Vitamin D, osteocalcin, and risk for adiposity as comorbidities in middle school children

Authors

  • Claudia Boucher-Berry,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, Cohen Children's Medical Center, New Hyde Park, NY, USA
    Search for more papers by this author
  • Phyllis W Speiser,

    Corresponding author
    1. Department of Pediatrics, Division of Pediatric Endocrinology, Cohen Children's Medical Center, New Hyde Park, NY, USA
    2. Hofstra North Shore LIJ School of Medicine, Hofstra University, Hempstead, NY, USA
    • Division of Pediatric Endocrinology, Steven & Alexandra Cohen Children's Medical Center of New York, 269-01 76th Ave., New Hyde Park, NY 11040.
    Search for more papers by this author
  • Dennis E Carey,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, Cohen Children's Medical Center, New Hyde Park, NY, USA
    2. Hofstra North Shore LIJ School of Medicine, Hofstra University, Hempstead, NY, USA
    Search for more papers by this author
  • Steven P Shelov,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, Cohen Children's Medical Center, New Hyde Park, NY, USA
    2. Hofstra North Shore LIJ School of Medicine, Hofstra University, Hempstead, NY, USA
    Search for more papers by this author
  • Siham Accacha,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, Winthrop University Hospital, Mineola, NY, USA
    Search for more papers by this author
  • Ilene Fennoy,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, New York Presbyterian Medical Center, New York, NY, USA
    Search for more papers by this author
  • Robert Rapaport,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, Mt. Sinai Medical Center, New York, NY, USA
    Search for more papers by this author
  • Yomery Espinal,

    1. Department of Pediatrics, Division of Pediatric Endocrinology, New York Presbyterian Medical Center, New York, NY, USA
    Search for more papers by this author
  • Michael Rosenbaum

    1. Department of Pediatrics, Division of Pediatric Endocrinology, New York Presbyterian Medical Center, New York, NY, USA
    Search for more papers by this author

Abstract

Nonclassic actions of vitamin D include potential regulation of immune function and glucose homeostasis. The bone-metabolism loop has recently been expanded to include osteocalcin, which appears to play a more direct role in pancreatic beta cell function and energy metabolism. We hypothesized that both vitamin D and osteocalcin would correlate negatively with indices of adiposity-related comorbidity risk in periadolescents, varying by ethnic group. We analyzed anthropometric, metabolic, and inflammatory markers from a multiethnic population of 106 school children 11 to 14 years of age studied as part of the Reduce Obesity and Diabetes (ROAD) consortium. As expected, 25-hydroxyvitamin D (25-OH vitamin D) was inversely correlated with intact parathyroid hormone (iPTH); total osteocalcin (OCN) and uncarboxylated osteocalcin (uOCN) were directly correlated with each other. OCN and uOCN concentrations correlated inversely with age. Vitamin D deficiency was most prevalent among East Asians (EA) and African Americans (AA). The highest lipid risk scores and homeostatic model for assessment of insulin resistance (HOMA-IR) values were seen in the South Asian (SA) group. Overall, adiposity measures were inversely correlated with OCN and iPTH, whereas such relationships were not observed for vitamin D. Acute insulin response to glucose challenge correlated negatively with uOCN in all subjects; however, lipid risk score correlated negatively with uOCN only in whites. The relationships between markers of calcium metabolism and body composition, glucose homeostasis, lipids, and inflammation all showed racial and ethnic differences. No consistent relationship was found between vitamin D and adiposity or vitamin D and glucose metabolism; instead vitamin D levels varied by race and ethnicity in this school-based group. These findings are consistent with the hypothesis that markers of calcium and bone metabolism may reflect risk for adiposity-related comorbidities in children. © 2012 American Society for Bone and Mineral Research

Introduction

As the prevalence of vitamin D deficiency has increased across all age groups, reports have proliferated documenting putative comorbidities.1 National Health and Nutrition Examination Survey (NHANES) data from 2008 indicate that 9% of the pediatric population in the United States has vitamin D levels below 15 ng/mL and are considered vitamin D–deficient.2 Another 61% of children have vitamin D levels between 15 and 30 ng/mL and are considered vitamin D–insufficient.2 Such cut-points vary by laboratory and are not strictly evidence-based.3 Recent Institute of Medicine guidelines suggest a universal definition of deficiency as a 25-hydroxyvitamin D (25-OH vitamin D) serum level below 20 ng/mL.4

Vitamin D is acknowledged to be especially important for normal bone growth and avoidance of rickets in childhood. The vitamin D–mediated accretion of calcium from childhood through early adulthood is a major determinant of bone integrity in old age.1 Apart from these functions, the vitamin D receptor has been found in the pancreas, brain, liver, and immune system,5 and it has been hypothesized that vitamin D modulates endocrine, central nervous system (CNS), hepatic, and immunological functions. A corollary to these hypotheses is that, beyond bone health, vitamin D deficiency might have negative implications for systemic health. Studies done in animals and in adult humans have reported negative correlations of circulating vitamin D concentrations with adiposity, insulin sensitivity, and immune function.6, 7 Increased vitamin D levels during weight loss has been independently correlated with improved insulin sensitivity8 and decreased circulating concentrations of proinflammatory cytokines and triglycerides.9

These links between vitamin D, obesity, and adiposity-related comorbidities suggest the involvement of a wider array of bone markers10–12 in relevant metabolic systems. One such marker of bone metabolism is osteocalcin (OCN). This noncollagenous protein, expressed in osteoblasts and adipocytes,12–14 exists in two forms, uncarboxylated and carboxylated, and is used as a marker for bone formation. In adults15, 16 and children,17 circulating OCN levels are positively correlated with insulin sensitivity, whereas low levels of uncarboxylated OCN (uOCN) correlate with beta cell dysfunction in prediabetic children.17 Knockout mice lacking the OCN gene have more visceral fat, higher insulin resistance, and higher glucose levels than wild-type mice. Replacing OCN reverses these effects in mice.11, 12 Interestingly, a vitamin D receptor binding site has been detected in the promoter region of the OCN gene.18 However, the exact relationship between vitamin D and OCN remains unclear, since vitamin D supplementation has no apparent effect on percent serum uOCN in adolescent girls,19 and vitamin D receptor polymorphisms are not correlated with serum OCN levels.20

Another hormone important in bone metabolism is parathyroid hormone (PTH). Typically, high levels of vitamin D inhibit PTH and vitamin D deficiency stimulates PTH secretion. Studies of intact PTH (iPTH) levels in adults21 and children22 have reported positive correlations of iPTH and body mass index (BMI), as would be expected from the negative correlation between adiposity and vitamin D. This inverse relationship between vitamin D and PTH is more variable among youth than in adults.23 There is also an independent negative correlation between circulating iPTH and insulin sensitivity in adults.24

To further explore these observations, we investigated the relationship of serum vitamin D, OCN, and iPTH to measures of adiposity, lipids, glucose metabolism, and inflammatory markers in children aged 11 to 14 years. We hypothesized that both vitamin D and OCN would correlate negatively with indices of adiposity-related comorbidity risk in periadolescents. Specifically, we asked whether observations in obese pediatric clinic populations would hold true in a school-based population sample. Other aims included recruitment of an ethnically varied population and analysis of such correlations by gender, ethnicity, and race.

Subjects and Methods

Study subjects

Blood samples were obtained from 106 middle school children (age 11–14 years, mean age 12.6 years; 61 males, 45 females). Subjects were enrolled in the Reduce Obesity and Diabetes (ROAD) study involving five ethnically diverse schools in the New York metropolitan area. The ethnic distribution of subjects was 22 African Americans, 21 white Americans, 24 Hispanic Americans, and 38 Asian Americans (28 East Asian, 10 South Asian); one subject could not be so classified and is not included in the table of ethnic distribution, but is included in the other data analyses. The ROAD project, a 5-year study, was directed through a research consortium consisting of Columbia University Medical Center, Maimonides Medical Center, Mt. Sinai School of Medicine, Cohen Children's Medical Center of New York, and Winthrop University Hospital that has been assembled via the Academy for Medical Development and Collaboration (AMDeC, New York, NY, USA). Detailed methods for this study have been described elsewhere.25 An initial population of 619 subjects was divided into four groups: African American, white, Hispanic, and Asian American. For this substudy, subjects were selected from initial screening tests performed in the fall of the first year they were enrolled. Within each group, subjects were ranked from the lowest to the highest values for measures of insulin secretion and resistance (glucose disposal index [GDI], acute insulin response [AIR], and homeostatic model for assessment of insulin resistance [HOMA-IR]). Subjects in the first three groups with the highest HOMA-IR in each 10-percentile rank were chosen, followed by the highest glucose disposal index in each 10-percentile rank to generate a total of 22 to 25 subjects per group. In the case of Asian-Americans, a similar procedure was used for the highest values of HOMA-IR and GDI in each 5-percentile rank to generate a total of 38 subjects to include a mixed population of East Asians (from China, Japan, Taiwan, and Korea) and South Asians (from India, Pakistan, Ceylon, and Bangladesh). Outliers for fasting insulin or inflammatory markers (interleukin 6 [IL-6], tumor necrosis factor–alpha [TNF-α], and c reactive protein [CRP]) more than 3 SD from the mean were eliminated prospectively on the assumption that such subjects were probably either not truly fasting or were ill. More specifically, five subjects were excluded for insulin >40 mIU/mL; three subjects were excluded because IL-6 was >4.2 pg/mL; two subjects were excluded because TNF-α was >9.6 pg/mL; and three subjects were excluded because CRP was >36.2 pg/mL. There was some overlap between these groups and, in total, 12 subjects were excluded from individual analyses. An additional two subjects were excluded because of inadequate sample collection. The clinical protocol was approved by the Institutional Review Board for each participating hospital, the school boards, and the Department of Health and Education. This study was conducted in conformity with the guiding principles for research involving humans.26 Written informed consent and assent were obtained from all parents and students, respectively.

Procedures

Height, weight, BMI, and waist circumference (at the iliac crest) were measured; BMI Z-score27 and waist Z-score28 were calculated. Body composition was measured by bioimpedance using an Omron Body Fat Analyzer HBF-300 (Omron Health Care, Inc., Vernon Hills, IL, USA). Dietary history of vitamin D and daily dairy intake were estimated from the Block Food Frequency Questionnaire.29 Fasting blood samples were assayed for glucose, insulin, 25-OH vitamin D, iPTH, uOCN, total osteocalcin (tOCN), CRP, IL-6, TNF-α, and adiponectin (ACRP30). Participants received an intravenous infusion of 0.5 gm/kg dextrose (maximum 25 gm) over 3 minutes. Blood was drawn for insulin concentrations at 1, 3, and 5 minutes after the infusion.

Laboratory analyses

Serum vitamin D was measured by radioimmunoassay (RIA) (Diasorin, Stillwater, MN, USA). The intra- and interassay variability were 8.2% and 10.5%, respectively. We were uncertain of the cut-points for vitamin D deficiency; therefore, measurements were analyzed as a continuous variable. tOCN levels were measured in serum samples using enzyme-linked immunosorbent assay (ELISA) (IDS, Fountain Hills, AZ, USA). The intra- and interassay variability were 1.8% and 2.7%, respectively. uOCN levels were also measured in serum samples using a separate ELISA (Takara Bio, Inc., Madison, WI, USA). The intra- and interassay variability were 4.4% and 9.5%, respectively. iPTH was measured by immunoradiometric assay (IRMA) (Scantibodies, Santee, CA, USA). The intraassay variability ranged from 3.2% to 4.8% and the interassay variability ranged from 3.6% to 6.8%. Vitamin D, OCN, and iPTH levels were all measured in duplicate. Glucose was measured by the hexokinase method (Glucose/HK; Roche Molecular Biochemicals, Werk Penzbuerg, Germany). Plasma insulin was measured by solid phase 125-I-RIA (Coata-count; DPC, Los Angeles, CA, USA). CRP, TNF-α, and ACRP30 were determined by ELISA. IL-6 was assayed by RIA (R&D Systems, Minneapolis, MN, USA).

Statistical analysis and calculations

Fat mass was calculated from the weight and the percent body fat using the following equation: fat mass = (weight)(% body fat)/100. The HOMA-IR was calculated as described.30 First-phase insulin release was examined using the AIR, which was calculated as the mean rise in circulating concentrations of insulin 3 and 5 minutes following administration of intravenous dextrose. Because insulin sensitivity and insulin release are significantly correlated, the GDI was calculated as log10(AIR × fasting glucose concentration/fasting insulin concentration) to examine insulin secretory capacity once corrected for insulin sensitivity.31, 32

Data are presented as mean ± SD unless otherwise indicated. Whenever possible, data were converted into Z-scores to control for the effects of age and gender. This was done for BMI and waist circumferences as indices of adiposity and truncal adiposity, and for total cholesterol, triglyceride, HDL, and LDL for lipids. In addition, we calculated a “lipid risk score” similar to the Framingham risk score,33 as the Z-scores for LDL plus triglycerides minus the Z-score for HDL. Further analyses of individual components of the lipid panel were conducted only if there were significant differences between ethnic groups in the lipid risk score, or if there was a significant correlation of bone markers with the lipid risk score. Multiple linear regression analyses were performed and partial correlation coefficients were calculated, using the entire data set for race and each bone marker as independent variables, and measures of glucose homeostasis, lipids, and inflammation as individual dependent variables. Similar analyses were conducted using BMI Z-score as an independent variable and each bone marker as a dependent variable. When race was a significant covariate, further analyses comparing individual ethnic groups were conducted. Using this technique significantly reduces the number of comparisons being made. The initial hypotheses were that bone markers are significantly correlated with fat (BMI Z-score), truncal adiposity (waist circumference Z-score), insulin resistance (HOMA-IR), insulin secretory capacity (GDI), lipids (lipid risk score), and inflammatory markers. Because these dependent variables are all interrelated, post hoc adjustments, which assume that the tests are independent, were not made.34–36

Correlation coefficients were calculated among: (1) vitamin D, tOCN, uOCN, and iPTH; (2) measures of adiposity as dependent variables, and bone markers as independent variables; and (3) measures of glucose homeostasis and inflammation as dependent variables, and bone markers as independent variables. In the event that both circulating concentrations of bone markers and other variables were found to be significantly correlated with adiposity, we performed an analysis of covariance (ANCOVA) adding age, gender, and % body fat as covariates. Normality of distribution was assessed by Wilk-Shapiro testing. Statistical significance was prospectively defined as Pα < 0.05. All p values <0.10 are reported.

Results

Demographic and anthropometric characteristics

All data were normally distributed once corrected for significant age, ethnic, and gender effects. Ethnic distribution is shown in Table 1; gender-specific data are shown in Table 2. The mean population BMI Z-score was 0.86 ± 1.3 and BMI Z-scores in whites were significantly greater than those for East Asian subjects. Total fat mass was significantly greater in African American, white American, Hispanic American and South-Asian American children compared to East-Asian American children. Mean waist Z-scores were 0.67 ± 1.04 without significant interethnic differences. Females were significantly taller.

Table 1. Baseline Subject Characteristics by Ethnic/Racial Group
VariableAfrican American (n = 22)White American (n = 21)Hispanic American (n = 24)East-Asian American (n = 28)South-Asian American (n = 10)All Subjectsa (n = 106)
  • Values are expressed as mean (SD) unless otherwise indicated. Bold values indicate statistically significant differences.

  • M = male; F = female; BMI = body mass index.

  • a

    Includes three subjects (1 male, 2 female) classified as “other” because ethnicity could not be ascertained.

  • *

    p < 0.05 versus East-Asian Americans.

  • **

    p < 0.05 compared with white and East-Asian Americans.

  • ***

    p < 0.01 compared with East-Asians Americans.

Gender (M/F)12/1010/1113/919/96/461/45
Age (years)12.9 (0.8)12.7 (1.0)13.3 (0.9)12.4 (0.8)13.1 (1.1)12.6 (1.0)
Height (cm)159.9 (7.9)156.5 (6.8)161.6(11.6)155.7 (8.5)160.5 (9.1)158.3 (9.2)
Weight (kg)60.9 (14.7)62.2 (15.2)63.0 (17.2)50.7 (16.2)56.2 (11.5)58.2 (15.9)
BMI (kg/m2)23.7 (5.1)24.9 (5.6)23.9 (5.2)20.6 (5.1)21.7 (3.3)22.9 (5.3)
BMI Z-score1.06 (1.24) 1.84 (1.52) *1.53 (5.19)0.88 (1.67)0.48 (0.73)0.86 (1.30)
Waist (cm)78.3 (11.9)80.4 (13.7)81.0 (13.7)72.5 (13.9)78.8 (12.4)77.7 (13.4)
Waist Z-score0.70 (0.90)0.86 (1.07)0.98 (1.22)0.37 (1.12)0.66 (0.91)0.67 (1.04)
% Body fat29.0 (9.1)29.7 (8.8)28.4 (8.3)25.5 (6.9)29.1 (7.7)28.0 (8.1)
Fat mass (kg) 18.6 (8.9) * 19.3 (9.1) * 18.8 (9.5) *13.5 (7.9) 16.7 (6.7) *17.1 (8.7)
Calcium intake (mg/day) 917 (678) **502 (275) 730 (368) **525 (339)644 (218)653 (438)
Vitamin D Intake (IU/day)158 (97)104 (73)128 (65)106 (80)129 (31)122 (77)
Dairy intake (servings/day) 1.45 (1.13) ***0.87 (0.66)1.07 (0.65)0.73 (0.64)1.03 (0.31)1.00 (0.78)
Table 2. Anthropometric and Biochemical Variables by Gender
VariableMales (n = 45)Females (n = 61)
  • Values are expressed as mean (SD) unless otherwise indicated. Bold values indicate statistically significant differences.

  • BMI = body mass index; OCN = osteocalcin; PTH = parathyroid hormone; HDL = high-density lipoprotein; LDL = low-density lipoprotein; QUICKI = quantitative insulin sensitivity check index; HOMA-IR = homeostatic model for assessment of insulin resistance; AIR = acute insulin response; GDI = glucose disposal index; IL-6 = interleukin 6; CRP = c-reactive protein; TNF-α = tumor necrosis factor-alpha.

  • a

    Calculated as the sum of the LDL and triglyceride Z-scores minus the HDL Z-score.

  • *

    p < 0.05 compared to males.

  • **

    p < 0.01 compared to males.

Demographics
 Age (years)12.7 (0.9)13.0 (0.9)
 Height (cm) 156.5 (7.0) 160.0 (9.9) *
 Weight (kg)56.4 (14.9)59.5 (16.5)
 BMI (kg/m2)22.9 (5.4)22.9 (5.1)
 BMI Z-score0.80 (1.07)0.94 (0.92)
 Waist circumference (cm)77.5 (13.1)77.8 (13.7)
 % Body fat29.4 (7.3)27.0 (8.6)
 Fat mass (kg)17.3 (8.1)16.9 (9.1)
 Calcium intake (mg/day)719 (481)549 (345)
 Vitamin D intake (IU/day) 135 (80) 103 (66) *
 Dairy intake (servings/day)1.08 (0.80)0.85 (0.76)
Bone mineralization markers
 25-OH vitamin D (ng/mL)18.7 (6.6)20.6 (9.7)
 25-OH vitamin D < 20 ng/mL (%)5152
 25-OH vitamin D < 30 ng/mL (%)8884
 Total OCN (ng/mL) 68.7 (52.3) 91.4 (47.0) **
 Uncarboxylated OCN (ng/mL) 23.9 (23.9) 40.6 (26.9) *
 Intact PTH (pg/mL)33.1 (12.9)30.2 (11.0)
Lipids
 Cholesterol (mg/dL) 172 (27) 159 (34) *
 Cholesterol Z-score 0.33 (1.01) −0.04 (0.76) *
 Triglycerides (mg/dL)81 (35)72 (39)
 Triglyceride Z-score0.02 (0.65)−0.05 (0.67)
 HDL (mg/dL)54 (12)52 (12)
 HDL Z-score0.25 (0.90)0.16 (0.88)
 LDL (mg/dL)101 (31)93 (23)
 LDL Z-score0.14 (0.46)0.03 (0.36)
 Lipid risk scorea−0.09 (1.47)−0.18 (1.36)
Glucose homeostasis
 Glucose (mg/dL)91 (7)94 (7)
 Insulin (mU/mL)10.7 (7.4)9.6 (6.8)
 QUICKI0.34 (0.03)0.35 (0.03)
 HOMA-IR2.52 (1.75)2.21 (1.57)
 AIR (mIU/mL)88.4 (91.4)98.0 (79.6)
 GDI2.84 (0.41)2.92 (0.23)
Inflammatory markers
 IL-6 (pg/mL)1.06 (0.71)1.02 (0.64)
 CRP (µg/mL) 2.81 (3.40) 4.82 (5.83) *
 TNF-α (pg/mL)2.24 (2.06)1.96 (1.33)
 Adiponectin (µg/mL) 13.7 (4.8) 11.1 (4.7) **

Metabolic measures

Daily vitamin D intake was significantly greater in males than females (Table 2), as reported elsewhere.37 Daily calcium intake was significantly higher in African and Hispanic Americans compared to white and East-Asian Americans (Table 1). Vitamin D in all subjects was significantly correlated with iPTH (r = −0.34, p < 0.001). The majority of subjects in all ethnic groups had circulating 25-OH vitamin D levels below 30 ng/mL, but African Americans, South-Asian Americans, and East-Asian Americans had 25-OH vitamin D levels below 20 ng/mL (Table 3). Vitamin D deficiency was most prevalent among East Asians (EA) and African Americans (AA). Diet recall surveys indicated that on average subjects consumed only one daily dairy serving with vitamin D intake estimated at <200 IU daily (Table 2). As expected, serum vitamin D levels correlated with both dairy servings and vitamin D intake by diet recall surveys in the whole cohort.

Table 3. Biochemical Variables by Ethnic/Racial Group
VariableAfrican American (n = 22)Caucasian American (n = 21)Hispanic American (n = 22)East Asian American (n = 28)South Asian American (n = 10)All Subjects (n = 106)a
  • Values are expressed as mean (SD) unless otherwise indicated. Bold values indicate statistically significant differences.

  • OCN = osteocalcin; PTH = parathyroid hormone; HDL = high-density lipoprotein; LDL = low-density lipoprotein; QUICKI = quantitative insulin sensitivity check index; HOMA-IR = homeostatic model for assessment of insulin resistance; AIR = acute insulin response; GDI = glucose disposal index; IL-6 = interleukin 6; CRP = c-reactive protein; TNF-α = tumor necrosis factor-alpha.

  • a

    Includes 3 subjects (1 male, 2 female) classified as “other” because ethnicity could not be ascertained.

  • b

    Calculated as the sum of the LDL and triglyceride Z-scores minus the HDL Z-score.

  • *

    p < 0.01 versus African Americans and East Asian Americans.

  • **

    p < 0.05 versus African Americans, Caucasian Americans, and Hispanic Americans.

  • ***

    p < 0.01 versus African Americans.

25-OH vitamin D (ng/mL)16.0 (6.3) 24.9 (9.4) * 22.5 (8.2) *16.1 (7.1)18.7 (7.5)20.6 (9.7)
 25-OH vitamin D < 20 ng/mL (%)681936795051
 25-OH vitamin D < 30 ng/mL (%)1007682979086
Total OCN (ng/mL)66.7 (40.3)66.8 (37.5)70.1 (50.3) 103.3 (52.6) **80.8 (27.5)91.4 (47.0)
Uncarboxylated OCN (ng/mL)28.4 (27.7)34.4 (23.2)34.4 (29.8)36.2 (29.1)40.7 (23.6)40.6 (26.9)
Intact PTH (pg/mL)34.8 (10.5) 26.7 (8.2) ***32.0 (14.4)31.2 (12.0)31.0 (14.8)30.2 (11.0)
Cholesterol (mg/dL)174 (43)165 (28)151 (24)168 (22)169 (30)164 (31)
Cholesterol Z-score0.40 (1.25)0.17 (0.84)−0.22 (0.66)0.12 (0.74)0.30 (0.82)0.12 (0.89)
Triglycerides (mg/dL)60 (22)80 (30)83 (36)78 (50)85 (42)76 (37)
Triglyceride Z-score−0.29 (0.39)0.05 (0.51)0.11 (0.63)0.02 (0.88)0.15 (0.77)−0.02 (0.66)
HDL (mg/dL)55 (11)50 (10)47 (12)59 (13)50 (10)53 (12)
HDL Z-score0.35 (0.85)−0.03 (0.73)−0.15 (0.91)0.59 (0.97)0.04 (0.74)0.20 (0.89)
LDL (mg/dL)107 (38)100 (25)87 (21)93 (21)102 (23)96 (27)
LDL Z-score0.25 (0.57)0.13 (0.38)−0.06 (0.32)0.01 (0.32)0.18 (0.35)0.08 (0.41)
Lipid risk scoreb−0.40 (1.19)0.20 (1.12)0.20 (1.46)−0.56 (1.64)0.29 (1.45)−0.14 (1.40)
Glucose (mg/dL)90 (8)93 (7)92 (7)95 (6)96 (5)93 (7)
Insulin (mU/mL)10.2 (6.6)9.0 (4.4)10.0 (5.4)9.7 (9.5)14.1 (7.9)10.1 (7.0)
QUICKI0.35 (0.03)0.35 (0.03)0.34 (0.02)0.35 (0.03)0.32 (0.02)0.35 (0.03)
HOMA-IR2.31 (1.51)2.30 (1.33)2.32 (1.24)2.30 (2.19)3.32 (1.86)2.33 (1.64)
AIR (mIU/mL)140.5 (128.3)70.4 (44.8)72.1 (37.1)105.3 (96.7)79.2 (26.6)94.2 (84.0)
GDI2.99 (0.27)2.76 (0.24)*2.82 (0.19)*3.03 (0.30)2.75 (0.280)2.89 (0.32)
IL-6 (pg/mL)1.10 (0.68)0.84 (0.45)1.09 (0.58)0.99 (0.54)1.44 (1.24)1.04 (0.67)
CRP (µg/mL)3.56 (4.40)3.71 (4.43)4.54 (6.12)4.30 (5.87)3.52 (2.88)3.95 (5.00)
TNF-α (pg/mL)2.26 (2.38)2.31 (1.70)2.71 (1.50)1.34 (0.67)1.93 (2.03)2.08 (1.68)
Adiponectin (µg/mL)13.7 (7.2)12.4 (4.3)11.3 (3.4)12.2 (4.9)11.4 (3.3)12.2 (4.9)

iPTH was significantly lower in white Americans than in African Americans, the groups with the highest and lowest vitamin D levels, respectively. (Table 3)

tOCN and uOCN correlated positively with each other. OCN and uOCN concentrations correlated inversely with age (Table 4). Females had significantly higher tOCN (91.4 ± 47.0 ng/mL versus 68.7 ± 52.3 ng/mL, p = 0.03) and uOCN (40.6 ± 26.9 ng/mL versus 23.9 ± 23.9 ng/mL, p = 0.002) (Table 2). The significant effect of gender on circulating concentrations of OCN remained significant even when corrected for BMI, % body fat, and fat mass (see Fig. 1). tOCN was significantly higher in East-Asian Americans than in all other ethnic groups except South-Asian Americans (Table 3). Although there is no standard range for tOCN and uOCN, levels found in our cohort were consistent with ranges demonstrated in previous pediatric studies.38

Figure 1.

(A) Regression of BMI and tOCN in males (open squares) and females (closed squares). Overall regression equation is: OCN (ng/mL) = −2.5[BMI (kg/m2)] + 141.1; r = −0.27, p = 0.008. When data were analyzed as a multiple linear regression, treating BMI and gender (dichotomous variable) as independent variables, the overall regression equation was improved to an adjusted r2 value of 0.12, p = 0.002. There were significant effects on of both BMI (semipartial r = −0.26, p = 0.009) and gender (semipartial r = −0.22, p = 0.025), indicating that tOCN relative to BMI was significantly lower in females. (B) Regression of % body fat and tOCN in males (open circles) and females (closed circles). Overall regression equation is: OCN (ng/mL) = −1.54 (% body fat) + 125.1; r = −0.25, p = 0.014. When data were analyzed as a multiple linear regression analysis treating % body fat and gender (dichotomous variable) as independent variables, the overall regression equation was improved to an adjusted r2 value of 0.10, p = 0.007. There were significant effects of both BMI (semipartial r = −0.23, p = 0.028) and gender (semipartial r = −0.20, p = 0.048), indicating that tOCN relative to % body fat was significantly lower in females. (C) Regression of fat mass and tOCN in males (open triangles) and females (closed triangles). Overall regression equation is: OCN (ng/mL) = −1.7 [fat mass (kg)] −110.9; r = −0.29, p = 0.004. When data were analyzed as a multiple linear regression, treating fat mass and gender (dichotomous variable) as independent variables, the overall regression equation was improved to an adjusted r2 value of 0.13, p = 0.001. There were significant effects on of both fat mass (semipartial r = −0.29, p = 0.004) and gender (semipartial r = −0.22, p = 0.027), indicating that tOCN relative to fat mass was significantly lower in females.

Males had higher total cholesterol (172 ± 35 mg/dL versus 159 ± 27 mg/dL, p < 0.05), total cholesterol Z-scores (0.33 ± 1.01 versus −0.04 ± 0.76, p < 0.05), ACRP30 (13.7 ± 4.8 µg/mL versus 11.1 ± 4.7 µg/mL, p < 0.01), and lower CRP (2.8 ± 3.4 µg/mL versus 4.8 ± 5.8 µg/mL) (Table 2). None of the subjects had evidence of diabetes.

Correlations between bone and anthropometric markers

Vitamin D was not significantly correlated with any anthropometric variables. Indices of body fatness (% fat, BMI, and BMI Z-score) and absolute fat mass correlated inversely with iPTH and OCN (Figs. 1 and 2; Table 4).

Figure 2.

Values of r for correlations between bone markers and indices of body fatness corrected for age and gender (BMI Z-score), fat mass, and fractional body fat content (% body fat). No significant correlations were found with 25-OH vitamin D or uOCN for any of these variables. *p < 0.05.

Table 4. Values of r and p for Significant Correlations
Variable25-OH vitamin DOCNuOCNiPTH
  • OCN = osteocalcin; uOCN = uncarboxylated osteocalcin; iPTH = intact parathyroid hormone; HA = Hispanic American; EA = East-Asian American; BMI = body mass index; CA = white American; SA = South-Asian American; AA = African American; HDL = high-density lipoprotein; LDL = low-density lipoprotein; HOMA-IR = homeostatic model for assessment of insulin resistance; AIR = acute insulin response; GDI = glucose disposal index; IL-6 = interleukin 6; CRP = c-reactive protein; TNF-α = tumor necrosis factor-alpha.

  • a

    Calculated as the sum of the LDL and triglyceride Z-scores minus the HDL Z-score.

Age (years)All: r = 0.20, p = 0.048All: r = −0.32, p = 0.015All: r = −0.25, p = 0.017 
  Boys: r = −0.38, p = 0.003Boys: r = −0.35, p = 0.01 
  Girls: r = 0.35, p = 0.032  
  EA: r = −0.56, p = 0.004EA: r = −0.54, p = 0.008 
 HA: r = 0.44, p = 0.046HA:r = −0.51, p = 0.021HA:r = −0.51, p = 0.021 
BMI (kg/m2) All: r = −0.27, p = 0.008Boys: r = −0.32, p = 0.020All: r = −0.20, p = 0.040
BMI Z-score All: r = −0.20, p = 0.048 All: r = −0.24, p = 0.016
Waist (cm)  Girls: r = 0.32, p = 0.048 
Waist Z-score CA: r = −0.45, p = 0.044  
% Body fat All: r = −0.25, p = 0.014 All: r = −0.26, p = 0.007
    Boys: r = −0.34, p = 0.009
    SA: r = −0.70, p = 0.023
 HA: r = −0.45, p = 0.050   
Fat mass (kg) All: r = −0.29, p = 0.004 All: r = −0.23, p = 0.021
  Boys: r = −0.29, p = 0.030Boys: r = −0.36, p = 0.007Boys: r = −0.25, p = 0.05
    SA: r = −0.64, p = 0.045
25-OH vitamin D   All: r = −0.34, p < 0.001
    Boys: r = −0.27, p = 0.05
    Girls: r = −0.46, p = 0.012
    HA: r = −0.58, p = 0.001
OCN  All: r = 0.65, p < 0.001 
   Boys: r = 0.78, p < 0.001 
   Girls: r = 0.43, p = 0.008 
   AA: r = 0.76, p < 0.001 
   CA: r = 0.56, p = 0.023CA: r = 0.61, p = 0.01
   EA: r = 0.89, p < 0.001 
   HA: r = 0.86, p < 0.001 
uOCN All: r = 0.65, p < 0.001  
  Boys: r = 0.78, p < 0.001  
  Girls: r = 0.43, p = 0.008 Girls: r = 0.33, p = 0.048
  AA: r = 0.76, p < 0.001  
 CA: r = 0.61, p = 0.01CA: r = 0.56, p = 0.023  
  EA: r = 0.89, p < 0.001  
  HA: r = 0.86, p < 0.001  
iPTHAll: r = −0.34, p < 0.001   
 Boys: r = −0.27, p = 0.05   
 Girls: r = −0.46, p = 0.012 Girls: r = 0.33, p = 0.048 
  CA: r = 0.61, p = 0.01  
 HA: r = −0.58, p = 0.001   
Cholesterol (mg/dL)   Boys: r = −0.27, p = 0.032
Cholesterol Z-score   Boys: r = −0.28, p = 0.031
Triglycerides (mg/dL)    
Triglyceride Z-score    
HDL (mg/dL)    
HDL Z-score    
LDL (mg/dL)AA: r = −0.52, p = 0.021   
 EA: r = 0.52, p = 0.005   
LDL Z-scoreAA: r = −0.50, p = 0.022   
   CA: r = −0.68, p = 0.003 
 EA: r = 0.51, p = 0.006   
Lipid risk scoreaAA: r = −0.44, p = 0.048   
   CA: r = −0.62, p = 0.008 
Glucose (mg/dL) All: r = 0.20, p = 0.049  
    HA: r = 0.52, p = 0.015
Insulin (mU/mL)    
HOMA-IR  CA: r = −0.72, p = 0.006 
AIR (mIU/mL)  All: r = −0.24, p = 0.043 
  CA: r = −0.56, p = 0.038CA: r = −0.53, p = 0.049 
GDIAll: r = −0.21, p = 0.047   
 Girls: r = −0.45, p = 0.013   
IL-6 (pg/mL)   AA: r = −0.50, p = 0.026
 CA: r = −0.45, p = 0.047   
CRP (µg/mL)    
TNF-α (pg/mL)  Boys: r = −0.37, p = 0.007 
  HA: r = 0.51, p = 0.022HA:r = 0.52, p = 0.015 
Adiponectin (µg/mL) EA: r = −0.44, p = 0.036EA: r = −0.54, p = 0.008 
   SA: r = −0.65, p = 0.042 

Correlations between bone markers and glucose metabolism

Vitamin D showed a weak negative correlation with GDI, a measure of insulin secretory capacity, in all subjects, but a stronger negative correlation in females. Greater significance was observed in the negative associations of uOCN with HOMA-IR, a measure of insulin resistance. Among whites, both OCN and uOCN were inversely related to AIR, a measure of beta cell function. AIR also correlated with uOCN in the group as a whole. Fasting glucose concentrations were positively correlated with iPTH (Table 4).

Correlations between bone markers and inflammatory markers

Vitamin D and iPTH correlated negatively with IL6 in Caucasians. OCN and uOCN correlated positively with TNF-α in Hispanic Americans, and negatively with adiponectin in East-Asian Americans (Table 4).

Discussion

To our knowledge, this is the first pediatric study to simultaneously examine vitamin D, iPTH, and OCN in relation to adiposity, glucose metabolism, and inflammation. Our major findings are that tOCN and iPTH, but not vitamin D itself, are negatively correlated with markers of adiposity. There were also inverse correlations between the uncarboxylated form of osteocalcin and beta cell function (AIR) and insulin resistance (HOMA-IR), respectively, which varied by racial and ethnic group. We found no direct correlation between vitamin D and OCN levels.

Our findings differ from previous studies showing vitamin D levels to be significantly lower in obese children and adults.5–7, 20–22 The absence of a direct correlation between vitamin D and adiposity measures in our cohort is consistent with the observation that vitamin D supplementation to raise serum levels had no effect on body composition over a 4-year period in adults.39 Because most of the students in this study were vitamin D deficient, we examined the possibility that the significant correlations of adiposity and vitamin D noted in other studies were asymmetrically evident in vitamin D–sufficient versus vitamin D–deficient populations. However, stratifying our population by vitamin D quartiles, analyzing those with levels below 20 ng/mL separately, or omitting apparent outliers, did not alter our results; no significant correlations between adiposity and vitamin D were noted in any cohorts. In multiple regression analysis vitamin D levels differed significantly by race and/or ethnicity, even when corrected for BMI Z-score. Vitamin D levels were lowest in African Americans and East Asian Americans, whose dark skin pigment may interfere with light-induced vitamin D synthesis.40 It is interesting to note that white Americans and Hispanic Americans had the highest vitamin D levels, and whites had the lowest PTH levels. While dietary habits and vitamin supplementation practices may differ in these groups, average calcium and vitamin D intakes were low in all students based on diet recall. The ethnic specific vitamin D levels and difficulties demonstrating a robust correlation between vitamin D and adiposity measures in children of color are similar to recently published data from Rajakumar and colleagues,41 whose Pittsburgh population was categorized solely as either African American or white. This group was only able to show a correlation between vitamin D and BMI in African Americans if they omitted an outlier with normal serum vitamin D and high BMI. However, unlike our cohort, the former group did demonstrate more highly significant inverse relationships between vitamin D levels and BMI as well as visceral fat in whites.41 Future studies examining vitamin D associations must consider race, ethnicity, gender, age, existing dietary habits, subject ascertainment mode (school or population-based versus clinic-based), and weight range of subjects (within normal weight range versus mainly overweight or obese).

OCN has been correlated inversely with both BMI and % body fat in adult humans and in mice.12, 42 OCN-null mice have excess visceral fat,12 whereas mice given OCN exhibit a dose-dependent decrease in fat mass.11 Thus, unlike vitamin D administration to humans, which does not affect adiposity,39 OCN diminishes adiposity in animal models. We have found a strong inverse association between serum OCN levels and several measures of adiposity in a school-based group of adolescents. In concordance with our results, Reinehr and Roth43 found that obese children tended to have lower OCN levels than the nonobese controls, and that osteocalcin was inversely correlated with HOMA-IR, a measure of insulin resistance. Interestingly, in the latter study, substantial weight loss in the obese group was associated with an increase in OCN levels. We do not have a mechanistic explanation for why a significant association between AIR and OCN was only seen in whites. It is quite possible that with a larger population, this association would be detected in other ethnic groups. It is also possible that other variables related to adiposity, such as pubertal status, which may affect OCN but were not assessed in this cohort, would account for some of the differences observed. This study was not powered to perform beta-analyses of nonsignificant associations. There is a complex interaction of age, pubertal status, and gender with OCN levels. In contrast to our study, and using a different OCN assay, the HELENA cross-sectional study of bone markers in adolescents found higher serum OCN levels in males compared with females.44 Other pediatric studies showed a marginally significant OCN increment in boys43; however, puberty was not addressed. Since OCN levels seem to rise earlier in pubertal female subjects compared with males,45 this could explain the interstudy gender discrepancies. In a longitudinal study of adults, older women had higher total OCN levels compared with men.46

With respect to glucose metabolism, mice with increased levels of OCN secondary to a gain-of-function mutation exhibit greater insulin sensitivity.12 The few retrospective studies done in humans, in both children and adults, have shown a comparable inverse relationship between OCN levels and measures of insulin resistance.42, 43, 47 Our data further refine these observations, indicating that it is the uncarboxylated form of OCN exerting major effects on beta cell function and insulin resistance, in concordance with observations in mice.11, 12 The only robust correlation for vitamin D in this regard was with insulin secretory capacity, as measured by GDI, in females.

Our observations of associations between OCN, iPTH, measures of absolute fat mass, and fractional fat mass, as well as gender effects on these parameters have not been previously reported in children. The correlation between OCN and body composition was most significant for absolute fat mass compared to other indices of adiposity (% body fat, BMI, or BMI Z-score). This suggests that OCN is more influenced by absolute fat mass than by the proportion of body fat. The existence of gender effects on the relationship of OCN to indices of body fatness indicates that there are primary gender differences in OCN production or clearance. In contrast to OCN, iPTH was most strongly correlated with % body fat, suggesting that iPTH is more influenced by the proportion of body fat, rather than by absolute fat mass. This type of relationship may be analogous to the relationships of leptin and circulating insulin concentrations to body fat: Leptin levels tend to vary by total fat mass,48 whereas insulin levels correlate better with relative fat mass; ie, % body fat. The absence of direct correlations between vitamin D and adiposity measures suggests that OCN and iPTH may exert effects independent from vitamin D.

Several lines of evidence indicate that the nonclassic actions of vitamin D include suppression of inflammatory cytokines (reviewed in Bikle49). Administration of exogenous vitamin D to adult subjects undergoing weight loss resulted in decreased circulating concentrations of proinflammatory cytokines,7 although there were no significant associations between vitamin D and cytokines prior to weight loss. The inflammatory markers analyzed in the present study all reflect cytokines produced in adipocytes. Low adiponectin levels have been particularly associated with high risk for cardiovascular morbidity in adult men,50 and with early atherosclerotic changes in children.51 Based on murine studies, Lee and colleagues12 postulated that tOCN regulates insulin sensitivity partially through ACRP30. In the elderly, there is a direct positive association of OCN and adiponectin, both negatively associated with insulin resistance.52 In our cohort, we found rather weak inverse correlations between vitamin D, PTH, and IL-6, and only in select ethnic groups, while OCNs displayed positive correlations with TNF-α, and negative correlations with ACRP30 in different ethnic populations. Reinehr and Roth43 found no relationship between OCN and adiponectin in a smaller cohort of obese children. The source of these apparent discrepancies is unknown, but it is possible that cross-sectional sampling in growing children does not readily reveal such relationships.

One strength of the present study is the inclusion of an ethnically diverse school-based periadolescent population, rather than clinic-based obese subjects. A weakness is that these data represent only a single set of cross-sectional samples from a 5-year study. Budget constraints prevented us from measuring these many parameters in the entire subject population. It is perhaps surprising that although OCN correlated with fat mass and other adiposity measures, it was uOCN that more strongly correlated with measures of insulin resistance. We attribute this to the relatively small subject population and gender-based variability of OCN. It is important to note that this field is under active investigation, and published reports have differed. Variances have also been detected in adult men between higher serum uOCN levels, which correlated with improved beta cell function, and higher OCN levels, which correlated with lower insulin resistance,15 and lower risk for type 2 diabetes.16 Yet another larger study of adult men found uOCN correlated inversely with insulin resistance.53 Definitive statements about the effects of race, ethnicity, gender, age, and pubertal status would require a much larger population.

In summary, our study expands upon the relationship between vitamin D, OCN, iPTH, adiposity, and insulin sensitivity in children. Osteocalcin seems to correlate better than vitamin D in all areas we explored: measures of adiposity, lipid risk score, insulin resistance, and inflammatory cytokines associated with metabolic syndrome. We have further demonstrated that race, ethnicity, gender, and age must be considered when evaluating metabolic phenotypes. Our findings should be considered hypothesis-generating and further studies need to be done to more fully evaluate these relationships.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgements

Funding was provided through AMDeC by the Starr Foundation as well as with support from NIH grant numbers UL1RR024156, UL1RR0023568 and CTSA grant UL1 RR024156 to Columbia University and NIH/NCRR GCRC grant M01 RR018535 to NSLIJHS. Dr. Speiser served as Dr. Boucher-Berry's research mentor; Drs. Carey, Rosenbaum and Shelov served on the scholarly oversight committee. Dr. Speiser, as corresponding author, accepts responsibility for data integrity. ROAD Collaborators helped recruit subjects, performed anthropometric measurements, collected questionnaire data and blood samples, and participated in monthly group discussions. We gratefully acknowledge the invaluable participation of all the students, teachers, and school administrators, as well as the NYC Board of Health and Department of Education. We thank the GCRC nurses for assistance with sample collection. We thank Serge Cremers, PhD for technical assistance, Steven Holleran, PhD for statistical guidance, and Drs. Alan Jacobson and Stuart Weinerman for helpful comments.

ROAD Collaborators: L. Altshuler and B. Lowell (Department of Pediatrics, Division of Behavioral & Developmental Pediatrics, Maimonides Medical Center, Brooklyn, NY, USA); A. Bhangoo, D. DeSantis, R. Gupta, A.A. Hassoun, and S. Ten (Department of Pediatrics, Division of Pediatric Endocrinology, Maimonides Medical Center, Brooklyn, NY, USA); S. Close, R. Conroy, A.M. Jean, R.L. Levine, K. Pavlovich, and E. Shamoon (Department of Pediatrics, Division of Pediatric Endocrinology, New York Presbyterian Medical Center, New York, NY, USA); L. Iazetti and M.L. Klein (Department of Pediatrics, Division of Pediatric Endocrinology- Mt. Sinai Medical Center, New York, NY, USA); and F.J. Jacques, L. Michel, and W. Rosenfeld (Department of Pediatrics, Division of Pediatric Endocrinology, Winthrop University Hospital, Mineola, NY, USA).

Authors' roles: Study design: CBB, PWS, MR, DEC, SPS, SA, IF, RR. Study conduct: CBB. Data collection: CBB, PWS, MR, YE. Data interpretation: CBB, PWS, MR, DEC, SPS, SA, IF, RR, YE. Drafting and revising manuscript: CBB, PWS, MR, DEC, SPS, SA, IF, RR. Approval of final manuscript: CBB, PWS, MR, DEC, SPS, SA, IF, RR, YE.

Ancillary