vitamin A deficient population improves lung function in the offspring (Checkley et al., 2010). These findings are in line with our observation that low maternal vitamin A intake may be a risk factor for development of CDH.

The differences in vitamin A intake and the increased risk estimates were statistically significant in normal weight mothers. This can partially be explained by confounding because of underreporting of EI by the overweight mothers in both cases and controls, which is a known phenomenon in nutritional epidemiology (Garriguet, 2008; Bothwell et al., 2009). This makes the estimation of the nutrient intake less reliable in the overweight group and may disturb the associations in the total study population. Because a continuous analysis would have ignored this phenomenon, we chose to stratify the population in three BMI categories. This is not a reason to apply the conservative Bonferroni correction on the data. A larger study (Yang et al., 2008) also describes a trend for an increased CDH risk in association with lower vitamin A intakes. Although this was not significant, the authors did not adjust for BMI nor investigated the mothers with normal weight separately. Further, one would expect a relationship between low vitamin A intake and CDH in the underweight group. However, we were not able to test this hypothesis because there was only one underweight case and no underweight controls. We did not find a relationship between overweight women and a lower risk on CDH. It could be possible that the higher amount of fatty tissue in overweight women and, thus the increased storage capacity of retinoids, influence the risk on CDH. However, as we have stated, the retinol homeostasis is strictly controlled, but this does not exclude a protective relationship between overweight women and CDH. Further research is needed to clarify this relationship.

The biomarkers retinol and RBP reflect the vitamin A intake of the previous 4 weeks, the period covered by the FFQ. The correlation between these biomarkers and vitamin A intake were marginal in the total study population and not significantly different between cases and controls (results not shown). In an earlier report, we described that the serum levels at birth were comparable in case and control mothers (Beurskens et al., 2010). This is in contrast to one report with lower numbers showing higher levels in case mothers at birth (Major et al., 1998). However, in these two other studies, the levels of retinol and RBP were not stratified for BMI category. Although adipose tissue has retinol storage capacity, there are no reports that a high percentage of adipose tissue influences the blood retinol concentration. The fact that only a small part of the vitamin A intake is reflected in the blood retinol and RBP concentrations may explain that the lower intake was also not reflected in lower serum levels. The assessment of vitamin A intake as RAE does not discriminate between all different forms of vitamin A (retinoids) and might neglect unique physiologic functions of carotenoids, which may be relevant to the development of the embryo (Willett, 1998). Additionally, serum retinol is tightly regulated, a mechanism that is conserved during pregnancy (Miller et al., 1998; Ross and Gardner, 1994). This may explain why the correlation between the estimated vitamin A intake and serum level of retinol is generally found to be low or absent, as was the case in the present study. The level of β-carotene may have a better correlation with vitamin A intake (Ascherio et al., 1992). However, we did not determine β-carotene levels because of its strong dependence of the time interval between sampling and the last meal. In addition, the proportion of carotenoids intake of RAE was comparable between cases and controls. This substantiates the suggestion that an increased CDH risk could be because of a decrease in dietary intake of retinoids other than retinol.

Furthermore, the accuracy of nutrient calculation depends on the representativeness of the food composition database. Moreover, the size, growth, processing, storage, cooking, and genetic variability of vegetables and fruit may lead to differences in carotenoid levels between food composition tables (Willett, 1998). In our study, the recall period of the FFQ was 4 weeks, which is an optimal time period to account for the large day-to-day variation of intake of vitamin A. Further, the FFQ was handed out at the moment of enrollment and the participants were instructed to fill out the questionnaire on the 4 weeks before the date of enrollment. Therefore, it is unlikely that knowledge of having a child with a congenital defect has resulted in a change of the diet or has influenced the FFQ data. Participants were included during the whole year, and the questionnaires thus represent all seasons, accounting for a seasonal variation in carotenoids (Willett, 1998). Moreover, in our study, the RAE was determined largely by the same food products in both cases and controls. Nevertheless, it would be interesting to investigate the intake of retinol and carotenoids separately and to study the intake of specific retinoids and measure the serum concentrations simultaneously.

We found no significant differences between the included and excluded cases. However, the excluded patients did not complete the questionnaires with information on parameters such as lifestyle factors and socioeconomic score (SES). Therefore, selection bias cannot completely be excluded. Parity differed between the included case and control mothers. If parity is considered a proxy for nutrition, we should have adjusted for this difference. However, age and education seemed to be highly correlated with parity. Therefore, adjustment would have resulted in overcorrection. Nevertheless, the risk of CDH with an intake under the RDA remained significant after adjustment for parity with a smaller 95% CI. The difference in age and parity between cases and controls can be explained by the fact that older women are more likely to have given birth previously. In case mothers, we found a lower educational level or SES. A low SES may be related to an increased risk of non-chromosomal congenital anomalies (Vrijheid et al., 2000), but this is not supported by other studies (Yang et al., 2006; Felix et al., 2008). Furthermore, a low SES is related to a poorer nutrition and, in the etiologic pathway, it precedes nutritional intake in the association with congenital anomalies (Shahar et al., 2005; Darmon and Drewnowski, 2008). Still, after adjustment for these factors, the risk estimate remained significantly increased.

A strong point of this study is the adjustment of the Goldberg equation (underreporting of energy intake) for BMR and PAL, which was based on several reports that described an increased BMR and a decreased PAL during pregnancy (King, 2002; Butte et al., 2004; Lof et al., 2005; Melzer et al., 2009). Our observation that nutritional underreporting increased with higher BMI, corresponds with other studies (Garriguet, 2008; Scagliusi et al., 2008; Bothwell et al., 2009), and improved the reliability of our data.

In conclusion, our study – in line with animal models – suggests that in normal weight mothers, a maternal dietary intake of vitamin A below the RDA may contribute to the risk for CDH. The number of study participants was relatively small and therefore larger sample sizes of preferably cohort studies are recommended to replicate our results. Based on the known teratogenicity of vitamin A and the preliminary nature of our findings, the current recommendation for pregnant women not to take supplements containing vitamin A still stands.

Acknowledgements

The authors acknowledge the contributions of Mrs. J. van Rhee and W. Keller for the inclusion of the participants, and Mr. K. Hagoort for editing the English language of the article.

Ancillary