Introduction

Vitamin D status, as assessed by 25-hydroxyvitamin D (25(OH)D) concentrations, has been a worldwide target of research [1]. According to a systematic review [2], deficient 25(OH)D concentrations (<50 nmol/L) were identified among 54% of pregnant women (42–72% in North America, 18–90% in Europe, 46% in Eastern Mediterranean, 66–96% in Southeast Asia, and 41–97% in the Western Pacific). Evidence mostly from developed countries points out to associations of vitamin D deficiency during pregnancy with perinatal complications as preeclampsia [3], gestational diabetes mellitus [3], spontaneous foetal loss [4], preterm birth, and adverse anthropometric and neurodevelopmental outcomes [5].

Decreased dietary sources of vitamin D and reduced exposure to sunlight, due to seasonality, time spent outdoors and type of clothing, have been identified as factors associated with gestational 25(OH)D concentrations <50 nmol/L in the Middle East (latitudes 12°N–42°N) [6]. Recent cross-sectional and cohort investigations have reported similar findings in China, Indonesia, Malaysia, Saudi Arabia and the Switzerland [7,8,9,10,11], but there is still a paucity of longitudinal studies on predictors of vitamin D status during pregnancy in tropical, low to middle-income countries, and notably in Latin America and Africa [2]. In such settings, the burden of malnutrition frequently clusters micronutrient inadequacies [12]. Among Brazilian infants, for instance, each 1 μmol/L increase in retinol concentrations was associated with a 1.38-fold higher prevalence of vitamin D sufficiency (95%CI 1.18; 1.61) [13], but comparable analyses among pregnant women are scarce. As vitamins D and A play immunomodulatory roles, novel investigations exploring their potential synergic relationship seem important in pregnancy, given the relevance of foetal-maternal immune balance and several inflammatory pathways for successful perinatal outcomes [14].

Also, a majority of studies has focused on the dichotomous outcome of vitamin D deficiency [2]. Some studies have suggested, however, that maternal 25(OH)D ≥50 nmol/L would not guarantee an adequate vitamin D status for the neonate, as assessed in the cord blood [15], and that the chance for preterm birth was 62% lower (95%CI 0.23; 0.63) among pregnant women with 25(OH)D ≥100 nmol/L compared to those with 25(OH)D <50 nmol/L [16].

The present study aimed to prospectively investigate predictors of 25(OH)D3 concentrations in the third trimester among pregnant women in the Brazilian Amazon. We provide a comprehensive characterisation of predictors of 25(OH)D3 at the 25th, 50th, and 75th percentiles of its distribution, in a low to medium-income region close to the tropics, to favour promotion of a healthier vitamin D status and timely prevention of perinatal complications.

Subjects and methods

Study design and participants

The Maternal and Child Health and Nutrition in Acre, Brazil (MINA-Brazil Study) [17] is a population-based birth cohort conducted in Cruzeiro do Sul (latitude: −7.7, longitude: −72.7), a municipality in the Brazilian Amazon area. The city had an estimated population of 87,673 inhabitants in 2018 [18], with a Human Development Index of 0.664, classified as medium (between 0.600 and 0.699) and below the Brazilian national average [19]. The present prospective analysis focused on participants from the urban area of Cruzeiro do Sul, whose recruitment took place during pregnancy in each of the 13 primary health care units of the Family Health Strategy, while booking an antenatal care appointment. Pregnant women were screened between February 2015 and January 2016 [17]. Inclusion criteria were gestational age up to 20 weeks according to the last menstrual period (LMP), fixed residence in the urban area and intention to give birth in Cruzeiro do Sul. We estimated to screen around 854 pregnant women, based on the number of deliveries in the maternity hospital (n = 1780, in 2013), the proportion of residents in the urban area (60%), and coverage of local primary health care services (80%). Eligible pregnant women were invited to a baseline interview and two clinical evaluations; gestational age based on LMP was confirmed by ultrasound examination during the antenatal period [20]. Multiple pregnancies and participants without complete information on 25(OH)D3 concentrations during follow-up were excluded for the present analysis.

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving participants were approved by the Institutional Review Board of the School of Public Health of the University of São Paulo (protocol number 872.613/2014). Written informed consent was obtained from all participants. In case of teenage pregnancies, the adolescent’s legal guardian provided consent.

Data collection and laboratory procedures

At baseline (median [interquartile range, IQR] gestational age: 16.9 [14.8, 19.1] weeks), trained research assistants conducted structured face-to-face interviews during home visits to collect data on each participant’s age, self-reported skin colour (as defined by the Brazilian Institute of Geography and Statistics: white, mulatto, black, yellow or indigenous), schooling (≤9 or >9 years), ownership of household assets for generating a wealth index (divided into tertiles) [21], assistance from the Bolsa Família conditional cash transfer programme (no or yes), living in a masonry household (no or yes), access to treated water (no or yes), septic destination of sewage (no or yes), number of residents in the household (≤4 or >4 people), pregnant woman as the main provider of family income (no or yes), presence of a smoker in the household (no or yes), parity (nulliparous, primiparous or multiparous), pre-pregnancy smoking (no or yes), and pre-pregnancy obesity (no or yes, according to body mass index [BMI] >30 kg/m² for adult women [22] and BMI-for-age z-score >2 for adolescents) [23]. The participant’s history of morbidities was also assessed during these interviews, including elevated blood pressure, hypertensive disorders, diabetes mellitus, gestational diabetes, kidney diseases (no or yes).

Two clinical evaluations were sequentially scheduled during the antenatal period at local primary health care units, in the second trimester (19.9 [18.2, 21.6] gestational weeks) and afterwards in the beginning of the third trimester (27.6 [26.9, 28.4] gestational weeks). In each evaluation, structured interviews and blood collection were performed by the study team.

Information was gathered on physical activity (practice of leisure walk, commuting and/or physical exercises in the last week; no or yes) [24], use of multiple micronutrient supplementation with vitamin D (no or yes), sunlight exposure (>30 min in the last week; no or yes) [25], and sun protection practices (use of sunscreen, shade structures and face protection; never/rarely/sometimes or very often/always). A sun protection score was conceived to depict the cumulative occurrence of factors influencing vitamin D endogenous synthesis [26, 27]. Non-white skin colour, no exposure to sunlight, and frequent sun protection practices accounted each for 1 point of in the score, varying from 0 to 5 points. Considering the midpoint of the score range, lower sun protection was defined at ≤2 points, while higher sun protection was denoted by >2 points.

Trained nursing technicians collected 8-hour fasting venous blood samples (approximately 10 mL) from participants in the second and third trimesters of pregnancy. Sterile vacuum blood collection tubes, with protection from light, were used for serum determination. Blood samples were allowed to clot and then were centrifuged within two hours of collection, in a swinging bucket centrifuge for 10 minutes at 3000 × g. Serum aliquots were immediately stored in amber conical microtubes and frozen at −20 °C before being sent on dry ice to the Laboratory of Human Nutrition (School of Public Health, University of São Paulo), where tubes were kept at −70 °C until analysis, within six months of each blood collection.

Serum concentrations of 25(OH)D3 in the second and third trimesters were determined on an isocratic HPLC system with UV detector with Chromsystems components and HPLC column, equilibrated and with test chromatogram (Chromsystems Instruments & Chemicals GmbH, Gräfelfing, Germany). Sample preparation was standardised strictly according to the manufacturer’s instructions, in light-protected reaction vials for precipitation and elution procedures, followed by the reading into the HPLC system. The limit of quantification was equivalent to 3.5 nmol/L, with linearity up to at least 624 nmol/L. The method presented intraassay and interassay coefficients of variation for 25(OH)D3 of 3.0% and 3.3%, respectively. Considering that vitamin D fortification or supplementation programs were not implemented at the population level in the study area or as a policy by the Brazilian Ministry of Health and that multiple micronutrient supplements referred by participants included exclusively 25(OH)D3 forms, it is possible to assume that serum 25(OH)D3 concentrations properly depicted vitamin D status. Vitamin D deficiency was defined as 25(OH)D3 <50 nmol/L and insufficiency as 25(OH)D3 <75 nmol/L [28].

According to serum retinol concentrations in the second and third trimesters, also analysed by standardised HPLC (Chromsystems Instruments & Chemicals GmbH, Gräfelfing, Germany), vitamin A insufficiency was defined as <1.05 µmol/L [29]. Vitamin A insufficiency during follow-up was depicted by combining information from both trimesters (never insufficient, insufficient in the 2nd or 3rd trimester, or always insufficient) [12]. C-reactive protein (CRP), a marker of acute inflammation, was measured in the third trimester of pregnancy only, using an ultrasensitive immunoturbidimetric assay (Roche Diagnostics, Basel, Switzerland). A cut-off point of 5 mg/L was used [30]. Blind samples of internal and external quality control were routinely used for each run. Coefficients of variation for these analyses were below 7%.

Seasonality at blood drawn was categorised according to the Amazonian weather as rainy season (November to March) or dry season (April to October). A combination of classifications in the second and third trimesters generated a variable covering seasonality during follow-up (always Amazonian rainy season, change in seasonality, or always Amazonian dry season).

Statistical analyses

The main outcome of interest was 25(OH)D3 concentrations in the third trimester of pregnancy. A conceptual framework with variables at distal, intermediate and proximal levels of determination, according to their influence on the outcome, was conceived based on evidence from the literature [31], to guide variable selection. Predictors of interest included sociodemographic conditions, obstetric history, lifestyle and pregnancy characteristics. First, median and IQR of 25(OH)D3 concentrations were compared across categories of predictors using Mann-Whitney or Kruskal-Wallis tests in unadjusted analysis.

Predictors of 25(OH)D3 concentrations were investigated by fitting simultaneous-quantile regression with bootstrapped standard errors at the 25th, 50th and 75th percentiles of its distribution, with the command sqreg in Stata. Quantile regression does not rely on normality assumptions and the estimation of coefficients at a given percentile is based on the reduction of the median absolute deviation, including a check function with asymmetric weights depending on the percentile. This approach may reveal effects of predictors at different segments of the outcome distribution, which could be especially useful in exploring 25(OH)D3 concentrations. The command sqreg, in particular, allows for the simultaneous estimation of all equations at once, with assessment of correlations between parameters for different percentiles and an estimate of the entire variance-covariance matrix of the estimators by bootstrapping [32]. Variables were included in the model if they were considered conceptually relevant and also following a data-driven approach based on the unadjusted analyses. Statistical significance (P < 0.05) was an additional criterion for retaining variables. With adjustment for the participant’s age and gestational age, coefficients and 95% confidence intervals (95%CI) for the association of each predictor with 25(OH)D3 concentrations in the third trimester were estimated at the 25th, 50th and 75th percentiles of the distribution. Differences in coefficients between percentiles of the 25(OH)D3 distribution were tested with interquantile regressions, by using the command iqreg (75th vs. 25th, 50th vs. 25th, and 75th vs. 50th percentiles).

A set of sensitivity analyses was performed. We run two alternative models excluding: (i) adolescent pregnant women; and (ii) all participants who reported a history of elevated blood pressure, hypertensive disorders, diabetes mellitus, gestational diabetes, or kidney diseases, considering the potential relationship of such conditions with vitamin D status. We also conceived models with additional adjustment for: (i) the use of multiple micronutrient supplements with vitamin D before the third trimester, although this was not a broad practice among the study population; and (ii) concentrations of CRP, as measured in the third trimester, taking into account that lower concentrations of retinol are associated with higher concentrations of acute phase proteins in inflammatory processes [30].

All analyses were performed using Stata 15.1 (Stata Corp, College Station, TX, USA).

Results

From 860 pregnant women initially screened in primary health care units, 699 were eligible according to inclusion criteria, of whom 41 refused to participate and 71 were not found for invitation and enrolment by the research team. Of 587 pregnant women included in the study, 448 had data on 25(OH)D3 concentrations in the second and third trimesters (Fig. 1). Among participants included in the analysis (n = 448), 29.2% had ≤9 years of schooling, in contrast to 32.4% among those not included (n = 139); 30.6% of those who were successfully followed up were from the lowest third of household wealth index as defined at baseline (P < 0.05). There were no differences regarding age, skin colour, type of household, sun exposure, seasonality at blood drawn, and nutritional status of vitamins D and A at the beginning of pregnancy.

Fig. 1: Flowchart of the recruitment and follow-up of participants in the MINA-Brazil study.
figure 1

Between February 2015 and January 2016,pregnant women were screened while booking an antenatal care appointment in the primary health care units of the urban area of Cruzeiro do Sul, Brazil. Reasons for exclusion and the final number of participants who completed a baseline interview and two clinical evaluations in the second and third trimesters of pregnancy are indicated.

Full size image

Mean age of participants was 25 years (SD 6) and 20% were teenagers; 15.2% self-declared as white, 76.8% as mulatto, 4.2% as black, 2.9% as yellow and 0.9% as indigenous. There was no statistical difference in 25(OH)D3 concentrations among non-white groups. In the third trimester of pregnancy, median concentration of 25(OH)D3 among participants was 79 nmol/L, with the 25th and 75th percentiles of the distribution of 25(OH)D3 at 49 nmol/L and 110 nmol/L, respectively.

Considering the cut-off points at 50 nmol/L and 75 nmol/L, prevalence of vitamin D deficiency and insufficiency in the third trimester of pregnancy was equal to 26% and 47%, respectively. Between the second and third trimesters of pregnancy, the cumulative incidence of vitamin D deficiency was 23%; that of vitamin D insufficiency was 37%.

Median 25(OH)D3 concentrations in the third trimester were lower among pregnant women who were the main providers of family income (P = 0.04; Table 1). While exposure to the Amazonian dry season and vitamin D status ≥75 nmol/L in the second trimester were positively associated with 25(OH)D3 concentrations in the third trimester (P < 0.001; Table 2), sustained vitamin A insufficiency throughout pregnancy was related to lower median 25(OH)D3 (P = 0.04).

Table 1 Median distribution of 25(OH)D3 concentrations in the third trimester of pregnancy according to sociodemographic characteristics of participants from the MINA-Brazil Study.
Full size table
Table 2 Median distribution of 25(OH)D3 concentrations in the third trimester of pregnancy according to obstetric history, lifestyle and pregnancy characteristics of participants from the MINA-Brazil Study.
Full size table

In adjusted simultaneous-quantile regression models (Table 3), we observed that 25(OH)D3 concentrations were significantly lower among pregnant women who were the main providers of family income, particularly at the 50th and 75th percentiles of the outcome distribution (−15 nmol/L, 95%CI −24; −3, and −22 nmol/L, 95%CI −36; −7, respectively), when compared to those who reported having a partner or other family members providing the majority of the income. A negative impact of sun protection practices was limited to pregnant women with 25(OH)D3 concentrations at the 25th percentile (−11 nmol/L, 95%CI −19; −3). As expected, exposure to the Amazonian dry season during follow-up and vitamin D status ≥75 nmol/L in the second trimester were consistently and positively associated with 25(OH)D3 concentrations in the third trimester of pregnancy. Interestingly, sustained vitamin A insufficiency during follow-up was associated with reductions in 25(OH)D3 concentrations (−27 nmol/L, 95%CI −40; −15 at the 25th percentile, and −17 nmol/L, 95%CI −33; −1 at the 50th percentile). Supplemental Fig. 1 illustrates the association of each predictor along the distribution of 25(OH)D3 concentrations in the third trimester of pregnancy. According to interquantile regression, significant differences were noted between coefficients estimated at each of the percentiles, as indicated in Table 3. Therefore, there is evidence that 25(OH)D3 concentrations were differently influenced by these predictors depending on the portion of the 25(OH)D3 distribution under examination.

Table 3 Predictors of 25(OH)D3 concentrations at the 25th, 50th and 75th percentiles of the distribution in the third trimester of pregnancy in the MINA-Brazil Study (n = 448).
Full size table

In sensitivity analysis, there were no statistically significant differences to the final model in analysis excluding teenage participants or participants who reported a history of elevated blood pressure, hypertensive disorders, diabetes mellitus, gestational diabetes, or kidney diseases. Concentrations of 25(OH)D3 were not associated with the use of multiple micronutrient supplements with vitamin D after inclusion of this variable in the simultaneous-quantile regression model, and magnitude and direction of the coefficients related to all other predictors remained unchanged. No significant changes for the association with vitamin A status were observed with additional adjustment for CRP concentrations >5 mg/L.

Discussion

In this prospective study, 25(OH)D3 concentrations in the third trimester of pregnancy were positively associated with exposure to the Amazonian dry season and 25(OH)D3 ≥75 nmol/L in the second trimester, when compared, respectively, to those who were exposed to the Amazonian rainy season during pregnancy and those who had a vitamin D status <75 nmol/L in the second trimester. On the other hand, pregnant woman who reported being the main provider of family income and had insufficient vitamin A status presented lower 25(OH)D3 concentrations in the third trimester when compared to their counterparts.

We observed a median 25(OH)D3 concentration of 79 nmol/L among pregnant women in the present study, conducted at latitude 7°S. Overall, 26% of participants were vitamin D deficient in the third trimester. This figure was somewhat higher in comparison with a Brazilian prospective cohort conducted in Rio de Janeiro (10.2% in the third trimester, latitude 22°S) [33], but similar to a previous estimate from a population-based cohort (n = 7507) in the Netherlands (27% at 20.6 weeks of gestational age, 52°N) [34]. There is a remarkable difference, however, in relation to the occurrence of vitamin D deficiency among only 1% of pregnant women from traditional nomadic and semi-nomadic ethnical groups in Tanzania (2° to 4°S) [35]. Diverse latitudes and lifestyle factors regarding sun exposure habits should be highlighted, as well as differences in Human Development Index (0.761 for Rio de Janeiro; 0.922 for the Netherlands; 0.521 for Tanzania) [33,34,35,36], with the potential impact of inequalities on maternal health and vitamin D status.

In this sense, we noted that 25(OH)D3 concentrations were significantly lower among pregnant women who were the main providers of family income, particularly at the 50th and 75th percentiles of the outcome distribution, when compared to participants whose households were headed by other family members. This factor may be an indicative of greater socioeconomic vulnerability. There is evidence from studies in the Middle-East and Africa that household-heads have lower income and schooling level (below college education) in female-headed households than in male-headed households [37, 38]. While some studies from low and middle-income settings could not establish a relationship of socioeconomic variables with vitamin D status [8, 9, 39,40,41], a Saudi observational study has shown a protective role of higher maternal schooling on vitamin D deficiency in the first trimester among pregnant women in a tertiary antenatal care clinic (OR 0.32, 95%CI 0.12; 0.86) [7].

Importance of the endogenous synthesis of vitamin D through exposure to UVB light is well known. In our study, we observed a positive association of exposure to the Amazonian dry season during follow-up and 25(OH)D3 concentrations. Consistently, in a cross-sectional study with 147 Thai pregnant women from the University Hospital of Siriraj, blood collection during the rainy season was positively associated (OR 2.62, 95%CI 1.18; 5.85) with 25(OH)D <50 nmol/L at delivery [39]. Sun protection practices had a negative impact on 25(OH)D3 for participants whose concentrations were at the 25th percentile of the outcome distribution (49 nmol/L). According to NHANES 2003–2006, participants who reported frequently protecting themselves from sunlight with shade structures and clothing had twice the odds of vitamin D deficiency (OR 2.16, 95%CI 1.41, 3.32 for those who reported staying in the shade, and OR 2.11, 95%CI 1.48; 3.00 for those who reported wearing long sleeves, when compared to their counterparts), whilst use of sunscreen was not associated with vitamin D deficiency [42]. Our findings suggest that proper counselling on sun protection practices may be critical in the antenatal care devised to pregnant women with deficient 25(OH)D3 concentrations (<50 nmol/L).

Our analysis corroborated with evidence on the positive synergistic relationship between vitamins D and A during pregnancy. Analogously to our results, baseline data of a supplementation trial in Tanzania among 1,078 HIV-positive pregnant women between 12 and 27 gestational weeks showed that 25(OH)D <80 nmol/L was 13% lower with each 0.35 μmol/L increase in retinol concentrations [43]. The literature points that these fat-soluble vitamins have an interdependent role in regulating immune mechanisms, besides cellular apoptosis and gene transcription [44, 45]. At the nuclear level, retinoid X receptor can heterodimerise with vitamin D receptor into a complex that controls the expression of several genes involved in the inhibition of proinflammatory transcription factors and synthesis of inflammatory molecules [46]. While physiological changes induce an inflammatory state during pregnancy, the present observational finding supporting such interaction is of special interest given that exacerbated inflammatory responses may lead to gestational complications [14], outlining important roles for vitamins D and A in the third trimester. Of note, these findings were not substantially changed with further adjustment for C-reactive protein. Considering the frequent inadequacies in vitamin A status reported in pregnancy [47, 48], consumption of its dietary sources should be encouraged within safe intake levels, favouring fruits and vegetables high in carotenoids, with potential benefits for 25(OH)D3 concentrations.

There are some limitations to this study. First, losses to follow-up, mainly due to difficulties in contacting participants to attend the evaluations, were more frequent among poorer pregnant women, probably because poorer families move more frequently than the wealthier ones. Poorer areas lack standardisation of street and household numbers, and have more unstable cellular networks. Thus, selection bias cannot be completely ruled out. Given that our study population was from the urban area of Cruzeiro do Sul, caution is needed when extrapolating findings for pregnant women from rural regions, as patterns of work and sun exposure may vary. Although this study does not have detailed information on the consumption of dietary sources of vitamin D, their contribution to vitamin D status is apparently limited [27], especially considering the context of absence of vitamin D fortification or supplementation programs, as it is the case in Brazil and in the study area. In this sense, analysis further controlling for the use of multiple micronutrient supplements with vitamin D could not identify this practice as a significant predictor of 25(OH)D3 concentrations in the present study. Lack of information regarding dosage and frequency of these supplements may also be regarded as a limitation, as well as the use of self-reported questionnaires. Validity parameters for the proposed sun protection score would be desirable, in addition to the measurement of CRP also in the second trimester of pregnancy and other micronutrients and hormones related to vitamin D metabolism, but these were not available for this investigation. The present study has several strengths. We highlight its prospective design, with repeated data on biomarkers collected under high-quality conditions by periodically trained researchers. Moreover, the MINA-Brazil Study allowed the investigation of factors associated with 25(OH)D3 concentrations during pregnancy in a tropical middle-income area. The quantile regression approach added to the understanding of predictors along the distribution of 25(OH)D3 concentrations in the third trimester of pregnancy.

In conclusion, among pregnant women in the Brazilian Amazon, exposure to the dry season and vitamin D status ≥75 nmol/L in the second trimester were positively associated with 25(OH)D3 concentrations in the third trimester. Conversely, being the main provider of family income and presenting with sustained vitamin A insufficiency during pregnancy were associated with decreased 25(OH)D3 concentrations. These findings may contribute with tailoring of prevention and treatment strategies for optimising vitamin D status, taking into consideration the influence of predictors at different 25(OH)D3 concentrations that pregnant women may exhibit during the antenatal period. Counselling on adequate exposure to sunlight and dietary sources of vitamin A, amid better quality antenatal care, may help ensuring a good nutritional status of vitamin D during pregnancy and benefit maternal and child health.