Effect of maternal prenatal and postpartum vitamin D supplementation on offspring bone mass and muscle strength in early childhood: follow-up of a randomized controlled trial

ABSTRACT Background Maternal vitamin D status during pregnancy and lactation is a modifiable factor that may influence offspring musculoskeletal outcomes. However, few randomized trials have tested the effects of prenatal or postpartum vitamin D supplementation on offspring bone and muscle development. Objectives The aim was to examine hypothesized effects of improvements in early-life vitamin D status on childhood musculoskeletal health in Dhaka, Bangladesh. Methods In a previously completed, double-blind, dose-ranging trial, healthy pregnant women (n = 1300) were recruited at 17–24 weeks’ gestation and randomly assigned to a prenatal/postpartum regimen of 0/0, 4200/0, 16,800/0, 28,000/0, or 28,000/28,000 IU cholecalciferol (vitamin D3)/wk until 26 wk postpartum. In this new report, we describe additional follow-up at 4 y of age (n = 642) for longer-term outcomes. Bone mineral content (BMC) and areal bone mineral density (aBMD) were measured by DXA. Grip strength was tested using a hand-held dynamometer. The primary comparison was children of women assigned to 28,000 IU/wk prenatally compared with placebo. Differences are expressed as means and 95% CIs. Results Total-body-less-head (TBLH) BMC, TBLH aBMD, and grip strength were similar in the combined high-dose prenatal (28,000/0 and 28,000/28,000 IU/wk) compared with placebo groups (mean difference [95% CI] = 0.61 g [–10.90, 12.13], 0.0004 g/cm2 [–0.0089, 0.0097], and 0.02 kg [–0.26, 0.31], respectively). In dose-ranging analyses, TBLH BMC and aBMD, whole-body BMC and aBMD, and grip strength in each of the prenatal vitamin D groups were not significantly different from placebo (P > 0.05 for all comparisons). Only head aBMD was greater in children of women assigned to the 28,000/28,000-IU regimen compared with placebo (mean difference [95% CI] = 0.024 g/cm2 [0.0009, 0.047], P = 0.042); the effect was attenuated upon adjustment for child height, weight, and sex (P = 0.11). Conclusions Maternal prenatal, with or without postpartum, vitamin D supplementation does not improve child BMC, aBMD, or grip strength at 4 y of age. The MDIG trial and present follow-up study were registered prospectively at www.clinicaltrials.gov as NCT01924013 and NCT03537443, respectively.


Sample size estimation, calculation of anticipated detectable differences, and power calculation for secondary outcomes
The target sample size was based on the primary objective of detecting a meaningful difference in total-body-less-head (TBLH) bone mineral content (BMC) at 4 years of age between the placebo and combined high-dose (28000 IU/week prenatal with or without 28000 IU/week postpartum) groups. However, we also performed power calculations for selected secondary outcomes. Including allowance for an expected attrition rate of 15%, the desired sample size for our primary outcome was calculated at 140 children from each of the 5 groups in the original MDIG trial, giving an overall target sample size of 700 children. Based on the MDIG participant follow-up registry, we anticipated that about half of the original MDIG trial sample (1298 pregnancies) would be available and willing to consider participation in a follow-up study (i.e., ~650 to 700 participants). Therefore, we aimed to enrol the maximal number of children available from the original cohort, and expected an even distributed across trial arms given the randomized trial design.
For lumbar spine areal BMD (aBMD), we estimated that a SMD=0.36 would be an approximate difference of 0.02 g/cm 2 or 4.8% assuming a mean of 0.46 g/cm 2 and CV of 13% (1). A separate originally planned analysis of lumbar spine aBMD was determined to be unfeasible given the additional time and cost of conducting the DXA scanning in a manner that would generate the necessary data.
O'Callaghan KM et al.  In extension of the between-group differences in our primary analytical approach, secondary analyses were based on the dose-response relationship using supplemental vitamin D intake as the continuous exposure variable. Assuming a 5% risk of a type I error and n=600 (where n=120 from each trial arm), we anticipated having 90% power to detect a minimum SMD=0.12 for every 10,000 IU/week increase in vitamin D intake. Previous research suggested this is a plausible dose-response effect (assuming 10,000 IU/week yields a ~25 nmol/L difference in 25(OH)D): 0.24 SMD increase in BMC at 20 years of age for each 25 nmol/L increase in maternal 25(OH)D was found in an Australian cohort (2), equating to a ~8 g increase in BMC per 10,000 IU/week.
For analyses of lean and fat mass, we considered that we would have 90% power to detect differences of 1.7, 1.5 and 0.6 percentage points, respectively, given distributions at age 4 years (3). In the dose-response analysis based on vitamin D dose as a continuous variable, we expected to be able to detect 0.5 %-point increases in lean and fat mass. O 1 Effect estimates for between-group differences calculated from multivariable linear regression models, with placebo (0;0 IU/week) as the reference group whereby intervention group reflects the vitamin D dose provided in IU/week, represented as a prenatal; postpartum supplementation regimen assigned to the child's mother from randomization (17-24 weeks of gestation) to 6 months postpartum. Regression models included the following selected covariates: maternal age at enrolment, maternal height at enrolment, household asset index, and duration of exclusive breastfeeding. Estimates of the 95% CIs were obtained using a bootstrap procedure with 1000 replications. aBMD, areal bone mineral density; BMC, bone mineral content; TBLH, total-body-less-head. 2 Values represent mean difference for each vitamin D supplementation group compared to placebo.  Table 5: Effect of maternal vitamin D supplementation on offspring total-body-less-head bone mineral content, totalbody-less-head bone mineral density and grip strength at age four years in all maternal vitamin D intervention groups relative to placebo, using multiple imputation by chained equations to address missing values 1 .  1 Effect estimates for between-group differences calculated from linear regression models, with placebo (0;0 IU/week) as the reference group whereby intervention group reflects the vitamin D dose provided in IU/week, represented as a prenatal; postpartum supplementation regimen assigned to the child's mother from randomization (17-24 weeks of gestation) to 6 months postpartum. The imputation models included the main exposure (treatment assignment), the outcome of interest, and auxiliary variables that helped improve the performance of the missing data procedure, including maternal age at enrolment, maternal height at enrolment, household asset index at enrolment, paternal education at enrolment, maternal adherence to the intervention, gestational age at birth, child sex, feeding pattern at 6 months, child anthropometry at 4 years of age, child hemoglobin concentration at 4 years of age, information related to venous blood collection at 4 years of age (successful/not successful), and household smoking status at 4 years of age. aBMD, areal bone mineral density; BMC, bone mineral content; TBLH, total-body-less-head.
O'Callaghan KM et al.  Supplementary Table 6: Effect of maternal vitamin D supplementation on offspring bone mineral content and bone mineral density at age four years relative to placebo among participants whose DXA report showed only negligible or no motion artifact 1 . 0.017 (-0.008, 0.043) 1 No or negligible motion artifact defined as whole body positioned within identified measurement parameters and no presence of additional artifacts within the measurement area (e.g. hand of caregiver/DXA technician). Effect estimates for between-group differences calculated from unadjusted linear regression models, with placebo as the reference group. Estimates of the 95% CIs were obtained using a bootstrap procedure with 1000 replications. aBMD; areal bone mineral density; BMC, bone mineral content; TBLH, total-body-less-head; WB, whole-body.        Table 8: Length-and height-for-age z-scores of girls from birth to 4 years of age, limited to placebo and high-dose prenatal vitamin D supplementation trial arms and stratified by study cohort.    1 Vitamin D deficiency defined as a 25-hydroxyvitamin D concentration <30 nmol/L at enrolment. Effect estimates for between-group differences calculated from unadjusted linear regression models, with placebo as the reference group. Estimates of the 95% CIs were obtained using a bootstrap procedure with 1000 replications. aBMD; areal bone mineral density; BMC, bone mineral content; TBLH, total-body-less-head; WB, whole-body. 2 Values represent mean difference for each vitamin D supplementation group compared to placebo.