The Norwegian Mother, Father and Child Cohort Study (MoBa) is a prospective population-based pregnancy cohort study conducted by the Norwegian Institute of Public Health (18). Participants were recruited from all over Norway from 1999 to 2008, and 41% of the invited women consented to participate. The cohort now includes approximately 114,500 children, 95,200 mothers and 75,200 fathers. MoBa data are linked to information from the Medical Birth Registry of Norway, a national health registry containing information about all births in Norway (19). The Norwegian Environmental Biobank is a sub-study within MoBa (20), established with the aim of biomonitoring nutrients and environmental contaminants in MoBa participants. Eligibility of participants in the Norwegian Environmental Biobank required available genetic data, available data from the three questionnaires during pregnancy and the first three postnatal questionnaires, the fathers’ questionnaire (21), and available maternal plasma, urine, and whole blood samples. In total, 2,999 women from MoBa were included in the Norwegian Environmental Biobank. Of these 2,988 had the biomarker 25OHD analyzed in pregnancy and were thus eligible for inclusion in the current study. Further, 28 participants (0.9%) with implausible reported daily energy intake below 1,076 kcal (4.5 MJ) or above 4,780 kcal (20 MJ) were excluded from the analyses. These cut-points are used in MoBa to exclude food frequency questionnaires (FFQs) of poor quality and are based on results from the validation study (22). Thus, the final analytic sample consisted of 2,960 pregnant women.

MoBa is conducted according to the guidelines laid down in the declaration of Helsinki and written informed consent was obtained from all participants. The establishment of MoBa and initial data collection was based on a license from the Norwegian Data Protection Agency and approval from The Regional Committees for Medical and Health Research Ethics. MoBa is currently regulated by the Norwegian Health Registry Act. The current study was approved by The Regional Committees for Medical and Health Research Ethics (REC 2019/770–12172). The current study is based on version 12 of the quality-assured data files released for research in January 2019 (23).

The women’s habitual food consumption and supplement use since becoming pregnant was assessed by a semi-quantitative FFQ including 255 food items answered in gestational week 22 (22, 24, 25). Intake frequency of each item in the FFQ ranged from never to more than eight times a day. Portion size was only specified for liquids, bread, and fruit, while standard Norwegian portion sizes for women were used for all other food items (26). Food intake (g/day) and energy intake (kcal/day) was estimated based on the given intake frequency and portion size. At the time of data collection, the following foods were fortified with vitamin D: margarine and butter (8.0 μg/100 g), fat-reduced milk (0.4 μg/100 g), and lactose free milk (0.4 μg/100 g). In the current study, all foods containing vitamin D were aggregated into eight food groups: fish, margarine, butter, milk, yoghurt, eggs, cheese, and mixed dishes/products containing milk, margarine, butter, and/or egg. The women’s intake of these foods (g/day), as well as their contribution to the daily vitamin D intake (μg/day) were calculated. The women’s intake of vitamin D from supplements was also calculated, based on the frequency and amount of supplement intake reported in the FFQ. The FFQ listed 13 commonly used cod liver oil/fish oil, vitamin, and mineral supplements followed by six open-ended spaces where respondents were instructed to record the name and manufacturer of supplement(s) used but not listed. For each supplement there were nine options for frequency, ranging from never to daily, and for dose there were three options for liquid supplements and four options for tablets/capsules. For nutrient calculation of the dietary supplements, a database containing the nutrient value of more than 1,000 different food supplements was constructed and continuously updated (24, 27).

The estimated vitamin D intake from foods and supplements has previously been validated in a sub-group of 119 women using a weighed food diary and biomarkers. The results showed fair agreement between reported vitamin D intake and 25OHD with Spearman’s rank correlation coefficient (0.32–0.45, p < 0.01) and <5% classified into opposing quintiles of vitamin D intake and 25OHD (24, 25).

The blood samples were collected in gestational week 18 (mean 18.5, SD 1.3) in connection with the routine ultrasound examination, offered to all pregnant women in Norway (20). Samples were analyzed at the National Institute for Health and Welfare in Helsinki, Finland. Plasma concentrations of 25OHD were analyzed in 2015 using the high through-put automated chemiluminescent microparticle immunoassay method (Architect ci8200 system, Abbott Laboratories, Abbott Park, IL, United States). Both 25OHD2 and 25OHD3 are measured and the sum is reported. The laboratory regularly partake in the Vitamin D External Quality Assessment Scheme and met the performance target set by the advisory panel (28). Coefficient of variation of control samples (Biorad Liquid Assayed Multiqual lot 45,680 high and low level) were 3.7–5.5%.

Maternal age at time of delivery, parity, and country of origin was obtained from the Medical Birth Registry of Norway. Information about education, pre-pregnant weight, height, smoking habits, and use of solarium was obtained from questionnaires, answered around gestational weeks 15 and 30. In gestational weeks 15, participants were also asked about use of vitamin D containing supplements (yes or no) by weekly intervals prior to conception, and in the first trimester.

Dietary intake and supplement use during the first half of pregnancy were described by categories of 25OHD, based on the reference values (<30 nmol/L, 30–49.9 nmol/L, 50–75 nmol/L, >75 nmol/L) in NNR 2012 (13) and National Academy of Medicine (Institute of Medicine) (12). Statistical difference in study population characteristics between the categories of 25OHD were assessed by Kruskal Wallis test.

The candidate determinant variables of vitamin D status were selected based on biological plausibility and included: season at blood sampling (two seasons; June–September, October–May), vitamin D intake from foods (<2.5, 2.5–4.9, 5.0–7.49, ≥7.5 μg/day), supplemental vitamin D intake (none, 0.1–4.9, 5.0–9.9, 10.0–14.9, ≥15.0 μg/day), energy intake (kcal/day), use of solarium during pregnancy (no, 1–5, ≥6 times), country of origin (Norway, other high income country, low/middle income country), education (<13, 13–16, >16 years), pre-pregnancy BMI (<18.5, 18.5–24.9, 25.0–29.9, ≥30.0 kg/m²), age (<25, 25–34, >34 years), smoking during pregnancy (yes, no), and parity (nulliparous, multiparous). The variables were checked for multicollinearity using Spearman correlation matrix, and all correlation coefficients were below 0.4. Variables were selected using stepwise backward selection (29) and p < 0.2 as cut-off for inclusion. Bootstrap subsampling was used to investigate and quantify the stability of selected variables (30). One hundred resamples were drawn and variable selection was repeated in each of the resamples. Variables included in >50% of the resamples were selected in the final model (30). No variable was excluded after the stability check. Results are presented as a full sample and stratified by season, as season is the main source of vitamin D and that there is a large seasonal variation in 25OHD concentrations in the Nordic countries (17), and pre-pregnancy BMI. The estimates of the determinants of 25OHD concentration in pregnancy were investigated using multivariable linear regression analysis.

The association between the total vitamin D intake and continuous 25OHD, in all women and stratified by season and pre-pregnancy BMI, was explored using restricted cubic splines. Three splines were modeled for the total vitamin D intake, positioned at percentiles 10, 50, and 90 (31). Models were adjusted for country of origin, age, and smoking during pregnancy. The models in all women and stratified by season were additionally adjusted for pre-pregnancy BMI. The overall associations between total vitamin D intake and 25OHD were performed by testing the coefficients of all spline transformations equal to zero. Non-linearity was examined by testing the coefficients of the second spline transformation equal to zero. The associations between use of vitamin D containing supplements before and during pregnancy and maternal concentrations of 25OHD were analyzed by Kruskal Wallis test.

Stata version 16 was used for all statistical analyses (Stata Corporation, College Station, Texas). Significance was accepted at p < 0.05.

Of the 2,960 participants, 72% had a high education (≥13 years) and 5.9% reported smoking during pregnancy (Table 1). Only 4.8% were born in a high-income country other than Norway, and 1.4% were born in a low or middle-income country. There were differences between vitamin D categories with regard to the characteristics of the study population in terms of, e.g., age, education, country of origin, and pre-pregnancy BMI. About 80% of the participants reported use of vitamin D supplement in pregnancy. Around 44% of all women had a total vitamin D intake below AR at 7.5 μg/day and 61% had an intake below RI at 10 μg/day (13). Median (25^th–75^th percentiles) 25OHD concentration was 51 (38–64) nmol/L.

Overall, the median vitamin D intake from foods was 3.1 μg/day, and the total vitamin D intake (from foods and supplements combined) was 8.3 μg/day (Table 2). Thus, the main contributor to vitamin D intake in pregnancy was vitamin D supplements (66%, Figure 1). The main food sources of vitamin D were fish and fortified margarine.

There was a difference in vitamin D intake from both foods and supplements between categories of 25OHD concentration, but no differences were observed for intake of energy or other nutrients (Table 2). Among the major food sources of vitamin D, there was a difference. In the consumption of margarine, but not fish or butter, between categories of 25OHD.

In a multivariable linear regression, determinants of 25OHD concentrations in pregnancy were (in descending order of the beta estimates) summer season, use of solarium, higher vitamin D intake from supplements, origin from high income country, lower pre-pregnancy BMI, higher age, higher vitamin D intake from foods, no smoking during pregnancy, higher education and energy intake (Table 3). Together, these variables explained 21% of the variation in 25OHD. During June to September, total vitamin D intake explained 1% of the variation in 25OHD, whereas during October to May, it was 7% (Table 3). Stratified by pre-pregnancy BMI, the total vitamin D intake explained 5% of the variation in 25OHD in women with pre-pregnancy BMI <25 kg/m², whereas in women with pre-pregnancy BMI ≥25 kg/m², it was 3% (Supplementary Table S1).

In the full sample of women, total vitamin D intake >5 μg/day was sufficient to reach predicted 25OHD >50 nmol/L (p-overall: <0.001, p non-linearity: <0.001, Figure 2). The effect of vitamin D intake on 25OHD seemed to level off at reported intakes >15 μg/day.

When total vitamin D intake was stratified by season of blood sampling, women who were sampled during June to September had a predicted 25OHD >50 nmol/L even at very low intakes (p-overall: 0.001, p non-linearity: 0.003, Figure 3A). However, during October to May, a total vitamin D intake ≥10 μg/day was required to reach predicted 25OHD >50 nmol/L (p-overall: <0.001, p non-linearity: 0.023, Figure 3B).

When total vitamin D intake was stratified by pre-pregnancy BMI, women with BMI <25 kg/m² had predicted 25OHD >50 nmol/L at vitamin D intakes ≥5 μg/day (p-overall: <0.001, p non-linearity: 0.004, Supplementary Figure S1A). Women with BMI ≥25 kg/m² required a vitamin D intake ≥15 μg/day to reach a predicted sufficient 25OHD (p-overall: <0.001, p non-linearity: 0.043, Supplementary Figure S1B).

There was a difference in 25OHD concentrations depending on duration and timing of vitamin D supplement use during pregnancy (Supplementary Table S2). Women who reported any use of vitamin D supplement, either before and/or in early pregnancy, or in mid pregnancy, or both, had a higher 25OHD compared to women who reported no use of vitamin D supplement either before or during pregnancy. Women who reported use of vitamin D supplement both before and/or in early pregnancy and in mid pregnancy had a higher 25OHD compared to those who started to take a supplement in mid pregnancy.

The main strength of our study is the availability of numerous potential determinants of vitamin D in a relatively large study population. In addition, questionnaires provided information both prospectively and retrospectively during pregnancy. The FFQ can be considered a suitable dietary assessment method in this study since only few foods contain significant amounts of vitamin D making food diaries less suitable and since the FFQ provides a fair estimation of the habitual diet over a specific period (49).

Some limitations should be considered in the interpretation of the results. Some studies have observed underestimation of 25OHD using Abbott Architect chemiluminescence immunoassays compared with liquid chromatography tandem mass spectrometry (50, 51), also in a pregnant population (52). Standardization of 25OHD was not possible as there was no plasma left for such analysis. If the assay method underestimated the 25OHD concentrations in our study, it would potentially lead to biased estimates of insufficiencies and deficiencies. Thus, the prevalence of vitamin D insufficiency and deficiency should be interpreted with caution. Further, some important variables were not available to us in the investigation of determinants of 25OHD, such as recent travels to southern latitudes and portion sizes. The participants included in the Norwegian Environmental Biobank were women with good compliance in MoBa. Thus, selection bias may be present and might negatively affect the representativeness and the external validity of the results. In addition, the lack of variation in country of origin in the study population limits the interpretation of the vitamin D intake and status in ethnic minorities. Other limitations related to the FFQ used in MoBa have previously been described (22, 25).

This study provides a thorough investigation of the vitamin D intake and other determinants of vitamin D status in pregnant women living in Norway. The contribution of the determinants is likely relevant to pregnant women also in other northern regions on corresponding latitudes. In addition, the response in 25OHD by total vitamin D intake might differ by season and pre-pregnancy BMI. Future studies should aim to investigate how vitamin D intake and status can be safely increased on a population level.