Vitamin D supplementation in pregnancy and lactation to promote infant growth

BACKGROUND Causes of early infant growth restriction remain incompletely understood. Where vitamin D deficiency is common, vitamin D supplementation during pregnancy and lactation may improve fetal-infant growth and other birth outcomes. METHODS We conducted a randomized, double-blind, placebo-controlled trial of maternal vitamin D supplementation from 17-24 weeks gestation until birth or 6 months postpartum. Participants were randomly allocated to five vitamin D and/or placebo supplementation groups: (A) 0 IU/week, (B) 4200 IU/week, (C) 16800 IU/week, or (D) 28000 IU/week in pregnancy, all with 0 IU/week postpartum; or, (E) 28000 IU/week in prenatal and postpartum periods. The primary outcome was length-for-age z-score at one year of age according to World Health Organization child growth standards. RESULTS Among 1164 infants assessed at one year of age (90% of 1300 pregnancies), there were no differences across groups in length-for-age z-scores (mean ±standard deviation): A: -0.93 ±1.05, B: -1.11 ±1.12, C: -0.97 ±0.97, D: -1.06 ±1.07, E: -0.94 ±1.00 (p=0.23). Groups were similar with respect to other anthropometric measures, birth outcomes, and morbidity. Vitamin D had dose- dependent effects on maternal and infant serum 25-hydroxyvitamin D and calcium, maternal urinary calcium excretion, and maternal parathyroid hormone concentrations. No clinical adverse events were attributed to the vitamin D intervention. CONCLUSIONS In a population with widespread prenatal vitamin D deficiency and fetal/infant growth restriction, maternal vitamin D supplementation from mid-pregnancy until birth or 6 months postpartum does not influence fetal or infant growth, and has no beneficial or harmful effects on numerous other birth and infant outcomes.

urinary calcium excretion, and maternal parathyroid hormone concentrations. No clinical adverse events were attributed to the vitamin D intervention.

CONCLUSIONS
In a population with widespread prenatal vitamin D deficiency and fetal/infant growth restriction, maternal vitamin D supplementation from mid-pregnancy until birth or 6 months postpartum does not influence fetal or infant growth, and has no beneficial or harmful effects on numerous other birth and infant outcomes. Vitamin D may influence fetal and postnatal growth through effects on calcium absorption 6 , parathyroid hormone (PTH) expression 7 , phosphate metabolism 8 , growth plate function 9,10 , and possible regulation of the insulin-like growth factor axis 11 . Meta-analyses of observational studies 12 and clinical trials 13 have suggested a possible beneficial effect of vitamin D on fetal growth, but most previous trials had important methodological limitations 13 . In a previous small trial in Bangladesh, we found that early postnatal linear growth was accelerated in infants born to vitamin D-supplemented mothers 14 .
In Bangladesh, 30% of newborns are SGA 4 and 36% of children under 5-years of age are stunted (height-for-age z-score <-2) 15 . Vitamin D deficiency is common in Bangladeshi women of reproductive age 16 . In this Maternal Vitamin D for Infant Growth (MDIG) trial, we aimed to evaluate the dose-dependent effects of prenatal vitamin D supplementation, with and without postpartum supplementation, on infant growth and other maternal, newborn and infant outcomes in Dhaka, Bangladesh.

Trial Design and Oversight
MDIG was a randomized, double-blind (participants and study personnel), placebo-controlled, dose-ranging parallel five-arm trial of maternal vitamin D supplementation 17 . The protocol and statistical analysis plan are available at NEJM.org. The study was overseen by a trial steering committee and an independent data and safety monitoring board. The protocol was approved by research ethics committees at The Hospital for Sick Children (Toronto, Canada; REB1000039072) and icddr,b (PR-13055). All authors contributed to finalizing the manuscript and attest to the completeness and accuracy of analyses and adherence to the protocol. The trial funder had no role in trial design, data collection, analysis, or interpretation of the results.

Participants
Generally healthy women between 17 and 24 weeks of gestation were enrolled after providing written informed consent between March 2014 and September 2015 at the Maternal and Child Health Training Institute (MCHTI), a public hospital in Dhaka, Bangladesh. Inclusion and exclusion criteria are shown in Table S1 in the Supplementary Appendix.

Interventions
Participants were randomly allocated at enrolment to one of five groups: 0 IU/week vitamin D from enrolment until delivery and 0 IU/week from 1 to 26 weeks postpartum (0;0 or 'placebo' group); 4200 IU/week prenatal and 0 IU/week postpartum (4200;0); 16800 IU/week prenatal and 0 IU/week postpartum (16800;0); 28000 IU/ prenatal and 0 IU/week postpartum (28000;0); or, 28000 IU/week prenatal and postpartum (28000;28000). A computer-generated simple randomization scheme was created independently by the trial statistician (A.R.W.). The master list linking unique participant identifiers to supplementation groups was held by the supplement manufacturer and not accessed by any study personnel until final unmasking.
Allocation concealment was ensured by using pre-labeled sequentially-numbered and otherwise identical supplement vials assigned to participants according to the allocation sequence. Oral vitamin D3 and placebo tablets were manufactured by Toronto Institute for Pharmaceutical Technology (Toronto, Canada). Vitamin D content of each batch of tablets was verified in product testing 17 . Tablets of varying doses were identical in appearance and taste.
Tablets were routinely administered under direct observation by study personnel; however, up to 4 consecutive doses could be unobserved when participants were unavailable for scheduled visits. Missed doses were administered up to 7 days late. Calcium (500 mg/day), iron (66 mg/day), and folic acid (350 μg/day) were provided to all participants throughout the intervention phase 17 . A mid-trial audit of self-reported calcium and iron-folic acid adherence revealed that >85% participants reported >85% adherence. If a participant reported non-study vitamin D or calcium supplement use for >1 week, study supplements were suspended until non-study supplement use was discontinued. Supplementation was discontinued in participants with confirmed hypercalcemia (see definition below), fetal or infant death, or a new condition or medication that could alter vitamin D metabolism.

Assessments
Study personnel contacted participants at weekly intervals from enrolment until 26 weeks postpartum, then every three months. Visits were conducted in the home or clinic, and included standardized questionnaires, point-of-care tests, anthropometry, and specimen collection ( Table S2 in  were measured according to standardized procedures by trained personnel, as previously described 17 and adapted from Intergrowth-21 st protocols 18 . Length, weight, HC, UAL, and RKL were measured at birth. Length, weight, and HC were measured at a randomly timed visit during the first 2 months, and then at 3, 6, 9 and 12 months of age. MUAC, UAL, and RKL were measured at 3, 6, and 12 months. Each parameter was measured independently by two study personnel; paired measurements were compared and repeated if the difference exceeded 7 mm for length, 5 mm for HC, UAL, MUAC, or RKL, and 50 g for weight. Means of the final pair of values were used in analysis. Missing, outlying or implausible values were identified and interrogated (Method 1 in the Supplementary Appendix) 19 . There was high inter-rater reliability and few measurements were dropped due to implausibility or temporal inconsistencies ( Table S3 in

Outcomes
The primary outcome was length-for-age z-score (LAZ) at one year (364-420 days) of age.
Secondary anthropometric outcomes included weight-for-age z-score, WFL z-score, BMI-for-age z-score, HC-for-age z-score, MUAC-for-age z-score, UAL, and RKL ( Table S4 in Supplementary Appendix). Stunting was defined as LAZ<-2. For measurements within 48 hours of birth, SGA was weight-for-age<10 th percentile (using Intergrowth-21 st newborn standards 20 ) and low birth weight (LBW) was <2500 g. Preterm birth was gestational age (GA) at birth <37 weeks based on last menstrual period and 2 nd trimester ultrasound (Table S1 in the Supplementary Appendix).
Vitamin D status was based on serum 25(OH)D 23 ; deficiency was defined as 25(OH)D<30 nmol/L 24 . The C3-epimer fraction was included in sensitivity analyses (Method 2 in the Supplementary Appendix). The primary safety measure was maternal total sCa measured at baseline, 30 weeks of gestation, delivery, 3-months and 6-months postpartum, or when hospitalized due to illness if feasible. Possible hypercalcemia was any sCa>2.60 mmol/L and confirmed hypercalcemia (primary safety outcome) was defined when sCa>2.60 mmol/L on a repeat specimen or a single sCa>2.80 mmol/L. Secondary safety indicators included infant sCa at 3 and 6 months of age and maternal urinary Ca:Cr at delivery. Maternal possible hypercalciuria was a single uCa:Cr>1 mmol/mmol. Participants with uCa:Cr>1 on two consecutive specimens (confirmed hypercalciuria) and/or symptoms of renal colic underwent abdominal ultrasound for uro-or nephrolithiasis. Infant uCa:Cr was measured at 6 months of age. Secondary clinical outcomes included gestational hypertension, delivery characteristics, stillbirth, congenital anomalies, rickets, clinical encounters, hospitalization and death ( Table S4 in the Supplementary Appendix).

Statistical Analysis
The primary analysis was a complete-case intention-to-treat analysis. Analysis of variance (ANOVA) was performed to compare LAZ at one year of age across all groups. To estimate the effect of prenatal vitamin D (IU/week), five pairwise comparisons were conducted using t-tests: 4200;0 versus placebo, 16800;0 versus placebo, 16800;0 versus 4200;0, 28000;0 versus placebo, and 28000;0 versus 16800;0. Statistical significance was tested at an overall alpha=0.05 (two-sided), applying the Holm test for multiple comparisons 25 . Sample size determination conservatively assumed that if each between-group comparison had a two-sided alpha=0.01 and 90% power, 220 analyzable participants per group would enable detection of a between-group difference in LAZ of at least 0.40 14 . To accommodate 15% attrition, we aimed to enroll 260 pregnant women in each arm.
The effect of postpartum vitamin D on LAZ at one year of age was assessed by the pairwise comparison of 28000;28000 versus 28000;0 (two-sided alpha=0.05) using a t-test. Secondary outcomes were compared across groups using ANOVA for continuous normally-distributed variables and Kruskal-Wallis tests for skewed distributions; Chi-square and Fischer's exact tests were used for categorical variables. Zero-inflated negative binomial models were used to compare incidence rates of clinical encounters, hospitalizations, and other adverse events.
Where a global test was significant at p<0.05, post-hoc pairwise comparisons were performed, applying the Holm test for multiplicity 25

Trial Population
1,300 pregnant women were enrolled and randomized (Figure 1). Baseline characteristics including vitamin D status were similar across groups ( Table 1; Table S5 and S6 in the Supplementary Appendix). Overall, 64% of women were vitamin D deficient. Groups did not differ by breast-feeding patterns or reported infant supplement use (Tables S7 and S8 in the Supplementary Appendix). Participants in primary analyses had higher average asset indices than those excluded but were otherwise similar ( Table S9 in  Across all groups, ≥90% of scheduled doses were received by >90% of women in the prenatal period and >80% of women in postpartum periods (Table S10 in the Supplementary Appendix).

Infant Growth Outcomes
Infant follow-up at 1 year of age was completed for 90% of pregnancies and 94% of infants alive at 1 year (Figure 1). Overall, mean LAZ at 1 year was -1.00 (SD 1.04) and the prevalence of stunting was 16%. Prenatal or postpartum maternal vitamin D supplementation did not increase or decrease infant length or other anthropometric outcomes by one year of age ( Table   2; Figures S1-S5 and Tables S11 and S12 in the Supplementary Appendix). The lack of effect of prenatal vitamin D on length was evident from birth ( Table 3) and was supported by sensitivity and stratified analyses (Tables S13-S22 in the Supplementary Appendix). Results of the multiple imputation analysis agreed closely with the complete case analysis (Table S18 in the Supplementary Appendix); therefore, the complete case analysis is shown in Table 2.

Biochemical effects of supplementation
Vitamin D had dose-dependent effects on maternal, cord blood and infant 25(OH)D (Table 3;

Secondary Clinical Outcomes
Delivery characteristics, duration of gestation, preterm birth, SGA, and LBW were similar across groups ( Table 3 and Table S28

DISCUSSION
In a setting of widespread vitamin D deficiency and fetal-infant growth restriction, vitamin D supplementation from mid-pregnancy to delivery or 6-months postpartum is safe but does not influence offspring growth patterns, and has no discernible effects on numerous pregnancy or infant clinical outcomes.
These findings do not support the hypothesis that prenatal vitamin D status in the 2 nd half of pregnancy is a determinant of newborn size, contrary to the conclusions of prior meta-analyses of observational studies 12 and mostly small trials 13 , but consistent with higher-quality trials in settings with lower prevalence of vitamin D deficiency or fetal growth restriction [26][27][28][29] . Our earlier trial in Bangladesh 14 and a study in the United Kingdom 30 found that prenatal vitamin D increased infant linear growth. However, they were small studies (n<135) with postnatal growth as a post-hoc outcome, and the between-group differences may have been due to chance. The present findings are broadly consistent with a meta-analysis of six trials of prenatal daily multiple micronutrient supplements (200 to 400 IU vitamin D) in low-and middle-income countries, which showed no effect on height at 2-8.5 years of age 3 .
There was no clear evidence of other health benefits of improved vitamin D status in the latter half of pregnancy or early infancy. In particular, purported effects of vitamin D on preterm birth 31 were not substantiated, consistent with a recent meta-analysis 13  Also, linear growth faltering was milder in participants than in national surveys 15 suggesting that participants had relatively better baseline health and access to care. For example, participants had high rates of facility deliveries (85%, versus 37% nationally) and C-sections (51%, which is typical of local health facilities but more than double the national average) 15 .
Consistent with WHO recommendations 34 , MDIG trial findings do not support routine vitamin D supplementation in pregnancy or lactation to improve birth outcomes or infant growth, even in communities with endemic vitamin D deficiency and fetal-infant growth restriction.
Supported by the Bill and Melinda Gates Foundation (OPP1066764). Figure 1. Trial flow diagram a Provisional screening refers to the initial eligibility assessment of pregnant women presenting for antenatal care at the Maternal and Child Health Training Institute; individual women may have been provisionally screened more than once during the trial enrolment period. b Detailed screening refers to the complete eligibility assessment supervised by a study physician. c Prenatal; postpartum vitamin D dose. d Participants not formally lost to follow-up but for whom data at the 1-year visit were not obtained.     days.

Figure legends and footnotes
e Post-hoc pairwise comparisons using t-tests showed significant pairwise differences between all groups except both groups who received a prenatal dose of 28000 IU/week, after adjusting for multiple comparisons using the Holm test.