Poor Glycemic Control Is Associated With Impaired Bone Accrual in the Year Following a Diagnosis of Type 1 Diabetes

Abstract

Context

Type 1 diabetes (T1D) is associated with an increased fracture risk across the life course. The effects on bone accrual early in the disease are unknown.

Objective

To characterize changes in bone density and structure over the year following diagnosis of T1D and to identify contributors to impaired bone accrual.

Design

Prospective cohort study.

Setting

Academic children’s hospital.

Participants

Thirty-six children, ages 7 to 17 years, enrolled at diagnosis of T1D.

Outcomes

Whole body and regional dual-energy X-ray absorptiometry and tibia peripheral quantitative computed tomography obtained at baseline and 12 months. The primary outcome was bone accrual assessed by bone mineral content (BMC) and areal bone mineral density (aBMD) velocity z score.

Results

Participants had low total body less head (TBLH) BMC (z = −0.46 ± 0.76), femoral neck aBMD (z = −0.57 ± 0.99), and tibia cortical volumetric BMD (z = −0.44 ± 1.11) at diagnosis, compared with reference data, P < 0.05. TBLH BMC velocity in the year following diagnosis was lower in participants with poor (hemoglobin A1c ≥7.5%) vs good (hemoglobin A1c <7.5%) glycemic control at 12 months, z = −0.36 ± 0.84 vs 0.58 ± 0.71, P = 0.003. TBLH BMC velocity was correlated with gains in tibia cortical area (R = 0.71, P = 0.003) and periosteal circumference (R = 0.67, P = 0.007) z scores in participants with good, but not poor control.

Conclusions

Our results suggest that the adverse effects of T1D on BMD develop early in the disease. Bone accrual following diagnosis was impaired in participants with poor glycemic control and appeared to be mediated by diminished bone formation on the periosteal surface.

This longitudinal study in children with new-onset type 1 diabetes found that deficits in bone density were present at diagnosis and that poor glycemic control was associated with lower bone accrual.

Despite technological advances in insulin delivery and glucose monitoring, type 1 diabetes (T1D) remains an incurable disease. Systemic complications of T1D have been recognized for decades and include retinopathy, neuropathy, nephropathy, and cardiovascular disease (1). It is now apparent that the skeleton is also adversely affected by this disease. Multiple epidemiologic studies have shown that T1D is associated with an increased fracture risk that is evident in childhood and continues across the entire lifespan (2–4).

The development of clinical strategies to improve bone health in patients with T1D has been hampered by an incomplete understanding of the factors underlying skeletal fragility in this condition. The insulin receptor is expressed in osteoblasts, and preclinical studies have shown that bone formation is reduced in T1D (5). These findings are supported by a meta-analysis of human studies demonstrating that markers of both bone formation and resorption were lower in participants with T1D compared with healthy controls (6). Mild to moderate deficits in bone mineral density (BMD) have been described in clinical cohorts (7), but few longitudinal studies have been conducted, and the natural history of the effects of T1D on bone mass remains poorly characterized.

The majority of patients with T1D develop the disease during childhood before the attainment of peak bone mass (8). Adolescence is a critical, but time-limited period for bone accrual (9). Therefore, an adverse effect of T1D on bone formation during these years could compromise peak bone mass and thereby contribute to life-long fracture risk. The current literature regarding the effects of T1D on skeletal development during childhood is limited by a lack of studies assessing bone outcomes early in the disease course. Furthermore, the effect of hyperglycemia on bone accrual has not been adequately examined.

The objectives of this study were to assess changes in bone mineral content (BMC), BMD, and bone structure in children over the first year following a diagnosis of T1D and to identify risk factors for impaired bone accrual in this population.

Methods

Participants

Children with incident T1D were recruited from the Children’s Hospital of Philadelphia (CHOP). Study visits were conducted within 30 days of diagnosis, and again 11 to 13 months later. Children aged 5 to 18 years were eligible. Those with a history of metabolic bone, renal, hepatic, or celiac disease, body mass index (BMI) >99th percentile for age and sex, or use of bone active medications (proton pump inhibitors, corticosteroids, growth hormone) in the 6 months before enrollment were excluded. Parental informed consent and participant assent were obtained and the study was approved by the CHOP institutional review board.

Bone and body composition outcomes

Whole body and regional dual-energy X-ray absorptiometry (DXA) scans were performed using a Hologic Delphi (n = 12 scans) or Horizon densitometer (n = 65 scans) (Hologic, Inc, Bedford MA; Apex software, version 13.5). DXA bone outcomes included total body less head (TBLH) BMC (g) and lumbar spine (LS, L1 through L4), left total hip, left femoral neck, and nondominant distal one-third radius areal BMD (aBMD, g/cm²). Fat mass index (FMI) and lean body mass (LBM) index (LBMI) were calculated from whole body fat and lean mass as [fat (or lean) mass (kg)/height (m)²]. Left tibia peripheral quantitative computed tomography (pQCT) scans were obtained using a Stratec XCT2000 (Orthometrix, White Plains, NY) with slice thickness of 2.3 mm and voxel size 0.4 mm as previously described (10). pQCT outcomes included trabecular volumetric BMD (vBMD, mg/cm³) assessed at the 3% metaphyseal site, and cortical vBMD, periosteal circumference (mm), endosteal circumference (mm), and cortical cross-sectional area (mm²) assessed at the 38% diaphyseal site. Quality control of both DXA and pQCT machines was monitored daily using manufacturer-specific phantoms. The coefficient of variation ranged from 1% (spine) to 2.5% (whole body) for DXA and from 0.5% to 1.6% for pQCT outcomes. Insulin pumps were removed before scans if feasible; whole body DXA data were excluded for six participants with tubeless insulin pumps or glucose sensors visible on DXA reports.

Biochemical markers of bone mineral and T1D status

Autoimmunity was confirmed with antiglutamic acid decarboxylase antibody, insulin autoantibody, and insulinoma antigen 2 antibody. Hemoglobin A1c (HbA1c) and c-peptide at diagnosis and 12-month visits were assessed by convenience venous sample in the CHOP laboratory by immunoturbidimetry (Siemens DCA Vantage System, Tarrytown, NY) and chemiluminescence, respectively. Interim HbA1c values from capillary samples at routine clinic visits were secondarily assessed. Poor glycemic control was defined as HbA1c ≥7.5%, consistent with American Diabetes Association guidelines (8). Calcium, magnesium, and phosphorus were performed by an automated clinical chemistry analyzer (Beckman AU680), intact PTH by electrochemoluminescent immunoassay, bone-specific alkaline phosphatase by immunoassay (Ostease reagent), and 1,25-dihydroxy vitamin D, 24,25-dihydroxyvitamin D3, 25-hydroxyvitamin D by immunoaffinity extraction and liquid chromatography-tandem mass spectrometry at the University of Washington (11). Beta-CrossLaps were performed at CHOP by electrochemoluminescent assay (Roche).

Other covariates of interest

Height was measured by stadiometer (Holtrain, Crymych, UK) and weight by digital scale (Scaletronix, White Plains, NY). Pubertal status was determined by palpation of breast tissue (females) and measurement of testicular volume (males) by a pediatric endocrinologist (12). Three-day diet recalls were performed by bionutritionists via telephone. Dietary data were analyzed using the Nutrition Data System for Research, version 2017 (Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN). Fracture was assessed by self-report. Time spent in physical activity (total and high impact) over the past week was estimated using the modified Slemenda questionnaire (13, 14).

Statistical analysis

Height, weight, and BMI were converted into sex specific z scores respective to age using the 2000 Centers for Disease Control and Prevention growth curves (15). Sex-specific z scores respective to age for FMI and LBMI were derived from contemporary National Health and Nutrition Evaluation Survey reference data using methodology developed by the authors (16). To account for interscanner variability in DXA outcomes, a regression equation derived from 33 volunteers scanned on both machines was used to adjust DXA outcomes of participants scanned on the Delphi to match those scanned on the Horizon. DXA BMC and aBMD outcomes were converted into sex- and race-specific (black vs nonblack) z scores respective to age using reference data for healthy children obtained from the Bone Mineral Density in Childhood Study (BMDCS), of which CHOP was a participating center (17). BMC and aBMD z scores were further adjusted for height z score to account for known associations between stature and DXA based z scores in children (17). Outcomes from pQCT were converted into sex- and race-specific z scores respective to age using local reference data collected in >600 healthy children as previously described (10, 18). Tibia cortical geometry outcomes were significantly associated with tibia length (P < 0.02, all measures) and therefore were adjusted for tibia length z score (18, 19).

Whole body and regional bone accrual was assessed by TBLH BMC velocity (change in BMC from baseline to 12-month visit/year) and LS, total hip, and femoral neck aBMD velocity, and expressed as a sex and pubertal stage-specific z score–adjusted for baseline BMC (or aBMD), age, height, and height velocity using reference data from BMDCS, as previously described (20). This method allowed us to relate bone accrual in study participants to that expected for healthy children after accounting for factors known to influence childhood bone accrual. Secondary analyses of bone accrual were performed with additional adjustments for weight and weight velocity and LBM and LBM velocity to test for an effect of body composition on bone accrual.

Statistical analysis was performed using Stata 13.1 (Stata Corp, College Station, TX). Statistical significance was defined by a two-sided P value <0.05 for all analyses. All variables were assessed for normality before analysis and expressed as mean ± SD for parametric and median [interquartile range (IQR)] for nonparametric distributions. Group differences between study participants vs the reference populations were assessed using Student single-sample t test for DXA and pQCT z scores and two-sample t tests vs published BMDCS data for body measures (9); changes in outcomes over the study period were assessed by paired t test or Wilcoxon signed-rank tests. Pearson’s R or Spearman’s ρ were used to assess correlations between linear variables. Multiple linear regression was used to investigate for possible confounding by measured covariates where indicated.

Results

A total of 41 participants (16 female, 78% white) with median age of 13.4 years (range, 7.5 to 17.7) were enrolled and completed the baseline study visit at median 16 (range, 1 to 34) days following diagnosis (Table 1). Of these, 36 participants (88%) completed the 12-month visit. Participants who dropped out were younger and more likely to be of nonwhite racial group compared with those who completed both visits, but did not differ by any other relevant parameter (21). All participants were positive for one or more antibody. Nine participants (27%) had moderate-to-severe diabetic ketoacidosis (serum bicarbonate <15 mmol/L) at presentation. Median HbA1c declined from 12.9% (IQR, 11.0 to 14.0) at diagnosis to 7.0% (IQR, 6.6 to 8.1) at 12 months, P < 0.001. At the 12-month visit, 15 participants (42%) had poor glycemic control by HbA1c. Participants in good vs poor glycemic control did not differ significantly by age, sex, race, BMI, pubertal stage, or growth velocity (Table 1). Mean total daily insulin dose at 12 months was 0.6 ± 0.2 units/kg/d and did not differ by glycemic control. However, c-peptide levels at 12 months were higher in those with good [0.3 (0.2 to 0.6) nmol/L] vs poor [0.2 (0.1 to 0.3) nmol/L] glycemic control, P = 0.02. All participants were started on a multiple daily injection insulin regimen, 31% were on insulin pumps at the 12-month visit. The prevalence of insulin pump use at 12 months did not differ in participants with good vs poor glycemic control (40% vs 20%, P = 0.21). Mean calcium and energy intake were higher (1422 ± 550 mg/d and 2289 ± 549 kcal/d) at baseline compared with 12 months (1292 ± 437 mg/d and 1926 ± 512 kcal/d, P < 0.01 for both), but did not differ by glycemic control at 12 months and were not associated with bone outcomes. Time spent in physical activity (total or high impact) did not differ by glycemic control, and was not correlated with insulin dose, insulin level, c-peptide level, or bone accrual.

Table 1.

Characteristics of 36 Participants With Incident Type 1 Diabetes Completing Both Study Visits, Overall and Stratified by Glycemic Control at 12 mo

	Entire Cohort, n = 36		Good Glycemic Control (HbA1c <7.5%), n = 21		Poor Glycemic Control (HbA1c ≥7.5%), n = 15
	Baseline	12 mo	Baseline	12 mo	Baseline	12 mo
Female, n (%)	14 (39)	14 (39)	10 (48)	10 (48)	4 (27)	4 (27)
Age, y (range)	14.2 (12.1–15.4)a	15.2 (13.0–16.5)	13.9 (12.0–15.0)	14.9 (13.0–16.0)	14.9 (12.2–16.0)	15.9 (13.2–16.9)
Pubertal stage, n (%)
1	6 (16.7)	5 (13.9)	4 (19)	4 (19.1)	2 (14.3)	1 (6.7)
2	4 (11.1)	1 (2.8)	1 (4.8)	0 (0)	3 (21.4)	1 (6.7)
3	6 (16.7)	2 (5.6)	5 (23.8)	1 (4.8)	1 (7.1)	1 (6.7)
4	10 (27.8)	7 (19.4)	5 (23.9)	4 (19.1)	5 (35.7)	3 (20)
5	9 (25)	21 (58.3)	6 (28.6)	12 (57.1)	3 (21.4)	9 (60)
Body mass index z score	0.3 (─0.5 to 0.8)	0.5 (─0.1 to 1.2)b	0.5 (─0.9 to 0.8)	0.7 (0.1–1.2)b	0.3 (─0.1 to 0.8)	0.3 (─0.1 to 1.1)
Growth velocity, cm/y	N/A	3.8 (1.2–6.9)	N/A	4.2 (1.3–6.9)	N/A	2.9 (0.5–6.4)
Racial group, n (%)
White	30 (83.4)	30 (83.4)	20 (95.2)	20 (95.2)	10 (66.7)	10 (66.7)
Black	3 (8.3)	3 (8.3)	0 (0)	0 (0)	3 (20)	3 (20)
Hispanic/Latino	3 (8.3)	3 (8.3)	1 (4.8)	1 (4.8)	2 (13.3)	2 (13.3)
Fracture history, n (%)	16 (44.4)	3 (8.3)c	9 (42.3)	2 (9.5)c	7 (46.7)	1 (6.7)c
Long bone	7 (43.8)	0 (0)	4 (44.4)	0 (0)	3 (42.9)	0 (0)
Other	9 (56.2)	3 (100)	5 (55.6)	2 (100)	4 (57.1)	1 (100)
Physical activity, h/d	0.9 (0.5–1.2)	1.1 (0.4–1.9)	1.1 (0.6–1.2)	1.3 (0.6–1.9)	0.8 (0.1–1.1)	0.3 (0.2–1.9)
High impact physical activity, h/d	0.3 (0.0–0.6)	0.5 (0.1–1.0)	0.3 (0.0–0.5)	0.6 (0.2–0.9)	0.3 (0.1–0.6)	0.2 (0–1.7)
Calcium intake, mg/d	1422 ± 550	1175 ± 385b	1444 ± 465	1161 ± 402b	1393 ± 663	1196 ± 374b
Hemoglobin A1c, %	12.9 (11–14)	7.0 (6.6–8.1)b	13.0 (11.6–14)	6.7 (5.8–6.9)b	11.2 (10.5–14)	8.9 (7.7–10.5)b,d
A1c ≥ 7.5	36 (100)	15 (41.7)	21 (100)	0 (0)	15 (100)	15 (100)
≥1 T1D ab positive, n (%)e	36 (100)	N/A	21 (100)	N/A	15 (100)	N/A
3 T1D ab positive, n (%)e	14 (38.9)	N/A	9 (42.9)	N/A	5 (33.3)	N/A

^aMedian (IQR), all such values.

^b P < 0.05, 12 mo compared with baseline.

^cIncident fracture since T1D diagnosis.

^d P < 0.05, poor glycemic control compared with good glycemic control.

^eGlutamic acid decarboxylase antibody, insulin autoantibody, or insulinoma antigen 2 antibody.

Characterization of DXA and body measure outcomes

T1D participants had significantly lower TBLH BMC (z score = −0.46 ± 0.76, P = 0.001) and femoral neck aBMD (z score = −0.57 ± 0.99, P = 0.002) at baseline compared with the BMDCS reference data; aBMD at other skeletal sites was not different. Laboratory markers of diabetes severity (HbA1c, glucose, insulin, c-peptide, bicarbonate) at presentation were not significantly associated with baseline BMC or aBMD z score at any skeletal site. Femoral neck aBMD z score significantly increased over the study period (mean change = 0.15 ± 0.37, P = 0.03); significant changes in BMC or aBMD z scores were not present at other sites (Table 2).

Table 2.

DXA and Tibia pQCT z Scores in 36 Participants With Incident T1D, Overall and Stratified by Glycemic Control

	Baseline	P a	12 mo	P a	12-mo Change	P (Change From Baseline)b
Entire Cohort (n = 36)
TBLH BMC z score	−0.46 ± 0.76	0.001	−0.28 ± 0.72c	0.04	0.08 ± 0.28c	0.11
LS aBMD z score	−0.25 ± 0.88	0.09	−0.29 ± 0.84	0.04	−0.04 ± 0.30	0.40
Total hip aBMD z score	−0.21 ± 0.99	0.22	−0.16 ± 0.96d	0.34	0.01 ± 0.3d	0.86
Femoral neck aBMD z score	−0.57 ± 0.99	0.002	−0.39 ± 0.92d	0.02	0.15 ± 0.37d	0.03
Distal 1/3 radius aBMD z score	0.09 ± 1.27	0.67	−0.03 ± 1.18e	0.87	−0.08 ± 0.48e	0.36
Tibia cortical vBMD z score	−0.44 ± 1.11	0.02	−0.41 ± 1.09	0.03	0.04 ± 0.53	0.69
Tibia cortical area z score	0.01 ± 0.95	0.97	−0.01 ± 0.94	0.98	−0.01 ± 0.29	0.85
Tibia periosteal circ. z score	−0.12 ± 0.91	0.42	−0.14 ± 0.87	0.31	−0.03 ± 0.20	0.44
Tibia endosteal circ. z score	−0.25 ± 1.08	0.18	−0.28 ± 1.09	0.13	−0.03 ± 0.17	0.24
Tibia trabecular vBMD z score	−0.39 ± 1.28	0.08	−0.25 ± 1.15e	0.22	0.06 ± 0.43e	0.41

	Baseline	P f	12 mo	P f	12-mo Change	P b /Pf
Good glycemic control (n = 21)
TBLH BMC z score	−0.72 ± 0.76	0.02	−0.41 ± 0.66c	0.26	0.21 ± 0.27c	0.001/0.01
LS aBMD z score	−0.36 ± 0.92	0.38	−0.38 ± 0.86	0.47	−0.02 ± 0.27	0.74/0.59
Total hip aBMD z score	−0.41 ± 0.92	0.15	−0.32 ± 0.86g	0.26	0.13 ± 0.25g	0.03/0.002
Femoral neck aBMD z score	−0.77 ± 1.06	0.14	−0.52 ± 0.94g	0.35	0.30 ± 0.36g	0.001/0.002
Distal 1/3 radius aBMD z score	−0.21 ± 1.35	0.09	−0.32 ± 1.16g	0.10	−0.01 ± 0.47g	0.99/0.29
Tibia cortical vBMD z score	−0.46 ± 1.11	0.89	−0.40 ± 1.12	0.98	0.06 ± 0.53	0.59/0.74
Tibia cortical area z score	−0.09 ± 0.94	0.48	0.01 ± 0.97	0.93	0.10 ± 0.30	0.14/0.005
Tibia periosteal circ. z score	−0.29 ± 0.91	0.19	−0.24 ± 0.92	0.45	0.05 ± 0.20	0.27/0.006
Tibia endosteal circ. z score	−0.44 ± 1.15	0.20	−0.49 ± 1.11	0.18	−0.04 ± 0.19	0.28/0.58
Tibia trabecular vBMD z score	−0.76 ± 1.20	0.04	−0.43 ± 1.05g	0.27	0.20 ± 0.46g	0.07/0.03
Poor glycemic control (n = 15)
TBLH BMC z score	−0.10 ± 0.72	0.02	−0.14 ± 0.78	0.26	−0.04 ± 0.23	0.50/0.01
LS aBMD z score	−0.10 ± 0.81	0.38	−0.18 ± 0.82	0.47	−0.07 ± 0.34	0.41/0.59
Total hip aBMD z score	0.09 ± 1.07	0.15	0.05 ± 1.07h	0.26	−0.17 ± 0.30h	0.06/0.002
Femoral neck aBMD z score	−0.27 ± 0.82	0.14	−0.22 ± 0.90h	0.35	−0.08 ± 0.27h	0.28/0.002
Distal one-third radius aBMD z score	0.52 ± 1.06	0.09	0.35 ± 1.14h	0.10	−0.18 ± 0.50h	0.19/0.29
Tibia cortical vBMD z score	−0.41 ± 1.15	0.89	−0.41 ± 1.08	0.98	−0.01 ± 0.57	0.99/0.74
Tibia cortical area z score	0.14 ± 0.98	0.48	−0.02 ± 0.93	0.93	−0.16 ± 0.19	0.005/0.005
Tibia periosteal circ. z score	0.11 ± 0.87	0.19	−0.02 ± 0.81	0.45	−0.13 ± 0.14	0.004/0.006
Tibia endosteal circ. z score	0.03 ± 0.93	0.20	0.01 ± 1.02	0.18	−0.02 ± 0.13	0.66/0.58
Tibia trabecular vBMD z score	0.13 ± 1.26	0.04	0.01 ± 1.27	0.27	−0.12 ± 0.31	0.15/0.03

DXA bone z scores adjusted for height z score, tibia cortical area, periosteal circumference, endosteal circumference, fat area, and muscle area adjusted for tibia length z score.

Abbreviation: circ., circumference.

^aVs BMDCS reference population, using one-sample t test.

^b12 mo vs baseline in T1D, paired t test.

^cExcludes 6 participants with visible pump or continuous glucose monitor on 12-mo scan.

^dn = 34.

^en = 35.

^fParticipants with good vs poor glycemic control.

^gn = 20.

^hn = 14.

Mean height and BMI z scores of T1D participants at baseline did not differ significantly from the BMDCS reference data (9). Weight (mean change in z score 0.28 ± 0.49, P = 0.002), BMI (mean change in z score 0.31 ± 0.56, P = 0.003), and FMI (mean change in z score 0.26 ± 0.37, P = 0.001) z scores all increased over the study; however, significant changes in height- or LBMI z scores were not observed (21).

Impact of glycemic control on bone accrual

Whole body and regional bone accrual (assessed by BMC and aBMD velocity z scores) in the 12 months following diagnosis for the entire cohort are shown in Fig. 1A and in the online repository (21). Mean BMC and aBMD velocity z scores did not differ significantly from the BMDCS reference data at any skeletal site. However, when stratified by glycemic control at 12 months, participants with poor control (HbA1c ≥7.5%) had significantly lower bone accrual for TBLH BMC, total hip aBMD, and femoral neck aBMD compared with those with good control (HbA1c <7.5%), P = 0.001, 0.002, and 0.002, respectively [Fig. 1B; (21)]. Additionally, greater reductions in HbA1c from baseline to 12 months were associated with greater TBLH BMC velocity z score (R = 0.44, P = 0.02), greater total hip aBMD velocity z score (R = 0.53, P = 0.002), and greater femoral neck aBMD velocity z score (R = 0.55, P = 0.001). The statistically significant negative relationship among TBLH, total hip, and femoral neck bone accrual and poor glycemic control was confirmed in sensitivity analyses that accounted for an effect of DXA machine, categorized glycemic control by mean HbA1c (including clinical measurements) over the study period, or defined bone accrual as the 12-month change in BMC or aBMD z score adjusted for height z score and relevant covariates (instead of BMC or aBMD velocity) (21).

Figure 1.

Bone accrual as assessed by 12-month BMC (TBLH) or aBMD (lumbar spine, total hip, femoral neck, distal one-third radius) velocity z scores (A) in the entire cohort or (B) stratified by glycemic control. Participants with poor glycemic control (HbA1c ≥7.5%) had significantly lower accrual of TBLH BMC (P = 0.0001), total hip aBMD (P = 0.003), and femoral neck aBMD (P = 0.002) compared with those with good control. Velocity z scores respective to sex and Tanner stage and adjusted for baseline BMC (aBMD), age, height, and height velocity. *Statistically significant difference in bone accrual by glycemic control, P < 0.05. Fem neck, femoral neck; rad, distal one-third radius; tot hip, total hip.

Because of a discrepancy between the 12-month changes in weight (significantly increased) vs height and lean mass z scores (no significant change), secondary analyses of BMC and aBMD velocity were undertaken that included further adjustments for weight + weight velocity or LBM + LBM velocity. Further adjustment for weight resulted in significantly lower bone velocity z scores for TBLH BMC, LS aBMD, total hip aBMD, and femoral neck aBMD compared with adjustment for height alone or height + LBM; further adjustment for LBM additionally resulted in higher LS aBMD velocity compared with height alone [Fig. 2; (21)]. The primary finding of lower TBLH, total hip, and femoral neck bone accrual in participants with poor glycemic control at 12 months persisted with further adjustment for either weight or LBM.

Figure 2.

Comparison of 12-month BMC (TBLH) or aBMD (LS, total hip, femoral neck, distal one-third radius) velocity z scores adjusted for (i) height + height velocity, (ii) height, height velocity + weight and weight velocity, or (iii) height, height velocity + LBM and LBM velocity. Additional adjustment for weight resulted in lower bone accrual vs compared with height alone or height + LBM for TBLH (P = 0.01 and 0.001, respectively), LS (P = 0.006 and 0.001, respectively), total hip (P = 0.003 and 0.005, respectively), and femoral neck (P = 0.002 and 0.001, respectively). Additionally, height + LBM adjustment resulted in greater LS aBMD velocity z score compared with height adjustment alone (P = 0.005). Ht, height; Vel, velocity; adj, adjusted; wt, weight. *Statistically significant difference in adjusted BMC or aBMD velocity z scores, P < 0.05.

Relationships Between DXA and pQCT bone outcomes

At the baseline visit, mean tibia cortical and trabecular vBMD z scores were −0.44 ± 1.11 and −0.39 ± 1.28, respectively. Cortical vBMD z score was significantly lower than the local reference data, P = 0.02. Neither outcome changed significantly over the study (21). Correlations between TBLH BMC velocity z score and change in tibia pQCT vBMD and cortical geometry z scores are shown in Fig. 3. In participants with good glycemic control at 12 months, TBLH BMC velocity z score was negatively correlated with change in cortical vBMD z score (R = 0.51, P = 0.05) and positively correlated with change in cortical area (R = 0.71, P = 0.003) and periosteal circumference (R = 0.67, P = 0.007) z scores; there was no correlation with change in endosteal circumference z score. There were no correlations between TBLH BMC velocity and any pQCT outcomes in participants with poor glycemic control.

Figure 3.

Pearson correlations between TBLH BMC velocity z score and 12-month change in (A) tibia cortical vBMD z score, (B) tibia cortical area z score, (C) tibia periosteal circumference z score, and (D) tibia endosteal circumference z score. TBLH velocity z scores respective to sex and Tanner stage and adjusted for baseline BMC (aBMD), age, height, height velocity; cortical area, periosteal circumference, and endosteal circumference z scores adjusted for tibia length z score.

Biochemical markers of bone mineral metabolism

Biochemical markers of bone and mineral metabolism are summarized in the online repository (21). The prevalence of vitamin D deficiency (25-OH vitamin D <50 nmol/L) was 5% at baseline and 8% at 12 months. Mean 25-OH vitamin D levels at 12 months were lower in participants with poor vs good glycemic control, 60.1 ± 13.7 vs 80.1 ± 19.9 nmol/L, P = 0.001. Additionally, 25-OH vitamin D level at 12 months was positively correlated with adjusted TBLH BMC velocity (R = 0.37, P = 0.04), change in adjusted tibia cortical area (R = 0.35, P = 0.03), and change in tibia periosteal circumference (R = 0.34, P = 0.04) z scores. To determine if the associations between poor glycemic control and bone accrual were confounded by vitamin D, linear regression models that included both variables were evaluated. In all cases, the negative association between poor glycemic control at 12 months and bone accrual remained after adjustment for the 12-month 25-OH vitamin D level (21). Analysis of correlations between bone mineral markers other than 25-OH vitamin D did not identify statistically significant relationships with HbA1c or bone outcomes.

Discussion

This study prospectively evaluated changes in bone density and structure occurring immediately after T1D diagnosis in children. On average, study participants had lower TBLH BMC, femoral neck aBMD, and tibia cortical vBMD at presentation of diabetes compared with reference data from healthy children. Furthermore, we found that rates of whole body, femoral neck, and total hip bone accrual in the year following diagnosis were impaired in participants with persistently poor glycemic control.

Few studies have investigated skeletal outcomes early in the course of T1D. A study in 32 young adults with new-onset T1D identified deficits in femoral neck (−0.38 ± 1.0) and LS (−0.61 ± 1.2) aBMD z scores (22), but did not evaluate bone accrual. A study in children that included 14 participants with DXA scans done within 3 months of diagnosis reported a mean LS aBMD z score of −0.65 ± 1.1 (23). The magnitudes of the previously described deficits were similar to our findings, and together suggest that the adverse effects of T1D on bone density arise early in the disease course. The natural history of T1D is characterized by a preclinical stage of autoimmunity and dysglycemia (24). We were unable to detect a clear association between biochemical markers of diabetes severity and degree of bone deficit at diagnosis. This might be the result of unmeasured factors including BMC and BMD before onset of autoimmunity, duration of preclinical disease, and degree of weight loss before presentation. It could also be related to the use of an HbA1c assay with an upper reporting limit of 14%, and convenience sampling of insulin and c-peptide. Future studies of bone accrual during the preclinical phase of T1D are needed.

Hyperglycemia is a defining characteristic of diabetes and contributes to the development of traditional diabetic complications (25). Preclinical studies have shown that exposure to hyperglycemia reduced the expression of genes involved in osteoblast mediated bone formation (26) and also impaired osteogenesis (27). The relationship between hyperglycemia and skeletal complications in humans is less clear. Relationships between HbA1c and fracture risk in T1D have been inconsistent in epidemiologic studies (3, 28). Longitudinal studies that investigated the effect of glycemic control and bone accrual in children have been similarly inconclusive. A 12-month study in adolescents with prevalent T1D reported an inverse relationship between HbA1c and changes in whole body BMC and tibia cortical area (29), whereas two others found no relationship between glycemic control and bone outcomes (30, 31). Our study included participants with a range of glycemic control at 12 months (range, 5.5% to 13.5%) and showed that bone accrual in participants with good glycemic control was not different from healthy reference data. A narrower range of glycemic control among participants in prior studies may have contributed to the perceived absence of an association between HbA1c and bone accrual (31). We defined poor glycemic control according to a widely accepted pediatric HbA1c target established to minimize the likelihood of diabetes complications. However, the level of hyperglycemia at which bone is adversely affected is not known. Larger studies including participants with a range of glycemic control will be required to determine this threshold.

Previously published pQCT cross-sectional data in children and adolescents with T1D have consistently demonstrated deficits in cortical area, whereas the results for cortical or trabecular vBMD have been mixed (29, 32–34). Cortical outcomes improved over time in two studies that reported longitudinal data (29, 31). Our study builds on previous reports by relating changes in tibia pQCT cortical and trabecular bone outcomes to whole body bone accrual in an attempt to investigate the mechanistic effects of T1D on the maturing skeleton. In participants with good glycemic control, the finding that TBLH BMC velocity z scores were positively associated with changes in cortical area and periosteal circumference, but negatively associated with changes in cortical vBMD is consistent with a pattern of skeletal recovery observed in other pediatric diseases (35). On the contrary, the absence of these relationships in participants with poor glycemic control suggests that hyperglycemia resulted in impaired bone formation (as opposed to increased bone resorption, which would have been suggested by increases in endosteal circumference). Correlation analyses between bone accrual and pQCT outcomes by glycemic group were conducted secondarily and not accounted for in a priori power calculations. Therefore, we cannot rule out the possibility that the absence of an significant correlation in participants with poor glycemic control was the result of insufficient sample size. However, these findings corroborate those from preclinical studies in which reduced bone formation in T1D rodents has been confirmed by histomorphometry and distraction osteogenesis (5, 36).

The primary findings of this study did not extend to LS or radius aBMD. Deficits in LS aBMD have been previously documented in children with T1D (37). Our data revealed a near-significant (P = 0.09) deficit at diagnosis, so it is likely that the spine was negatively affected. More surprising was the relatively low LS aBMD velocity, especially given the high rate of bone accrual observed at the femoral neck (another trabecular site). Vertebral fractures can alter LS aBMD and have been reported in young adults with T1D (38). We did not perform lateral spine imaging so were unable to investigate for this possibility. We found no evidence of a T1D effect at the distal one-third radius. There was greater variability in the height-adjusted aBMD z scores at this site, raising the possibility of an unexpected confounder or measurement artifact. Notably, the risk of hip fracture in T1D is elevated disproportionally to other skeletal sites (39). The fracture distribution also differs from healthy people, in which the proportion of fractures occurring in lower vs upper extremities is greater in people with T1D (3). A recent preclinical study found evidence of impaired bone mechanosignaling in the hyperglycemic environment (40). This raises the possibility of diminished bone formation at weight-bearing skeletal sites in association with T1D, however this hypothesis requires further study.

The strengths of this study included the use of BMC and aBMD velocity z scores that were derived from a robust dataset of healthy children and accounted for factors known to affect bone accrual including age, sex, growth, pubertal stage, and baseline BMC (or aBMD). BMC and aBMD are cumulative outcomes; therefore, the assessment of bone accrual allows for a more sensitive evaluation of bone health status than can be obtained from a cross-sectional study. The results of sensitivity analyses that included further adjustment of BMC and aBMD velocity for changes in body composition and the alternative assessment of bone accrual by absolute change in BMC or aBMD z score support the robustness of the observed relationship between glycemic control and bone accrual.

Limitations included an inadequate sample size to fully explore for differences in bone accrual by sex and racial/ethnic group and their potential effect on associations with glycemic control. DXA and pQCT bone z scores were sex and race specific to take into account the skeletal effects of these variables, but secondary analyses by sex and racial group were not possible. Similar to our findings, previous studies have shown that black vs white racial group is associated with poorer glycemic control (41) and lower vitamin D levels (42). Both of these factors appeared to be risk factors for impaired bone accrual in our study. Therefore, our findings should be interpreted with caution, because all three black participants were in the group with low BMD accrual. We did not collect data on socioeconomic status or environmental factors beyond nutritional intake and physical activity. The absence of detectable relationships between markers of bone turnover and skeletal outcomes or glycemic control may have been due to inadequate power given the effects of sex and growth on these analytes (43). Participants with good glycemic control at 12 months had lower BMC and aBMD at diagnosis, which may have contributed to their greater bone accrual. Statistical models were adjusted for baseline measures to account for this possibility. However, validation of these findings in a larger sample is required. Deficits in aBMD are associated with an increased risk of fracture in children (44), but may not fully explain the high fracture risk observed in T1D (45). Deficits in bone strength that arise from impaired bone quality may additionally contribute to diabetes related fracture risk (46). We did not assess advanced glycation end products or use techniques that allow for investigation of bone quality.

In summary, our data add to the existing knowledge of the adverse skeletal effects of T1D in two important ways. First, we found that children with T1D had detectable deficits in bone density at the time of diagnosis, suggesting that the negative effect on bone health emerges very early in the disease course. Second, we found that poor glycemic control was associated with lower rates of bone accrual in the first year following diagnosis. Further longitudinal studies of longer duration are needed to determine if early and sustained reductions in HbA1c are necessary and/or sufficient to normalize bone accrual and reduce fracture risk in patients with T1D.

Glossary

Abbreviations:

aBMD

areal bone mineral density

BMD

bone mineral density

BMC

bone mineral content

BMDCS

Bone Mineral Density in Childhood Study

BMI

body mass index

CHOP

Children’s Hospital of Philadelphia

DXA

dual-energy X-ray absorptiometry

FMI

fat mass index

HbA1c

hemoglobin A1c

IQR

interquartile range

LBM

lean body mass

LBMI

lean body mass index

LS

lumbar spine

pQCT

peripheral quantitative computed tomography

T1D

type 1 diabetes

TBLH

total body less head

vBMD

volumetric bone mineral density