Overview of design
We designed this study to determine whether oral Mg supplementation was safe and acceptable to adolescents. Identification of effect size and determination of compliance with Mg supplements were primary objectives. We directed the study toward 8- to 14-yr-old girls because of the coincident rapid bone accretion and relative Mg undernutrition. Subjects were recruited from local pediatricians. We selected volunteers whose estimated dietary Mg intake was less than 220 mg/d; the major reason for exclusion was a greater Mg intake established by a 3-d dietary record, as described below. Parents were informed of the purpose of the study and dietary Mg status after screening.
The study was a randomized, placebo-controlled, double-blind, yearlong interventional trial of magnesium oxide compared with placebo. Tests were performed in the Clinical Research Centers at Yale University School of Medicine. The study protocol was approved by the Yale Human Investigation Committee. After baseline evaluation, subjects were evaluated at 1, 6, and 12 months after initiating supplementation. Subjects were contacted at 1- to 2-month intervals to assess safety and compliance. If any untoward events occurred during the study, subjects were instructed to contact the study coordinator.
Recruitment and enrollment
After initial contact of eligible subjects by pediatricians via office posting or letter, we explained the project in detail by telephone. If inclusion/exclusion criteria were met, written consent from parents and assent from children were obtained. The eligible study population consisted of premenarchal healthy Caucasian adolescent females, aged 8–14 yr. A registered dietitian interviewed the parent and child to obtain dietary details. After analysis of dietary data, individuals with average Mg intakes less than 220 mg/d were invited to participate in the yearlong supplementation trial. Those participating underwent physical examination by a research nurse trained in pediatric endocrinology. Tanner stage of breast development was recorded.
Inclusion criteria were as follows: Caucasian ethnicity, a ratio of weight-to-height between the third and 97th centiles, and the absence of bone disease. Exclusion criteria were as follows: scoliosis, onset of menses, use of chronic medications (retinoids, thyroid hormone, GH, glucocorticoids, oral contraceptives, anticonvulsants, diuretics, or supplements providing pharmacological dosages of vitamins A or D).
Randomization and intervention
Subjects were randomized in blocks of four to receive either Mg oxide or placebo (1:1 ratio), using a random number table. Study personnel and subjects were blinded to treatment. Mg was supplemented twice daily in a capsule containing powdered magnesium oxide (300 mg of elemental Mg per day). Identically appearing encapsulated methylcellulose powder served as placebo. Capsules were provided in calendar-coded cards with two capsules in each sealed blister. One- to 3-month supplies were distributed throughout the study. Monthly telephone contact by the study coordinator assessed safety and encouraged compliance.
At entry and after 6 and 12 months of supplementation, densitometric measures of the lumbar spine and hip were performed. BMC was chosen as the primary skeletal outcome variable because it is a direct measure and is not confounded by changes in bone area that occur during growth. Height and weight were recorded, and a complete biochemical profile was obtained at these times and additionally after 1 month of supplementation, as shown in and . Follow-up visits and blood sampling generally occurred in the mid-afternoon as to not interrupt school schedules.
Anthropomorphic data, dietary intake, and biochemistry at enrollment
Biochemistry values through the course of Mg supplementation (mean ± sd)
Densitometric measures of bone mass accrual
Bone densitometry was performed using dual-energy x-ray absorptiometry (Hologic QDR 4500W bone densitometer; Hologic, Bedford, MA) at four hip sites: femoral neck, trochanter, the intertrochanteric regions of the femoral diaphysis (which taken together are the total hip BMC), and Ward’s area. Anterio-posterior scans of the lumbar spine were obtained and were analyzed using pediatric software (Legacy Low Density Spine-revision C; Hologic). All scans were performed by one of two technicians with experience in performing bone densitometry in children. All scans were reviewed to ensure comparable definition of regions of the hip for serial scans within the same subject.
Serum and urinary biochemical determinations were performed by the Clinical Chemistry Laboratory at Yale-New Haven Hospital. Total serum and urinary calcium was determined by flame-atomic absorptiometry (model 2380; PerkinElmer, Norwalk, CT). Serum and urinary magnesium, phosphorus, and creatinine were measured using auto-analyzer technology. The urinary bone resorptive marker, NTx (Ostex, Seattle, WA), the N-telopeptide of type I collagen, was assayed by kit methodology. Serum immunoreactive PTH, 25-OHD, and 1,25(OH)2
D were measured as described (28
), as was serum osteocalcin (29
). Tubular reabsorption of phosphate (TRP) was calculated according to the formula:
from serum and concurrent timed urine samples.
A questionnaire on general food preferences was used to estimate daily Mg intake. Those with an estimated Mg intake less than 220 mg/d were provided detailed instructions for keeping an ongoing 3-d diet diary for detailed analysis. Instructions for completing the food record were provided in a face-to-face meeting using food models, and printed descriptions of portion sizes. Subjects were asked to record brand names of consumed foods, estimate portion sizes using household measurements, and describe food preparation. Subjects were asked to record their intake over two weekdays and one weekend day.
The completed record was reviewed by the registered dietitian, and any incomplete information was clarified by telephone contact. Results of the food record were provided to subjects, and those with an average Mg intake of less than 220 mg/d were invited to participate. A second 3-d food record was completed by participants midway through the study to assess consistency of intake. Nutrient analysis was performed by a registered research dietitian using the Food Processor Program (ESHA Research, Inc., Salem, OR).
Measures of compliance
Pill counts were performed, and percentage of missed doses was calculated. Fractional excretion of Mg (FEMg), a standard physiological parameter representative of Mg intake, was calculated at baseline and at 1, 6, and 12 months of supplementation, according to the formula:
from serum and concurrent timed urine samples.
Statistical analyses for BMC and bone mineral density (BMD) were performed in SAS version 8.2 (SAS Institute Inc., Cary, NC). A P value of 0.1 (one-sided) was used as the level of significance for all tests.
The primary objective of the analysis was to evaluate the magnitude and variability of the incremental BMC changes in the treatment group compared with the placebo group after 12 months of treatment. The secondary objectives were to assess trends in BMC and BMD as related to treatment for each maturity group and for each skeletal site examined. Maturity groups consisted of a prepubertal-early pubertal group (Tanner stage 1 or 2 at enrollment) and a mid-late pubertal group (Tanner stage 3 or 4 at enrollment). The primary hypotheses were tested using analysis of covariance models with repeated measures over three hip regions [femoral neck, total hip (encompassing the femoral neck, trochanteric, and intertrochanteric regions of the femoral diaphysis), and Ward’s area]. In the model, the incremental BMC change from baseline to 12 months was the outcome variable. The treatment (which has two levels) and maturity group (which has two levels) served as fixed effects, and baseline BMC served as a covariate to adjust the baseline effect. Within-subject covariance was adjusted by an unstructured variance-covariance pattern matrix. In addition to the factors and covariates described above, we also tested the following interactions: treatment by maturity group, treatment by location, and treatment by baseline BMC (or BMD). All interactions were tested at the 0.1 level. If none of the interactions was significant, the absolute difference of increase between baseline and the 12-month parameters for treatment and placebo groups was tested by the above-described ANCOVA model in an overall analysis. The associated 95% confidence interval was calculated as well. Least squares means and se values of BMC (and BMD) increases were calculated in the model and were plotted for each treatment group as a whole as well as for each Tanner group. If a significant treatment and maturity group interaction was found, we applied the same model for each maturity group to assess the treatment effect. Lumbar spine changes were of a markedly different magnitude and were therefore analyzed separately, using similar methodology.
Biochemical data were analyzed using analysis of covariance, employing SAS. Treatment comparisons of these parameters over the time frame of the study were made using a model that accounts for dependence of observations obtained from the same patient by modeling the correlation structure. Treatment, time, the interaction between treatment and time, and baseline levels of the outcome were included in the model as fixed effects. A secondary subgroup analysis was performed examining the effects of treatment and time of therapy with pubertal staging. Where appropriate, result of biochemical data are expressed as least squares means. A one-sided significance level of 0.05 was used to compare treatment vs. placebo groups, unless otherwise stated.