Pregnant and non-pregnant women were enrolled at a clinic in Dhaka, Bangladesh (24°N) from July 2009 to February 2010 if they were aged 18 to <35 years, held permanent residence in Dhaka at a fixed address, and planned to stay in Dhaka for at least four months (Figure 1). Reasons for exclusion were a known medical condition, self-reported current use of any dietary supplements containing vitamin D, use of anti-convulsant or anti-mycobacterial medications, severe anemia (hemoglobin concentration <70 g/L), or hypertension at enrollment (systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg on at least two measurements). Pregnant women were excluded if they had major risk factors for preterm delivery (e.g., preterm labor or previous preterm delivery), pregnancy complications or had previously delivered an infant with a congenital anomaly or perinatal death. Non-pregnant women were excluded if they were possibly pregnant (e.g., missed recent menses) or lactating.

All participants in this study received a dose of vitamin D (70,000 IU) at baseline. Primary PK analyses involved participants who did not receive any additional vitamin D throughout follow-up (“single-dose group”). However, to enhance the assessment of 25(OH)D response and safety during the first week of follow-up, an additional cohort of participants who continued to receive weekly vitamin D doses beginning on day 7 (“weekly-dose group”) contributed biochemical data to the present analysis for the first 7 days after the 70,000 IU dose (i.e., up to the time preceding their 2nd dose). Findings related to the effect of weekly dosing will be reported elsewhere. Participants were enrolled in stages according to a design that enabled interim analyses and the testing of supplementation regimens in non-pregnant participants prior to their initiation in pregnant women: non-pregnant participants were enrolled in the summer (July to September 2009); pregnant women who received only a single dose were enrolled during the 30th week of gestation in August-September 2009; and, pregnant participants who received the initial dose followed by weekly doses were enrolled at 27 to <31 completed weeks of gestation in February 2010. The study was reviewed and approved by the Institutional Review Board at The Johns Hopkins Bloomberg School of Public Health and the International Center for Diarrheal Disease Research, Bangladesh (ICDDR,B). All participants gave signed informed consent prior to participation. The trial was registered at ClinicalTrials.gov (NCT00938600).

Vitamin D3 (cholecalciferol) 70,000 IU (1.75 mg) was administered directly by study personnel. The dose was selected to be intermediate between the doses previously studied in the only two rigorous single-dose vitamin D3 pharmacokinetic studies published at the time our study was designed (50,000 10 and 100,000 IU 11 ), thus providing reassurance in terms of probable safety as well as enabling coherent between-study comparisons. The vitamin D3 supplement (Vigantol Oil, Merck KGaA, Germany) was a liquid formulation (20,000 IU D3/mL). The batch of Vigantol Oil used in the study had a concentration of 20,697 IU/mL according to the manufacturer’s certificate of analysis (May, 2009). The stability of the vitamin D3 was established by independent testing of unused Vigantol Oil at the end of the study (June 2010) in the laboratory of Dr. Reinhold Vieth 12 , which revealed a concentration of 19,300 IU/mL (96.5% of the labeled concentration). Participants were advised not to take other vitamin D-containing supplements during the study period, but were permitted to take other micronutrient supplements (including calcium). All pregnant participants were provided standard iron and folic acid supplementation.

Study personnel assessed participants at least weekly. Non-pregnant participants who received only the single dose participated in weekly follow-up for 10 weeks; pregnant women in the single-dose group were assessed at least weekly until delivery, and then at least three times between delivery and discharge from the study at one-month post-partum. Visits involved a checklist of symptoms related to hypo- and hypercalcemia (decreased appetite, weight loss, vomiting, fever or chills, constipation, abdominal pain, excessive thirst, frequent urination, muscle weakness, back, arm, or leg pain, confusion, or depression), blood pressure measurement, and confirmation of fetal viability.

Abnormal urinalyses, hypertension, reported severe symptoms, or persistence of any mild symptomatic complaints (i.e., decreased appetite, weight loss, vomiting, fever or chills, constipation, abdominal pain, excessive thirst, frequent urination, muscle weakness, back, arm, or leg pain, confusion, or depression) for two consecutive visits prompted referral to the study physician for further evaluation. Participants were referred to an antenatal care physician at the maternity clinic for treatment of urinary tract infections, hypertension, or other medical problems that arose. Participants with obstetric complications were transported to a local tertiary-care hospital with advanced neonatal care facilities. All costs of medical and obstetric care were borne by the study.

Participants provided up to six scheduled blood specimens and at least seven urine samples during the 10-week follow-up period beginning on the day of supplement administration (Figure 2). To limit the burden of specimen collection on each individual, yet still enable robust group-level pharmacokinetic and safety analyses, participants were assigned to one of two sampling schedules (A or B) to enhance coverage of the follow-up period (Figure 2). During the first week, specimens were collected at baseline and then additionally on either day 2 or 4 to monitor for possible early transient elevations in serum calcium and to minimize the chance of missing a possible early peak in [25(OH)D]. In the single-dose only groups, blood collection thereafter was scheduled predominantly in the first month because this was when the peak [25(OH)D] 11 and the highest risk of hypercalcemia were anticipated. In pregnant women in the single-dose only group, the specimen collection schedule was continued in the postpartum period if delivery occurred prior to 10 weeks from enrollment. In pregnant participants, venous cord blood samples were collected immediately following delivery of the placenta.

Serum samples (separated from maternal venous and umbilical vein blood) and random spot urine specimens were maintained at +4°C prior to same-day transfer to the laboratory. Sera were frozen at −20 °C. Aliquots for the 25(OH)D assay were shipped at ambient temperature from Dhaka to Toronto (25(OH)D is stable under a range of conditions). Total serum [25(OH)D] was measured with the Diasorin Liaison Total assay in the laboratory of Dr. Reinhold Vieth (Mount Sinai Hospital, Toronto) according to a method previously described 13 . This laboratory participates in and meets the performance targets of the International Vitamin D External Quality Assessment Scheme 14 . Mean within-run coefficient of variation (CV%) was 7.8% (5.8% for specimens with values < 150 nmol/L) and mean between-run CV% was 10.5% (9.0% for specimens <150 nmol/L). Ancillary serum and urine biochemical tests were performed using the AU640 Olympus Autoanalyzer (Olympus Corporation, Japan) at ICDDR,B.

The primary pharmacokinetic (PK) outcome measure was the serum [25(OH)D]; incremental changes from baseline (Δ[25(OH)D]) were calculated as an individual’s absolute [25(OH)D] at each visit minus her baseline [25(OH)D]. The primary safety-related outcome was maternal albumin-adjusted serum calcium concentration ([Ca]), calculated using a conventional formula: [Ca]+(0.02*(40-albumin)). The reference range for albumin-adjusted serum calcium was set at 2.10 – 2.60 mmol/L, the upper limit of which was a conservative threshold relative to those used by: the local laboratory in Dhaka (2.62 mmol/L), the US Institute of Medicine (IOM) 1997 dietary reference intakes (DRIs) for vitamin D (2.75 mmol/L) 15 , and the IOM revised 2011 vitamin D DRIs (2.63 mmol/L) 1 . An albumin-adjusted serum calcium concentration >2.60 mmol/L prompted a repeat measurement on a new specimen as soon as possible. Confirmed hypercalcemia was a priori defined as albumin-adjusted serum calcium concentration > 2.60 mmol/L on both specimens, since hypercalcemia caused by vitamin D intoxication would not be expected to resolve within a few days without intervention.

The urinary calcium:creatinine ratio (ca:cr) was expressed as mmol Ca/mmol Cr, and 1.0 was considered the nominal upper limit of the reference range 16 . Any episode of urinary ca:cr>1.0 mmol/mmol prompted a repeat urine ca:cr measurement within one week. In addition, a ca:cr > 0.85 mmol/mmol that was also a 2-fold or greater increase over the lowest previously observed ratio in the same participant prompted repeat urine assessment. Persistent hypercalciuria was defined as ca:cr > 1.0 mmol/mmol on two consecutive tests, or on two non-consecutive measurements that occurred in the presence of persistent symptoms suggestive of possible hypercalcemia. Persistent hypercalciuria or persistent ca:cr > 0.85 mmol/mmol (under the conditions listed above) were indications for unscheduled measurement of serum calcium.

Continuous outcome variables were described by means, standard deviations (SD), and 95% confidence intervals (95% CI). Non-normally distributed variables (including [25(OH)D]) were described by geometric means with 95% CI’s, medians and interquartile ranges (IQR), and were log-transformed for modeling. In the primary PK analysis, the following model-independent PK parameters were estimated for each individual in the single-dose only groups (N=31): 1) maximum observed [25(OH)D] (Cmax); 2) maximum observed Δ[25(OH)D] above baseline (ΔCmax); 3) timing of Cmax in days (Tmax); and 4) area under the Δ[25(OH)D]-time curve (AUC), which was interpreted as a global measure of vitamin D3 bioavailability. Individual participants’ AUCs were estimated manually by the trapezoidal method, and negative Δ[25(OH)D] values were zeroed so that the AUC represented the positive area above baseline. AUC was estimated for the first month to enable comparisons to other published PK studies 10 11 . AUC_28/35 was calculated for either 0 to 28 days or 0 to 35 days, depending on the timing of the blood sampling (Figure 2); similarly, AUC_56/70 was calculated for the period 0 to 56 days or 0 to 70 days. An individual’s average Δ[25(OH)D] during the first 28 days (ΔCavg₂₈) was calculated by dividing AUC₂₈ by 28; for between-study comparisons, this measure was expressed per 40,000 IU (1 mg) vitamin D3 by dividing ΔCavg₂₈ by the dose administered (1.75 mg). Cmax, ΔCmax, Tmax, AUC_28/35, and AUC_56/70 were summarized within groups by geometric means and 95% CIs, and then log-transformed for one-way analyses of variance (ANOVA) to test for differences between the pregnant and non-pregnant groups. To plot the longitudinal change in [25(OH)D] over time using all available data (N=61), mean [25(OH)D] at each visit were predicted from a linear regression model using a random intercept for each participant, with each visit represented by its own fixed indicator variable. Cross-sectional differences in Δ[25(OH)D] between pregnant and non-pregnant groups at specific days of follow-up were compared by ANOVA. Changes in biochemical ([Ca] and Ca:Cr) and clinical outcomes from baseline were analyzed using generalized estimating equations (GEE) to account for repeated measures. The association between cord venous [25(OH)D] and the corresponding maternal [25(OH)D] closest in time to delivery was analyzed using Pearson correlation.

The target sample size of at least 12 analyzable participants per single-dose group was originally justified as follows: assuming two samples per subject (baseline and peak), a standard deviation for the ΔCmax of 20 nmol/L and an intra-subject correlation of 0.6, we anticipated that at least 12 women in each group would enable the estimation of the mean ΔCmax with 95% confidence bounds of ±10 nmol/L. In all analyses, P values less than 0.05 were considered to be statistically significant, with corrections for multiple comparisons using the Holm method, applied where appropriate 17 . Analyses were conducted using Stata version 10 and 11 (Stata Corporation, College Station, Texas).

There was substantial inter-individual variation in the shape and magnitude of 25(OH)D responses to a single oral dose of 70,000 IU vitamin D3. However, the population-average pattern consisted of an abrupt increase in [25(OH)D] in the first week, followed by a peak within the first three weeks, and then a gradual return to baseline over the ensuing two months in both non-pregnant and pregnant participants (Figure 3). The average [25(OH)D] remained marginally above baseline at ten weeks after supplementation. There were minor differences between the pregnant and non-pregnant groups in the average Δ[25(OH)D] throughout follow-up (Table 1). In particular, [25(OH)D] rose more rapidly and the peak average occurred earlier in the non-pregnant group. This was demonstrated by the significantly greater Δ[25(OH)D] on day 2, the significantly lower Δ[25(OH)D] on day 21, and the slightly earlier occurrence of Tmax in non-pregnant vs. pregnant women (Table 1). Moreover, there was greater variance in the early Δ[25(OH)D] in non-pregnant vs. pregnant participants (Figure 3). The highest [25(OH)D] in any non-pregnant participant was 164 nmol/L, whereas the maximum in any pregnant participant was 116 nmol/L. On average, pregnant women had slightly lower absolute Cmax, but the mean maximal rise in [25(OH)D] (i.e., ΔCmax) and AUC were similar in pregnant and non-pregnant women (Table 1). Overall, the [25(OH)D] was an average of 19 nmol/L (95% CI, 14 to 25) higher than baseline during the first month after supplementation, which corresponded to a gain of approximately 12 nmol/L per mg of the vitamin D3 dose (Table 1).

The supplement was tasteless and well tolerated and there were no supplement-related adverse events (Table 3). The stillbirth and newborn deaths were explained by medical problems, and there was no evidence that either was related to the vitamin D supplementation, given their timing (i.e., did not occur at peak [25(OH)D]) and the absence of biochemical evidence of vitamin D toxicity in the mother (Table 3). Postmortem examinations were not feasible in the study setting. Two other AEs resolved without complications and occurred in the absence of evidence of vitamin D toxicity (Table 3). Pregnancy and birth outcome metrics were consistent with expectations for the source population (Table 4).

^a. Table includes participants with at least one serum albumin-adjusted calcium concentration measurement greater than 2.60 mmol/L or a serious clinical adverse event at any time during follow-up, among participants in the single-dose only group.

^b. All values refer to serum concentrations in maternal venous blood except for the row labeled “cord blood”. “Range” refers to the participant’s range of values during the entire follow-up period.

^a. In a sample of 113 deliveries at the study site (October 2009 to January 2010) for which there was a recalled first day of last menstrual period, the mean gestational age at birth was estimated to be 39.7 weeks (±2.2).

^b. Only includes the 12 liveborn infants.

^c. In a consecutive sample of 362 liveborn infants delivered at the study site (October 2009 to January 2010), the mean birth weight was 2780 g (±440).

^d. In a consecutive sample of 369 deliveries at the study site (October 2009 to January 2010), there were 199 cesarean deliveries (54%).

^e. There was one stillbirth. In a consecutive sample of 369 deliveries at the study site (October 2009 to January 2010), there were 7 stillbirths (2%).

^f. There was one neonatal death at 3 days of age.

Changes in average serum calcium concentrations (Figure 4) and urinary calcium excretion (Figure 5) occurred during the early phase of [25(OH)D] escalation. In non-pregnant participants, a transient increase in albumin-adjusted serum [Ca] from baseline was notable on day 4 (Table 5; Figure 4). The corresponding change in unadjusted total serum [Ca] was smaller and non-significant, and the raised adjusted [Ca] coincided with a lower average serum albumin on day 4 (difference versus baseline, -1.23 g/L; 95% CI, -2.12 to −0.34). In pregnant participants, there was an initial increase in albumin-adjusted [Ca] beginning on day 2 that persisted until nearly the end of the observation period (Figure 4), but the difference from baseline was only statistically significant on day 7 (Table 6). The unadjusted total [Ca] did not vary greatly from baseline and serum albumin remained relatively stable until the end of the 70-day follow-up, when many of the participants were post-partum.

^a. Number of specimens (there may have been multiple specimens for a single participant during a given follow-up period).

^b. Geometric means.

^c. Mean at day 4 was higher than baseline (+0.05; 95% CI, 0.03 to 0.07) and this remained statistically significant after correction for multiples testing.

^d. Means at days 7 and 14 were significantly higher than baseline, after correction for multiple testing.

^e. Mean at day 70 was significantly lower than baseline, after correction for multiple testing.

^a. Number of specimens (there may have been multiple specimens from a single participant during a given follow-up period).

^b. Geometric means.

^c. Mean at day 7 was higher than baseline (+0.05; 95% CI, 0.01 to 0.08), which remained statistically significant after correcting for multiple testing.

^d. Means at days 7, 14, 21, 28, 35, 56, and 63 were significantly higher than baseline, after correcting for multiple testing.

There were no episodes of confirmed hypercalcemia according to the study definition, and no isolated albumin-adjusted [Ca] values greater than the recent IOM upper limit of normal of 2.63 mmol/L. One pregnant participant had a single albumin-adjusted [Ca] = 2.61 mmol/L at one-week postpartum (70 days after dose administration) corresponding to a normal total [Ca] (2.51 mmol/L; serum albumin concentration was 35.8 g/L) that was within the reference range on repeat testing 4 days later (Table 3; Figure 6). A further follow-up one week following the first abnormal result was also normal (albumin-adjusted serum [Ca] of 2.44 mmol/L). This participant also had two non-consecutive episodes of urinary ca:cr higher than 1.0 mmol/mmol during follow-up (Figure 6). Her serum biochemical patterns were consistent with the expected changes in the perinatal period, including a gradual increase in albumin-adjusted serum [Ca] towards the end of the antenatal period and a rapid increase in serum albumin in the post-partum period 18 . Furthermore, there was no temporal association between the rise in [25(OH)D] and either the occurrence of isolated peaks in urine ca:cr or the isolated elevated [Ca] (Figure 6).

None of the participants manifested persistent hypercalciuria according to the study definition, or using a more conservative threshold of 0.85 mmol/mmol. In non-pregnant participants, the Ca:Cr increased from baseline but differences were only statistically significant at day 7 and 14 (Table 5). In pregnant participants, the increases in average Ca:Cr above baseline were more persistent and were statistically significant on all days except day 42, 49, and 70 (Table 6). There was no overall difference in the average ca:cr between non-pregnant and pregnant participants (P= 0.857).

Among participants who had received a single dose at baseline and for whom cord blood specimens were collected (N=12), the geometric mean cord serum [25(OH)D] was 50 nmol/L (95% CI, 40 to 62; range, 29 to 80). All cord serum albumin-adjusted [Ca] were within the normal range. The cord ln[25(OH)D] was moderately correlated with the maternal ln[25(OH)D] closest to the time of delivery (Pearson rho=0.64, P=0.02), and the average ratio of cord:maternal [25(OH)D] (N=12) was 0.88 (95% CI, 0.76 –1.02).