Exploring Associations Between Maternal Vitamin D Binding Protein, Vitamin D, and Offspring Asthma/Recurrent Wheeze

Abstract

Background

Vitamin D binding protein (DBP) transports vitamin D metabolites and regulates vitamin D levels in circulation. Additionally, maternal vitamin D levels during pregnancy play an important role in lung development and childhood asthma occurrence.

Objective

In this post-hoc analysis of a randomized controlled trial, the joint associations of maternal 25-hydroxyvitamin-D (25OHD) and DBP with offspring asthma/recurrent wheeze, for both the full cohort and stratified by maternal asthma status, are analyzed. Additionally, associations between estimated maternal free 25OHD and offspring asthma/recurrent wheeze are investigated.

Methods

518 participants from the Vitamin D Antenatal Asthma Reduction Trial were included in this analysis. The primary outcome was offspring asthma/recurrent wheeze by age 3. Maternal plasma DBP levels were measured for 517 and 516 participants at 10–18 and 32–38 weeks of gestation, respectively. Total 25OHD and DBP levels were compared across haplotypes of the Group-specific Component (GC) gene, which codes for DBP. Logistic regression models estimated the relationships between maternal DBP, total 25OHD, and offspring asthma/recurrent wheeze. In addition, offspring asthma/recurrent wheeze was modeled as a function of estimated maternal free 25OHD.

Results

Maternal DBP levels generally increased as pregnancy progressed. Maternal DBP and total 25OHD levels varied significantly across GC haplotypes. A significant positive interaction effect between maternal DBP and total 25OHD on offspring asthma/recurrent wheeze risk was observed for the subset of mothers without asthma. For the subset of mothers with asthma, we observed a significant negative association between estimated maternal free 25OHD and offspring asthma/recurrent wheeze, surpassing the effects of DBP or total 25OHD individually.

Conclusions

Our study gives insight into the interplay between vitamin D metabolites during pregnancy and their associations with offspring asthma/recurrent wheeze. These results also suggest that maternal free vitamin D during pregnancy may be more biologically relevant than total vitamin D for offspring respiratory health.

Introduction

Asthma is the second most prevalent chronic respiratory disease, affecting approximately 3.6% of the global population [1]. Notably, asthma is the most common chronic respiratory disease among children, with an estimated 14% of children worldwide affected [1]. The risk of developing disease in childhood, adolescence, and adulthood is heavily influenced by prenatal perturbations [2]. This is especially true of asthma, where factors such as maternal asthma can result in increased risk of offspring asthma [3].

Vitamin D plays critical roles for healthy fetal development during pregnancy, including maternal immune system modulation, fetal bone development, and fetal lung development [4]. Yet, there is widespread vitamin D deficiency, especially during pregnancy, determined by the circulating levels of 25-hydroxyvitamin D₃ (25OHD, our primary vitamin D metabolite) [5]. The Vitamin D Antenatal Asthma Reduction Trial (VDAART) investigated whether vitamin D supplementation (4,400 IU vs. 400 IU daily) early in pregnancy (10–18 weeks of gestation) would decrease the risk of offspring asthma/recurrent wheeze and other adverse outcomes such as preeclampsia [6]. Although initial findings were statistically insignificant (OR = 0.73; 95% CI = [0.53, 1.00]) [7], a meta-analysis controlling for baseline 25OHD levels revealed a statistically significant result (OR = 0.43; 95% CI = [0.19, 0.96]) [8, 9]. Although VDAART and other clinical trials have not shown conclusive evidence of the protective effect of vitamin D on childhood asthma [10], their findings have pointed towards its potential preventative role. Therefore, additional data on other molecules involved in vitamin D metabolism and signaling, such as vitamin D binding protein, may lead to a clearer understanding of vitamin D’s role in the development of childhood asthma.

Vitamin D binding protein (DBP) is encoded by the Group-Specific Component (GC) gene and serves to transport various vitamin D metabolites throughout the body, primarily 25OHD [11]. Approximately 85% of circulating vitamin D metabolites are bound, primarily to DBP but also to albumin with much lower affinity. Bound vitamin D metabolites have a longer half-life in circulation [12]. DBP thus acts as a circulating reservoir of vitamin D metabolites, including 25OHD.

DBP is highly polymorphic [12]. Early studies of DBP identified three common variants—GC-1f, GC-1s, and GC-2—based on their different electrophoretic properties [13, 14], followed by studies that mapped these isoforms to specific alleles in the GC gene [15]. Different variants were found to affect binding affinities with vitamin D metabolites, and thus the half-life and total concentrations of vitamin D metabolites [16]. Therefore, it is important to account for genomic variation in GC when investigating DBP and 25OHD.

Given DBP’s importance in vitamin D metabolism, this study aims to investigate the role that DBP may have in modulating the effect of 25OHD levels on offspring asthma/recurrent wheeze risk, while accounting for GC polymorphisms. Furthermore, we investigate the relationship between estimated free 25OHD and offspring asthma/recurrent wheeze.

Methods

VDAART Design and Participants

VDAART is a randomized, double blind, placebo-controlled clinical trial of vitamin D supplementation (4,000 IU plus a multivitamin with 400 IU daily) versus placebo (placebo pill plus a multivitamin with 400 IU daily) in pregnant women, given after baseline measurements were taken, to prevent adverse pregnancy outcomes and “asthma and recurrent wheeze” in their children. Eligible participants were women between the ages of 18 and 39 years with the estimated gestational ages between 10 and 18 weeks (singleton) who had a history of physician-diagnosed asthma or atopy, or a partner (biologic father of the child) with a history of physician-diagnosed asthma or atopy, and no history of smoking or current nicotine product use. Eligible participants were screened per VDAART protocol and enrolled if eligible. A participant flow chart is available in Figure S1. Details of the trial design, protocol, and primary and secondary outcomes have been published [6, 7, 17]. For blood sample collection and processing details and covariate assessment details, see [6, 7]. Details on race/ethnicity definitions and assessment can be found in Litonjua et al. (JAMA 2016), under supplement 2 [7]. VDAART was approved by the Institutional Review Boards of the participating clinical centers and the data coordinating center, with pregnant women signing informed consent at the enrollment visit covering both primary and secondary data analyses. In this study, we included pregnant women enrolled in VDAART who provided any postnatal data, up to the child’s age of 3 years, and who had at least one DBP measurement.

Primary Outcomes

In this study, the primary outcome was offspring asthma/recurrent wheeze incidence by the age of 3 years, using the same definitions as those in VDAART. In statistical modeling, children who did not develop asthma/recurrent wheeze by the age of 3 years were treated as the referent group. Ancillary models that were explored also included total 25OHD and DBP as outcomes.

Measurement of Vitamin D Binding Protein

The levels of vitamin D binding protein (DBP) were measured for maternal plasma samples taken at the first and third trimester of pregnancy (10–18 weeks and 32–38 weeks, denoted by M1 and M2, respectively), where baseline first trimester measurements (M1) were taken before administering the intervention. DBP levels were not measured on cord blood samples. DBP levels were measured by Heartland Assays, Inc. (Ames, Iowa) using Aviva Systems Biology Corporation Human Vitamin D Binding Protein ELISA Kit (Catalog No. OKIA00086). Of the 806 participants in the intent-to-treat cohort of VDAART, 518 whose samples had previously been genotyped were selected for the measurement of DBP levels.

Both the inter- and intra-assay coefficients of variation were less than 10%. No Limit of Detection was determined. The Limit of Quantification was 3.13 ng/mL. The lower and upper limits were 3.13ng/mL and 200 ng/mL, respectively. Samples falling outside of this range were retested.

Estimation of Free Vitamin D Levels

The free hormone hypothesis postulates that the biological activity of a biomolecule depends primarily on its unbound concentration, and there is existing evidence that this applies to vitamin D metabolites [12]. Due to the absence of directly measured free 25OHD levels, estimates of these levels were derived from total 25OHD and DBP levels, following similar calculations in [12]. Specifically, free 25OHD levels were estimated by the ratio of total 25OHD to DBP levels.

$$\mathit{\text{free}}250HD=\frac{\mathit{\text{total}}250HD}{DBP}$$ (1)

GC Diplotype

The Illumina Human OmniExpressExome bead chip was used for genomic sequencing of maternal blood samples. Genotypes at rs7041 and rs4588, located in exonic regions of the GC gene, were used to infer diplotypes corresponding to combinations of DBP variants GC-1f, GC-1s, and GC-2. Haplotype and diplotype definitions follow Powe et al. [18] and are detailed in Table S1. GC diplotype was included as a categorical covariate in all models analyzing DBP levels, 25OHD levels, or asthma/recurrent wheeze incidence. Those with unknown GC diplotype, due to incomplete genotype data, were labeled “Unknown” for statistical modeling.

Additional Variables

The covariates included in the statistical modeling were split into a pre-partum and post-partum set. The pre-partum covariate set included maternal age (years), maternal asthma status, self-reported maternal race/ethnicity, maternal baseline BMI, clinical site, maternal baseline total 25OHD (for models with total 25OHD at M2), season of sampling (for models with 25OHD), gestational age of sampling (weeks; for models with DBP), and treatment group; these corresponded to models whose outcome derived from pre-partum samples (i.e., maternal samples). The post-partum covariate set included gestational age at delivery (weeks) and offspring sex. For models whose outcomes come from post-partum samples (i.e., cord blood, offspring), both the pre-partum and post-partum covariate sets were included. Additionally, the percent change from the first to third trimester (M1 to M2) for 25OHD (free and total) and DBP were calculated as $100\times \frac{X_{M2}-X_{M1}}{X_{M1}}$ and investigated for associations with offspring asthma/recurrent wheeze incidence.

Statistical Analysis

For bivariate analyses between offspring asthma/recurrent wheeze incidence and various clinical and demographic variables, we performed Mann-Whitney-Wilcoxon tests for continuous variables and Pearson’s χ² tests for categorical variables, and the Type I error rate was controlled at 0.05. 20 models were estimated (logistic regression for dichotomous outcomes, linear regression for continuous outcomes) for each of three sample sets (full cohort and subset of mothers with or without asthma), for a total of 60 models. Observations with any missing data in model variables were removed prior to model fitting. In addition to the full cohort analysis, we conducted a post-hoc stratified analysis for the subset of mothers with and without asthma, due to maternal asthma being a known risk factor for offspring asthma [19]. We performed model diagnostics to assess linearity of the independent variable to linked-response relationship, multicollinearity, outlier influence, and model goodness-of-fit.

Given the exploratory nature of this analysis, we prioritized sensitivity over strict control of type I error. Therefore, the false discovery rate (FDR) was controlled at 0.2 within each stratum (full cohort, subset of mothers with or without asthma), using Benjamini-Hochberg adjusted p-values [20]. Furthermore, bootstrapped confidence intervals (bCI) were obtained for model coefficients by taking 10,000 samples drawn with replacement from the study cohort, fitting models to these bootstrapped samples, and obtaining a background distribution of coefficient estimates. The 95% bCI ranged from the 2.5th percentile to the 97.5th percentile of the background distribution. R version 4.2.0 was used for all analyses (R Foundation for Statistical Computing). Unadjusted & adjusted p-values and bCIs for independent variables of interest, as well as complete model results for select models, can be found in Tables S2–S3.

Results

Study Cohort Characteristics

Of the 518 VDAART participants included in this study, 137 gave birth to children who developed asthma/recurrent wheeze by age 3 (Table 1). Average third trimester (M2) total 25OHD levels were significantly lower in mothers of children who developed asthma/recurrent wheeze by age 3, compared to those whose children did not (30.8 vs. 34.4 ng/mL; p = 0.0040). Average third trimester (M2) free 25OHD levels were also significantly lower in mothers whose children developed asthma/recurrent wheeze (0.0482 vs. 0.0562; p = 0.0017), though baseline (M1) free 25OHD levels did not differ significantly between groups.

Table 1.

Characteristics of study subjects stratified by offspring asthma/recurrent wheeze status by the age of three years. Differences between groups were assessed using Mann-Whitney-Wilcoxon tests and Pearson’s χ² tests for continuous and categorical variables, respectively.

	Offspring asthma/recurrent wheeze (by age 3)
	No (N=381)	Yes (N=137)	Total (N=518)	p-value
Total 25OHD (ng/mL)(M1)				0.0858
Mean (SD)	24.2 (10.5)	22.5 (9.89)	23.8 (10.3)
Median [Min, Max]	23.2 [5.50, 80.8]	21.9 [4.50, 64.7]	22.8 [4.50, 80.8]
Missing	1 (0.3%)	0 (0%)	1 (0.2%)
Total 25OHD (ng/mL)(M2)				0.0040
Mean (SD)	34.4 (14.0)	30.8 (15.0)	33.5 (14.4)
Median [Min, Max]	33.1 [5.67, 90.1]	28.9 [6.08, 81.3]	31.7 [5.67, 90.1]
Missing	1 (0.3%)	0 (0%)	1 (0.2%)
Total 25OHD (% change, M1 to M2)				0.3129
Mean (SD)	63.3 (96.3)	51.3 (81.0)	60.1 (92.6)
Median [Min, Max]	38.3 [−77.9, 586]	37.5 [−58.0, 479]	37.9 [−77.9, 586]
Missing	2 (0.5%)	0 (0%)	2 (0.4%)
DBP (μg/mL)(M1)				0.5718
Mean (SD)	518 (129)	524 (129)	520 (129)
Median [Min, Max]	516 [178, 1100]	525 [210, 990]	518 [178, 1100]
Missing	1 (0.3%)	0 (0%)	1 (0.2%)
DBP (μg/mL)(M2)				0.3474
Mean (SD)	667 (205)	680 (208)	670 (206)
Median [Min, Max]	626 [106, 1620]	637 [295, 1750]	630 [106, 1750]
Missing	1 (0.3%)	1 (0.7%)	2 (0.4%)
DBP (% change, M1 to M2)				0.7975
Mean (SD)	32.5 (39.7)	33.9 (40.4)	32.9 (39.8)
Median [Min, Max]	24.8 [−78.6, 205]	24.3 [−34.6, 245]	24.5 [−78.6, 245]
Missing	2 (0.5%)	1 (0.7%)	3 (0.6%)
Estimated Free 25OHD (M1)				0.0530
Mean (SD)	0.0499 (0.0260)	0.0447 (0.0212)	0.0485 (0.0249)
Median [Min, Max]	0.0451 [0.00767, 0.204]	0.0410 [0.00899, 0.123]	0.0442 [0.00767, 0.204]
Missing	2 (0.5%)	0 (0%)	2 (0.4%)
Estimated Free 25OHD (M2)				0.0017
Mean (SD)	0.0562 (0.0282)	0.0482 (0.0258)	0.0541 (0.0278)
Median [Min, Max]	0.0530 [0.00957, 0.172]	0.0426 [0.00818, 0.138]	0.0506 [0.00818, 0.172]
Missing	2 (0.5%)	1 (0.7%)	3 (0.6%)
Free 25OHD (% change, M1 to M2)				0.1813
Mean (SD)	32.9 (88.4)	22.4 (72.9)	30.1 (84.7)
Median [Min, Max]	11.0 [−80.7, 643]	1.42 [−82.2, 273]	8.82 [−82.2, 643]
Missing	4 (1.0%)	1 (0.7%)	5 (1.0%)
Maternal Age				0.0017
Mean (SD)	28.2 (5.53)	26.5 (5.71)	27.8 (5.63)
Median [Min, Max]	28.5 [18.0, 39.4]	25.9 [18.0, 39.5]	27.9 [18.0, 39.5]
Maternal BMI (M1)				0.4931
Mean (SD)	28.6 (7.41)	29.2 (7.71)	28.8 (7.49)
Median [Min, Max]	27.2 [16.3, 56.4]	27.5 [17.8, 54.4]	27.2 [16.3, 56.4]
Missing	14 (3.7%)	4 (2.9%)	18 (3.5%)
Maternal Asthma Status				0.0010
No	244 (64.0%)	65 (47.4%)	309 (59.7%)
Yes	137 (36.0%)	72 (52.6%)	209 (40.3%)
Maternal Race/Ethnicity				0.0116
Caucasian (Non-Hispanic)	121 (31.8%)	31 (22.6%)	152 (29.3%)
African American (Hispanic and Non-Hispanic)	134 (35.2%)	70 (51.1%)	204 (39.4%)
Caucasian (Hispanic)	56 (14.7%)	20 (14.6%)	76 (14.7%)
Other	60 (15.7%)	12 (8.8%)	72 (13.9%)
Missing	10 (2.6%)	4 (2.9%)	14 (2.7%)
Site				0.0028
Boston	97 (25.5%)	43 (31.4%)	140 (27.0%)
San Diego	166 (43.6%)	37 (27.0%)	203 (39.2%)
St. Louis	118 (31.0%)	57 (41.6%)	175 (33.8%)
Treatment				0.1542
Vitamin D Supplementation	188 (49.3%)	78 (56.9%)	266 (51.4%)
Placebo	193 (50.7%)	59 (43.1%)	252 (48.6%)
Child Sex				0.0005
Female	201 (52.8%)	48 (35.0%)	249 (48.1%)
Male	180 (47.2%)	89 (65.0%)	269 (51.9%)

DBP=vitamin D binding protein; 25OHD=25-hydroxyvitamin-D; M1=first trimester of pregnancy; M2=third trimester of pregnancy; BMI=body mass index.

The maternal asthma rate for the mothers of children who developed asthma/recurrent wheeze was 52.6%, compared to a rate of 36.0% for mothers of children who did not develop asthma/recurrent wheeze (p = 0.0010). Average maternal baseline total 25OHD levels were 21.55 ng/mL (SD=10.33) and 22.4 ng/mL (SD=10.19) for the subsets of mothers with and without asthma, respectively. There was a significantly higher rate of male offspring among those who developed asthma/recurrent wheeze (65% vs. 47.2%; p = 0.0005). While the distribution of BMI did not significantly differ by offspring asthma/recurrent wheeze status for the full cohort, we saw a statistically significant difference when considering the subset of mothers with asthma (Figure S2). This difference did not hold for the subset of mothers without asthma. Incidence rates for quartiles of DBP and 25OHD can be found in Table S4. Baseline characteristics were similar between those with and without DBP measurements, except for baseline 25OHD, maternal age, and study site (Table S5).

Maternal DBP Levels Increase During Pregnancy

To explore the relationship between DBP and offspring asthma/recurrent wheeze, DBP levels were measured for 517 and 516 maternal blood samples taken at baseline (M1, pre-intervention) and the third (M2, post-intervention) trimester, respectively, with 515 participants having samples at both timepoints (Table 1). Maternal DBP levels did not differ significantly by offspring asthma/recurrent wheeze status or treatment group. Mean DBP levels increased from 520 μg/mL to 670 μg/mL between the first and third trimesters. 85% of mothers showed an increase in DBP levels, with an average relative change of 41%, while only 15% of mothers showed a slight decrease in DBP (Figure S3). DBP increase vs. decrease status between trimesters was not significantly associated with vitamin D levels or treatment group.

Total 25OHD Varies by GC Diplotype

We performed bivariate analyses between GC diplotypes and various clinical and demographic variables, finding significant differences across diplotypes for total 25OHD levels (M1, M2, % change), DBP levels (M2, % change), free 25OHD levels (M1, % change), self-reported maternal race/ethnicity, and collection site (Table S6). Baseline levels of DBP were similar across GC diplotypes (Figure 1). However, we observed significant differences in DBP levels at M2 and in DBP % change from M1 to M2 across GC diplotypes, with the 1s/1s group showing the highest levels and largest % change.

Figure 1.

Boxplots of maternal vitamin D binding protein (DBP) levels across Group-specific Component (GC) diplotypes. P-values come from a Mann-Whitney-Wilcoxon test. Observations with unknown GC gene diplotype were removed for the generation of this plot. Sample sizes for each GC diplotype group are displayed above the corresponding box. M1=first trimester of pregnancy; M2=third trimester of pregnancy; p=p-value.

Maternal total 25OHD levels also varied significantly across GC diplotypes. The 1f/1f diplotype showed the lowest baseline total 25OHD levels, while the 1s/1s diplotype showed the highest (Figure 2). This pattern across diplotypes was maintained for third trimester measurements, though differences were attenuated in the vitamin D supplement group. While all diplotypes showed similar % changes in total 25OHD from M1 to M2 in the placebo group, the supplement group showed significant differences across GC diplotypes in % changes, with the 1f/1f, 1f/2, and 2/2 groups showing the greatest increases (Figure 2).

Figure 2.

Boxplots of maternal total 25-hydroxyvitamin-D (25OHD) levels across Group-specific Component (GC) diplotypes. For M2 and percent change measurements, observations were stratified by treatment group. P-values come from a Mann-Whitney-Wilcoxon test. Observations with unknown GC gene diplotype were removed for the generation of this plot. Sample sizes for each GC diplotype group are displayed above the corresponding box. M1=first trimester of pregnancy; M2=third trimester of pregnancy; p=p-value.

Maternal DBP and 25OHD Levels Show Joint Effect on Offspring Asthma/Recurrent Wheeze Risk

Due to DBP’s important role in vitamin D metabolism, we investigated whether DBP modulates the relationship between 25OHD and asthma/recurrent wheeze. We fit a logistic regression model of offspring asthma/recurrent wheeze incidence at age 3, including an interaction term between DBP and 25OHD at M2, in the full cohort and separately within each stratum of maternal asthma status. The interaction term between DBP and 25OHD levels at M2 was statistically significant in the subset of mothers without asthma (p_adj = 0.1913), but not in the full cohort (p_adj = 0.5651) or the subset of mothers with asthma (p_adj = 0.8580). For the subset of mothers without asthma, the odds ratio per 16 ng/mL increase in total 25OHD at M2 (which equals the average increase in 25OHD from M1 to M2 for those in the supplementation group) at the 1^st quartile (544 μg/mL), 2^nd quartile (630 μg/mL), and 3^rd quartile (745 μg/mL) of DBP at M2 was 0.8378, 1.0437, and 1.4003, respectively.

The interaction of 25OHD and DBP with offspring asthma/recurrent wheeze risk for the subset of mothers without asthma at M2 is visualized in Figure 3. For lower levels of DBP, there was an increased risk of offspring asthma/recurrent wheeze with decreasing 25OHD levels. On the other hand, for higher DBP levels, an increase in 25OHD was associated with increased asthma/recurrent wheeze risk. This reveals a J-shaped response curve of offspring asthma/recurrent wheeze risk when considering the joint change in DBP and 25OHD, with offspring asthma risk lowest near the mid ranges of each variable and highest in the extremes (especially upper extreme). Results from a post-hoc power analysis reveal adequate power for this model (0.75 power for the observed effect size and standard error), while the other interaction models have less power given the observed standard errors (Figure S4). Model results for the full and stratified analyses are given in Table S3.

Figure 3.

Predicted probabilities of offspring asthma/recurrent wheeze for 25-hydroxyvitamin-D (25OHD) × vitamin D binding protein (DBP) interaction model (subset of mothers without asthma at third trimester (M2)). For the subset of mothers without asthma, the 25OHD × DBP interaction model results were used to calculate predicted probabilities of offspring asthma/recurrent wheeze incidence. Predicted probabilities (Pred. probs.) were computed across a range of 25OHD and DBP values (bounded by the observed ranges for these variables in the subset of mothers without asthma). Continuous covariate values were fixed to the sample medians (maternal age (years) = 28.2, BMI = 26.9, gestational age at delivery (weeks) = 39.4, gestational age at M2 (weeks) = 34, baseline total 25OHD (ng/mL) = 22.45). Categorical variables were fixed to be the following: Maternal race/ethnicity = Caucasian (non-Hispanic), Site = Boston, Child gender = female, Treatment = placebo, Season of measurement at M2 = summer, GC diplotype = 1f/1f.

Estimated Maternal Free 25OHD Levels Show Stronger Association with Offspring Asthma than Total 25OHD

It is hypothesized that the free, rather than bound, levels of 25OHD are more biologically relevant. Due to the absence of directly measured free 25OHD levels, these were estimated using existing maternal DBP and total 25OHD measurements (Methods). For the full cohort and separately for the subsets of mothers with and without asthma, we fit logistic regression models of offspring asthma/recurrent wheeze incidence by age 3, with maternal free 25OHD levels (M1, M2, or % change) as our main independent variable of interest. These were compared to models of offspring asthma/recurrent wheeze with either total 25OHD or DBP as the main independent variable. Complete model results are given in Table S3.

For the full cohort and the subset of mothers without asthma, there was no significant association between free 25OHD (M1, M2, or % change) and offspring asthma/wheeze (p_adj > 0.76 for each model, Table S2). For the subset of mothers with asthma, there were significant associations for free 25OHD at M2 (OR_1SD = 0.57336, p_adj = 0.14989) and % change of free 25OHD (OR_1SD = 0.37865, p_adj = 0.02526), but the association for free 25OHD at M1 was non-significant (OR_1SD = 1.20189, p_adj = 0.75727), where OR_1SD represents the odds ratio per 1-standard-deviation increase of the independent variable (Table S2). Free 25OHD at M2 and % change of free 25OHD show stronger associations with offspring asthma/recurrent wheeze risk than either total 25OHD or DBP individually (Figure 4).

Figure 4.

Odds ratios with confidence intervals for logistic regression models of offspring asthma/recurrent wheeze (subset of mothers with asthma). To facilitate comparison across models, odds ratios correspond to a 1-standard-deviation increase in the explanatory variable. Significant odds ratios are in red. DBP=vitamin D binding protein; 25OHD=25-hydroxyvitamin-D; SD=standard deviation; y.o.=years old.

Logistic regression model diagnostics indicated approximate linearity between independent variables of interest and logit-transformed response, no substantial collinearity of independent variables, no unduly influential observations, and significantly better goodness-of-fit compared to null models.

Discussion

This exploratory study focused on newly measured maternal DBP levels of plasma samples from VDAART participants and their associations with 25OHD levels and offspring asthma/recurrent wheeze risk. We observed significant differences in maternal total 25OHD levels at M1 and M2 across GC diplotypes, along with significant differences in % change of total 25OHD across diplotypes for the vitamin D supplement group only. We also investigated the interactive effect between maternal total 25OHD and DBP on offspring asthma/recurrent wheeze risk by the age of 3 years. For the subset of mothers without asthma, this interaction effect was statistically significant; at higher levels of DBP, 25OHD levels were associated with an increased risk of offspring asthma/recurrent wheeze by 3 years. We also observed a stronger association between estimated free 25OHD and offspring asthma/recurrent wheeze incidence than either total 25OHD or DBP individually.

Maternal DBP levels did not differ significantly between treatment groups. However, DBP levels did differ between timepoints. We observed a general increase in DBP levels from the first to third trimester, although 15% of mothers showed a relatively small decrease. It is hypothesized that the measured increase or decrease in DBP levels is primarily due to the gestational age at measurements. DBP levels naturally increase during pregnancy, likely due to increased estrogen levels [21]. During pregnancy, DBP concentrations reach a maximum around week 28 of gestation [22]. Furthermore, mothers with an observed decrease in DBP had significantly higher DBP levels at M1 and a later gestational age at M1. Therefore, the variability in DBP increase vs. decrease is most likely due to the timing of an individual’s peak DBP level with respect to their gestational age at measurements M1 and M2.

There are three common variants of DBP (GC-1f, GC-1s, and GC-2) defined by combinations of alleles (i.e., haplotypes) at two loci (rs4588 and rs7041) in the GC gene [18]. The combination of haplotypes from a pair of chromosomes defines a diplotype. Earlier studies have suggested that GC haplotype can influence DBP levels, DBP-25OHD binding affinities, and 25OHD levels [12, 16, 18]. We compared DBP and 25OHD levels across GC diplotypes and found that DBP levels at M2 and % change in DBP significantly differ by GC diplotype, with the 1s/1s group showing the largest DBP levels at M2 and the largest % change in DBP, which corroborates previous studies [18]. Furthermore, we observed a significant difference in maternal total 25OHD levels across GC diplotypes, with the 1f/1f and 2/2 groups showing the smallest baseline levels, and the 1s/1s group showing the largest. This closely aligns with results from other studies of non-pregnant and pediatric populations [23, 24]. Additionally, our results show a significant difference in total 25OHD % change for the vitamin D supplement group only, with the 1f/1f group showing the largest increase. This suggests that vitamin D supplementation may be most beneficial, in terms of producing a larger increase in 25OHD levels, for individuals with the 1f/1f diplotype. However, current results should be validated in other cohorts before GC genotypes and DBP levels can begin to be used to inform vitamin D supplementation strategies during pregnancy.

In the subset of mothers without asthma, we observed a significant positive interaction effect between maternal DBP and 25OHD (M2) on offspring asthma/recurrent wheeze risk. This implies that for higher DBP levels (above 830 μg/mL), an increase in 25OHD is associated with an increase in asthma/recurrent wheeze risk. The reason behind this unintuitive result can be statistical or biological. While we can safely rule out sampling bias, as the bCI for this interaction term excludes zero, the potential biological mechanisms underlying this finding remain unclear.

We next investigated the association of estimated free 25OHD levels with offspring asthma/recurrent wheeze incidence. In our analysis, free 25OHD was approximated as the ratio of total 25OHD levels to DBP levels. Out of necessity, this is a more simplistic estimation than in [12], which considers albumin levels and binding affinities. Albumin measurements did not exist for our samples; furthermore, 25OHD-to-DBP binding affinities vary between GC genotypes and populations [12]. As our simplified approximation can potentially overestimate the free 25OHD by not accounting for the smaller portion of 25OHD bound to albumin, future studies should aim to refine these results by using either the more accurate approximation of free 25OHD or direct free 25OHD measurements. For the subset of mothers with asthma, we found that higher levels of free 25OHD in the third trimester, along with larger % changes in free 25OHD from the first to third trimesters, are significantly associated with reduced offspring asthma/recurrent wheeze risk. These models included treatment group as a covariate. Importantly, these effects were stronger than the individual effects of total 25OHD or DBP. This could suggest that maternal free 25OHD levels may be more biologically relevant to offspring asthma/recurrent wheeze etiology when compared to total 25OHD. These models also revealed male offspring to be at increased risk of asthma/recurrent wheeze. This corroborates previous studies that have investigated sex differences for pediatric asthma [25].

A primary limitation of this study is the unavailability of similar mother-child cohorts to VDAART with existing DBP measurements in addition to vitamin D levels, which precludes us from performing replication studies to assess the generalization of our results to other populations. Additionally, DBP measurements were not obtained for cord blood samples. Another limitation is the absence of directly measured free vitamin D. As mentioned previously, our estimation of free vitamin D is necessarily simpler than other estimation methods [26]. Additionally, this study explored multiple associations. Given the exploratory nature of this work, a less stringent FDR threshold of 0.2 was used, to maintain statistical power and decrease the risk of Type II errors [27]. As can be seen in Table S2, some, but not all, of our results are robust to stricter thresholds such as FDR < 0.1 or FDR < 0.05. Finally, not all clinical variables in our data were included in our models. While the inclusion of certain omitted variables during model diagnostics showed significant associations with the response (e.g., gestational diabetes for asthma/recurrent wheeze as response, C-reactive protein levels for DBP levels as response), this did not affect inference on the independent variables of interest. Given these limitations, findings should be interpreted with caution, and future confirmatory studies are needed to validate these associations.

In conclusion, we observed evidence that GC diplotype affects DBP and total 25OHD levels and associates with differential response to vitamin D supplementation. We also found a significantly positive interaction effect between DBP and total 25OHD on offspring asthma/recurrent wheeze incidence for the subset of mothers without asthma. Additionally, for the subset of mothers with asthma, our free 25OHD estimates show a stronger association with offspring asthma/recurrent wheeze incidence than total 25OHD or DBP individually. These observations highlight that DBP and free vitamin D can provide important missing links for explaining the relationship between maternal vitamin D and offspring asthma.

Abbreviations

25OHD

25-hydroxyvitamin-D

bCI

bootstrap confidence interval

DBP

vitamin D binding protein

FDR

false discovery rate

GC

Group-specific Component

OR

odds ratio

SD

standard deviation

VDAART

Vitamin D Antenatal Asthma Reduction Trial

Data Sharing

Data described in the manuscript, code book, and analytic code will be made available upon request pending application and approval.