Objective
We sought to determine whether 4000 IU/d (vs 2000 IU/d) of vitamin D during pregnancy is safe and improves maternal/neonatal 25-hydroxyvitamin D [25(OH)D] in a dose-dependent manner.
Study Design
A total of 257 pregnant women 12-16 weeks’ gestation were enrolled. Randomization to 2000 vs 4000 IU/d followed 1-month run-in at 2000 IU/d. Participants were monitored for hypercalciuria, hypercalcemia, and 25(OH)D status.
Results
Maternal 25(OH)D (n = 161) increased from 22.7 ng/mL (SD 9.7) at baseline to 36.2 ng/mL (SD 15) and 37.9 ng/mL (SD 13.5) in the 2000 and 4000 IU groups, respectively. While maternal 25(OH)D change from baseline did not differ between groups, 25(OH)D monthly increase differed between groups ( P < .01). No supplementation-related adverse events occurred. Mean cord blood 25(OH)D was 22.1 ± 10.3 ng/mL in 2000 IU and 27.0 ± 13.3 ng/mL in 4000 IU groups ( P = .024). After controlling for race and study site, preterm birth and labor were inversely associated with predelivery and mean 25(OH)D, but not baseline 25(OH)D.
Conclusion
Maternal supplementation with vitamin D 2000 and 4000 IU/d during pregnancy improved maternal/neonatal vitamin D status. Evidence of risk reduction in infection, preterm labor, and preterm birth was suggestive, requiring additional studies powered for these endpoints.
With avoidance of sunlight exposure due to lifestyle changes, concerns regarding skin cancer, and the resultant widespread use of sunscreen, fewer Americans are meeting their needs for vitamin D. A study published in 2002 by the Centers for Disease Control and Prevention and our laboratory at the Medical University of South Carolina (MUSC) revealed that 42% of African American women in their childbearing years exhibited vitamin D deficiency (hypovitaminosis D). More recent publications suggest that the rate of deficiency is higher than previously reported.
Until recently, there was no Recommended Dietary Allowance (RDA) for vitamin D, only an adequate intake, which remained at 200 IU/d of vitamin D for decades. A 2010 review of the vitamin D requirements by the Institute of Medicine (IOM) resulted in a revised RDA of 600 IU/d of vitamin D, and suggested that fewer Americans are deficient than previously reported. Using the IOM definition of deficiency of <20 ng/mL (50 nmol/L); however, 2 recent large studies of vitamin D status in pregnant women living at latitude 32°N (South Carolina) showed that African American women were 8 times as likely as Hispanic women, and 20 times as likely as Caucasian women, to have vitamin D deficiency. Yet little information exists that addresses the vitamin D requirements of the pregnant woman and her fetus.
The recent Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD)-sponsored randomized controlled trial of vitamin D supplementation using 400, 2000, or 4000 IU/d of vitamin D 3 starting at 12 weeks of gestation showed that 400 IU/d was woefully inadequate in achieving vitamin D sufficiency. Optimal 1,25-dihydroxyvitamin D 3 [1,25(OH) 2 D] production (ie, the point at which its concentration has reached steady-state) was found to occur when total circulating 25-hydroxyvitamin D [25(OH)D] concentration was at least 40 ng/mL (or 100 nmol/L). In addition, only the 4000 IU group achieved optimal status in all women throughout pregnancy, including second trimester, irrespective of race. The translation of vitamin D supplementation at these doses to other women receiving care in a nonuniversity health setting has become essential.
This study was undertaken to shed light on the vitamin D requirements of pregnant women receiving their prenatal care at community health centers. The study’s primary hypothesis was that vitamin D supplementation of 2000 or 4000 IU/d during pregnancy was safe and effective in achieving vitamin D sufficiency and would result in improved maternal and neonatal health status. It was further hypothesized that the 4000 IU dose would result in a greater increase in 25(OH)D than the 2000 IU dose. The coprimary outcome measures were the change in circulating 25(OH)D concentration from baseline to the completion of pregnancy in the mother, and her neonate’s 25(OH)D concentration at birth.
Materials and Methods
Study design
This was a 2-center, randomized, double-blinded study of vitamin D supplementation (Food and Drug Administration Investigational New Drug #66,346; ClinicalTrials.gov # NCT00412087 ). The study was approved by the MUSC Institutional Review Board (IRB) for Human Research (HR# 16476) and the Palmetto Baptist Hospital (Columbia, SC) IRB for Human Research (PH IRB# 2007-25). Written informed consent was obtained from each participant. Women <16 weeks’ gestation were eligible for participation in the study.
A precedent NICHD vitamin D supplementation trial during pregnancy had begun in 2004 with baseline 25(OH)D analysis revealing marked deficiency among darker pigmented women. Further, because 400 IU/d of vitamin D had already been shown to be ineffective in maintaining adequate vitamin D status during pregnancy and because the majority of women being recruited in this study were either African American or Hispanic, with darker pigmentation and a greater likelihood of vitamin D deficiency, a control group that would receive 400 IU/d of vitamin D was considered unethical by the scientific review committee as well as the research team. Hence, a control arm was not included a priori in the study design.
Study setting
This study was conducted from Nov. 21, 2006, through June 30, 2010, at Eau Claire Cooperative Health Center (ECCHC) in Columbia, SC, and Northwoods Community Health Center (NCHC) in North Charleston, SC. ECCHC and NCHC are both Section 330 community health centers. At least 60% of the ECCHC and NCHC patient populations are of racial/ethnic minority groups, including African American and Hispanic women. The populations being served by ECCHC and NCHC are among the poorest in South Carolina; the majority of women are 100% below the federally designated poverty level and many women rank 200% below federal poverty level.
Inclusion and exclusion criteria
The inclusion criteria were: maternal age ≥16 years; confirmed singleton pregnancy of <16 completed weeks of gestation at the time of enrollment; and intention to receive ongoing prenatal care at the community health center where consent was obtained.
Mothers with preexisting calcium or parathyroid conditions or who required chronic diuretic or cardiac medication therapy, including calcium channel blockers, were not eligible for enrollment into the study. Mothers with active thyroid disease (eg, Graves, Hashimoto, or thyroiditis) also were not eligible to participate in the study; however, mothers on thyroid supplement with normal serological parameters could participate in the study if they were without any other endocrine dysfunction.
Randomization and intervention
Upon enrollment into the study, expectant mothers’ vitamin D status was assessed by measuring total circulating 25(OH)D and parathyroid hormone (PTH). Based on this initial 25(OH)D level, the randomization to 2000 or 4000 IU/d of vitamin D 3 was stratified using a cut point of 32 ng/mL. Randomization lists were generated by computer prior to the start of the study. Randomization assignment was blinded to all participants and to the investigators except for the study biostatistician. Dose groups were identified for logistical purposes using 6 letters (3 per dose group) as an additional measure against inadvertent unblinding.
Adherence to medication regimen
Adherence to the vitamin D supplementation regimen was measured by maternal self-report and pill counts at each follow-up visit. If a woman missed 1 prenatal visit, her next month’s supply of vitamins was mailed to her or delivered to her residence. If a woman had >2 missed visits or if she failed to take at least 50% of the prescribed vitamin D pills, she was exited from the study.
Study protocol
Gestational age at enrollment
Gestational age was based on last menstrual period. If a woman was unsure of this date, the obstetrical estimate at the time of the visit was used. If, at the 20-week fetal ultrasound, it was determined by the obstetrician that the gestational age was incorrect, the revised gestational age was used.
Initial study visit
Baseline blood and urine samples were obtained following each participant’s consent at the initial visit (10 to <16 weeks). Irrespective of enrollment gestational age, vitamin D supplementation did not begin before the 12th week of gestation (12 and 0/7 weeks).
Subsequent study visits
Participants were followed with monthly study visits, which continued until delivery. These visits coincided with routine obstetrical visits. There was 1 additional visit with mother and infant 2 weeks’ postpartum.
Completion of questionnaires
Participants completed questionnaires used in the NICHD vitamin D pregnancy trial, which included sociodemographic information, baseline health status, and medical history at the first visit. At the second visit, the Block Food Frequency Questionnaire (Block, Berkeley, CA) was completed to ascertain generalized eating pattern, with specific calculation of calcium and vitamin D intake. An interim maternal health history questionnaire also was completed at each visit with the assistance of the study coordinator to ascertain adverse events, and type and frequency of acute illnesses such as respiratory, gastrointestinal, and other viral and/or bacterial illnesses. A review of medications and doctor’s visits was obtained at that time. After delivery, the newborn record of each infant was reviewed for mode of delivery, birthweight (grams), and gestational age.
Blood and urine samples
Maternal blood samples were collected at the first visit, then every other obstetrical visit and at the time of delivery. Maternal urine samples were collected at each visit. Cord blood was obtained at delivery. If the cord blood sample could not be obtained, a neonatal blood sample was drawn within 2 weeks of delivery.
Materials
Source of vitamin D
Vitamin D tablets (1600 and 3600 IU) were manufactured by Tishcon Corp (Westbury, NY) a Good-Manufacturing-Practice facility. Hoffman-La Roche Ltd (Basel, Switzerland) supplied the cholecalciferol content contained in the vitamin D tablet manufactured by Tishcon Corp. The tablet vitamin concentration was verified by the company every 6 months and by an independent laboratory chosen by the investigators (Heartland Assays, Ames, IA) using high-performance liquid chromatograph with ultraviolet light detection to ensure the tablets met label claim throughout the study; these results were reported to the MUSC Investigational Drug Department. Tablets were maintained in the MUSC Investigational Drug Division of Pharmacy until the time that they were dispensed to enrolled subjects at ECCHC or NCHC.
Source of prenatal vitamins
Prenatal vitamins (400 IU vitamin D 3 /tablet) prescribed at study entry were Myadec multivitamin-multimineral supplement (distributed by Pfizer Consumer Healthcare, Morris Plains, NJ). Those mothers unable to swallow a prenatal vitamin were given a Flintstones Complete chewable vitamin (Bayer Healthcare, Morristown, NJ) (400 IU vitamin D 3 per vitamin).
Study measures
Maternal sociodemographic questionnaire
Upon enrollment in the study, each mother was asked to complete a sociodemographic questionnaire to ascertain maternal age, race, educational level, occupation, and insurance status.
Race/ethnicity definition
Each mother was asked to describe the racial/ethnic group to which she belonged, by selecting any applicable categories from African American, Caucasian, Hispanic, American Indian, Asian, and other.
Pregnancy intake and surveillance survey
Upon enrollment, each woman was asked to complete a health assessment questionnaire to ascertain her use of medications (checklist) and over-the-counter preparations that may have influenced vitamin D/calcium homeostasis. Additional questions concerned use of cigarettes and alcohol, and overall health status.
Pregnancy health status, and labor and delivery characteristics and complications
Characteristics of each mother’s health status and complications during pregnancy, labor, and delivery were recorded. Complications at the time of delivery were listed according to American Congress of Obstetricians and Gynecologists definitions. In addition, if the mother required hospitalization, a copy of the hospital record was obtained after she signed a release of medical information form. Any acute illnesses or development of pregnancy-related conditions that were not preexisting also were recorded. When appropriate, the Data Safety and Monitoring Committee (DSMC) and the IRB were notified of any adverse events.
Season
The season that each blood sample was drawn was defined as spring (April through May), summer (June through September), fall (October through November), and winter (December through March).
Maternal body mass index measurement
Prepregnancy height and weight of each mother were recorded at the first outpatient visit to determine body mass index (weight [kg]/height [m 2 ]). During subsequent visits, only the mother’s weight was recorded, and the initial height and updated weight were used to calculate body mass index at each outpatient visit.
Neonatal growth parameters
At the postpartum visit, the infant’s weight in grams, head circumference in centimeters, and length in centimeters were recorded. The growth parameters were then plotted using Fenton growth curves, which facilitate the calculation of z-scores and permit the more precise assessment of growth of infants who are born preterm.
Laboratory measurements
Maternal and cord blood/neonatal total circulating 25(OH)D assays
A rapid, direct radioimmunoassay developed in an author laboratory (B.W.H.) and manufactured by Diasorin Corp (Stillwater, MN) was used to measure total circulating 25(OH)D concentration in serum samples as previously described. Based on clinical laboratory classifications, a priori, deficiency was defined as total circulating 25(OH)D 20 ng/mL (50 nmol/L), insufficiency as ≥20-32 ng/mL (≥50-80 nmol/L), and sufficiency as ≥80 nmol/L (≥32 ng/mL). The interassay and intra-assay coefficient of variation was ≤10%.
Maternal and infant concentrations of serum calcium, creatinine, and phosphorus
Maternal serum total calcium, creatinine, and inorganic phosphorus were measured bimonthly (maternal) and at delivery (cord blood) by MUSC Clinical Chemistry Laboratory using standard methodology and laboratory normative data.
Monthly maternal urinary calcium:creatinine ratio
Urinary calcium and creatinine were measured monthly for each mother by the MUSC Clinical Chemistry Laboratory (to convert mg/dL of calcium to mmol/L, multiply the value by 0.25. To convert mg/dL of creatinine to mmol/L, multiply by 0.088), then expressed as a ratio (urinary calcium [mg/dL]: creatinine [mg/dL] ratio).
Maternal and cord blood concentrations of circulating intact PTH
Intact PTH (iPTH) was measured in serum samples by immunoradiometric assay as previously described. The adult normal range for iPTH in our laboratory is 1.3-5.4 pmol/L. In general, higher vitamin D levels are associated with lower iPTH; as vitamin D status improves, iPTH declines.
Safety monitoring
All study participants were monitored monthly for hypervitaminosis D. The first sign of hypervitaminosis D is hypercalciuria, of which, urinary calcium:creatinine ratio is the most sensitive indicator. Operationally, we defined a priori caution limits for hypervitaminosis D as a urinary calcium:creatinine (mg/dL) ratio ≥0.8. The study’s DSMC reviewed the quarterly summary reports that were generated for all subjects enrolled in the study. Whenever any patient was to exceed the caution limit, a specific case study was to be initiated to examine the contribution of confounding factors (eg, diet, sunlight exposure). Operationally, we were to stop vitamin D 3 supplementation if the urinary calcium:creatinine ratio (measured monthly) exceeded 1.0 or if the circulating 25(OH)D level (measured bimonthly) exceeded 100 ng/mL , and the DSMC and IRB were notified immediately. The principal investigator of the study reviewed all laboratory results on a weekly basis to identify potentially abnormal values.
Statistical methods
Sample size and power considerations
A total of 148 participants were to be enrolled in the study with 74 per supplementation arm. For one primary endpoint of change in 25(OH)D level between baseline and final measurements, this sample size would support the detection of a 10-ng/mL difference between dose groups with 80% power using a 2-sided t test at α = 0.05. This calculation assumed that the SD of 25(OH)D measurements at a single time point was approximately 10, that there would be a low correlation (ρ = 0.25) between the baseline and final measurements, and that a substantial proportion (up to 50%) of participants may be lost to follow-up. This calculation was robust to changes in the assumption regarding the magnitude of correlation between measurements made over time; if a higher correlation were to be present, the study’s power would be increased.
Statistical analyses
Primary analysis
The a priori primary data analysis focused on comparisons of change in serum 25(OH)D between the 2 dose groups. While participants were randomized within each stratum, as described above, the strata were extremely imbalanced at the close of the study due to the infrequent occurrence of 25(OH)D levels >32 ng/mL ( Figure 1 ). Thus, the primary comparison for planning purposes was a 2-sample test of dose group differences in the change in 25(OH)D levels, regardless of stratum. This comparison also was performed using the Wilcoxon rank sum test as a sensitivity analysis to the normality assumption. The latter test is reported as the primary analysis. Statistical significance is claimed for P < .05. Due to the multisite implementation of the study, we also report results controlled for site using multivariable models.
Secondary analysis
In secondary analyses, multilevel mixed-effects models were used to estimate the average monthly rate of change in 25(OH)D, compare this rate between dose groups, and explore the effects of covariates on the rate of change. These models included fixed effects for dose group, time, and the group-time interaction, and a random intercept effect, with additional covariate effects as required. Time was considered a continuous variable, measured in months rather than assuming structured visit occurrences. An unstructured covariance matrix was assumed. The same approach was used to evaluate the longitudinal association between 25(OH)D and calcium, iPTH (log transformed), phosphorus, and urinary calcium, creatinine, and calcium:creatinine levels. The cumulative occurrence of pregnancy complications was compared between dose group levels using logistic regression. All analyses were performed using software (SAS, version 9.3; SAS Institute Inc, Cary, NC).
Participant attrition and missing data
Because the primary endpoint was change in 25(OH)D from baseline to delivery, the primary analysis was restricted to participants who remained in the study until delivery and provided a blood sample within 6 weeks prior to delivery, at delivery, or at the postdelivery visit (completers-only analysis). Typically, multiple imputation would be used to impute missing values in support of the favored intention-to-treat analytic approach. Because the multiple imputation model for this analysis would have required variables also measured in the final blood sample, however, it could not be used to impute cases with a missing final blood sample. Thus, to assess the primary findings’ robustness to assumptions about the missing data, we performed a sensitivity analysis under the following assumptions: cases with missing endpoints experienced no change in both groups; experienced the group-specific median change observed in completers; experienced no change in the 2000 IU group and minimal change in the 4000 IU group. In the secondary analyses using multilevel mixed-effects models for longitudinal modeling, all available data points were used, as it is not necessary to delete cases with missed time points when using this approach.
Materials
Source of vitamin D
Vitamin D tablets (1600 and 3600 IU) were manufactured by Tishcon Corp (Westbury, NY) a Good-Manufacturing-Practice facility. Hoffman-La Roche Ltd (Basel, Switzerland) supplied the cholecalciferol content contained in the vitamin D tablet manufactured by Tishcon Corp. The tablet vitamin concentration was verified by the company every 6 months and by an independent laboratory chosen by the investigators (Heartland Assays, Ames, IA) using high-performance liquid chromatograph with ultraviolet light detection to ensure the tablets met label claim throughout the study; these results were reported to the MUSC Investigational Drug Department. Tablets were maintained in the MUSC Investigational Drug Division of Pharmacy until the time that they were dispensed to enrolled subjects at ECCHC or NCHC.
Source of prenatal vitamins
Prenatal vitamins (400 IU vitamin D 3 /tablet) prescribed at study entry were Myadec multivitamin-multimineral supplement (distributed by Pfizer Consumer Healthcare, Morris Plains, NJ). Those mothers unable to swallow a prenatal vitamin were given a Flintstones Complete chewable vitamin (Bayer Healthcare, Morristown, NJ) (400 IU vitamin D 3 per vitamin).
Study measures
Maternal sociodemographic questionnaire
Upon enrollment in the study, each mother was asked to complete a sociodemographic questionnaire to ascertain maternal age, race, educational level, occupation, and insurance status.
Race/ethnicity definition
Each mother was asked to describe the racial/ethnic group to which she belonged, by selecting any applicable categories from African American, Caucasian, Hispanic, American Indian, Asian, and other.
Pregnancy intake and surveillance survey
Upon enrollment, each woman was asked to complete a health assessment questionnaire to ascertain her use of medications (checklist) and over-the-counter preparations that may have influenced vitamin D/calcium homeostasis. Additional questions concerned use of cigarettes and alcohol, and overall health status.
Pregnancy health status, and labor and delivery characteristics and complications
Characteristics of each mother’s health status and complications during pregnancy, labor, and delivery were recorded. Complications at the time of delivery were listed according to American Congress of Obstetricians and Gynecologists definitions. In addition, if the mother required hospitalization, a copy of the hospital record was obtained after she signed a release of medical information form. Any acute illnesses or development of pregnancy-related conditions that were not preexisting also were recorded. When appropriate, the Data Safety and Monitoring Committee (DSMC) and the IRB were notified of any adverse events.
Season
The season that each blood sample was drawn was defined as spring (April through May), summer (June through September), fall (October through November), and winter (December through March).
Maternal body mass index measurement
Prepregnancy height and weight of each mother were recorded at the first outpatient visit to determine body mass index (weight [kg]/height [m 2 ]). During subsequent visits, only the mother’s weight was recorded, and the initial height and updated weight were used to calculate body mass index at each outpatient visit.
Neonatal growth parameters
At the postpartum visit, the infant’s weight in grams, head circumference in centimeters, and length in centimeters were recorded. The growth parameters were then plotted using Fenton growth curves, which facilitate the calculation of z-scores and permit the more precise assessment of growth of infants who are born preterm.
Laboratory measurements
Maternal and cord blood/neonatal total circulating 25(OH)D assays
A rapid, direct radioimmunoassay developed in an author laboratory (B.W.H.) and manufactured by Diasorin Corp (Stillwater, MN) was used to measure total circulating 25(OH)D concentration in serum samples as previously described. Based on clinical laboratory classifications, a priori, deficiency was defined as total circulating 25(OH)D 20 ng/mL (50 nmol/L), insufficiency as ≥20-32 ng/mL (≥50-80 nmol/L), and sufficiency as ≥80 nmol/L (≥32 ng/mL). The interassay and intra-assay coefficient of variation was ≤10%.
Maternal and infant concentrations of serum calcium, creatinine, and phosphorus
Maternal serum total calcium, creatinine, and inorganic phosphorus were measured bimonthly (maternal) and at delivery (cord blood) by MUSC Clinical Chemistry Laboratory using standard methodology and laboratory normative data.
Monthly maternal urinary calcium:creatinine ratio
Urinary calcium and creatinine were measured monthly for each mother by the MUSC Clinical Chemistry Laboratory (to convert mg/dL of calcium to mmol/L, multiply the value by 0.25. To convert mg/dL of creatinine to mmol/L, multiply by 0.088), then expressed as a ratio (urinary calcium [mg/dL]: creatinine [mg/dL] ratio).
Maternal and cord blood concentrations of circulating intact PTH
Intact PTH (iPTH) was measured in serum samples by immunoradiometric assay as previously described. The adult normal range for iPTH in our laboratory is 1.3-5.4 pmol/L. In general, higher vitamin D levels are associated with lower iPTH; as vitamin D status improves, iPTH declines.
Safety monitoring
All study participants were monitored monthly for hypervitaminosis D. The first sign of hypervitaminosis D is hypercalciuria, of which, urinary calcium:creatinine ratio is the most sensitive indicator. Operationally, we defined a priori caution limits for hypervitaminosis D as a urinary calcium:creatinine (mg/dL) ratio ≥0.8. The study’s DSMC reviewed the quarterly summary reports that were generated for all subjects enrolled in the study. Whenever any patient was to exceed the caution limit, a specific case study was to be initiated to examine the contribution of confounding factors (eg, diet, sunlight exposure). Operationally, we were to stop vitamin D 3 supplementation if the urinary calcium:creatinine ratio (measured monthly) exceeded 1.0 or if the circulating 25(OH)D level (measured bimonthly) exceeded 100 ng/mL , and the DSMC and IRB were notified immediately. The principal investigator of the study reviewed all laboratory results on a weekly basis to identify potentially abnormal values.
Statistical methods
Sample size and power considerations
A total of 148 participants were to be enrolled in the study with 74 per supplementation arm. For one primary endpoint of change in 25(OH)D level between baseline and final measurements, this sample size would support the detection of a 10-ng/mL difference between dose groups with 80% power using a 2-sided t test at α = 0.05. This calculation assumed that the SD of 25(OH)D measurements at a single time point was approximately 10, that there would be a low correlation (ρ = 0.25) between the baseline and final measurements, and that a substantial proportion (up to 50%) of participants may be lost to follow-up. This calculation was robust to changes in the assumption regarding the magnitude of correlation between measurements made over time; if a higher correlation were to be present, the study’s power would be increased.
Statistical analyses
Primary analysis
The a priori primary data analysis focused on comparisons of change in serum 25(OH)D between the 2 dose groups. While participants were randomized within each stratum, as described above, the strata were extremely imbalanced at the close of the study due to the infrequent occurrence of 25(OH)D levels >32 ng/mL ( Figure 1 ). Thus, the primary comparison for planning purposes was a 2-sample test of dose group differences in the change in 25(OH)D levels, regardless of stratum. This comparison also was performed using the Wilcoxon rank sum test as a sensitivity analysis to the normality assumption. The latter test is reported as the primary analysis. Statistical significance is claimed for P < .05. Due to the multisite implementation of the study, we also report results controlled for site using multivariable models.
Secondary analysis
In secondary analyses, multilevel mixed-effects models were used to estimate the average monthly rate of change in 25(OH)D, compare this rate between dose groups, and explore the effects of covariates on the rate of change. These models included fixed effects for dose group, time, and the group-time interaction, and a random intercept effect, with additional covariate effects as required. Time was considered a continuous variable, measured in months rather than assuming structured visit occurrences. An unstructured covariance matrix was assumed. The same approach was used to evaluate the longitudinal association between 25(OH)D and calcium, iPTH (log transformed), phosphorus, and urinary calcium, creatinine, and calcium:creatinine levels. The cumulative occurrence of pregnancy complications was compared between dose group levels using logistic regression. All analyses were performed using software (SAS, version 9.3; SAS Institute Inc, Cary, NC).
Participant attrition and missing data
Because the primary endpoint was change in 25(OH)D from baseline to delivery, the primary analysis was restricted to participants who remained in the study until delivery and provided a blood sample within 6 weeks prior to delivery, at delivery, or at the postdelivery visit (completers-only analysis). Typically, multiple imputation would be used to impute missing values in support of the favored intention-to-treat analytic approach. Because the multiple imputation model for this analysis would have required variables also measured in the final blood sample, however, it could not be used to impute cases with a missing final blood sample. Thus, to assess the primary findings’ robustness to assumptions about the missing data, we performed a sensitivity analysis under the following assumptions: cases with missing endpoints experienced no change in both groups; experienced the group-specific median change observed in completers; experienced no change in the 2000 IU group and minimal change in the 4000 IU group. In the secondary analyses using multilevel mixed-effects models for longitudinal modeling, all available data points were used, as it is not necessary to delete cases with missed time points when using this approach.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
