If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author: Scott M. Tintle, MD, Orthopaedics, USU-Walter Reed Department of Surgery, Walter Reed National Military Medical Center, 8901 Rockville Pike, Bethesda, MD 20889.
The objective of this study was to analyze changes in serum markers of bone turnover across multiple decades in osteoporotic women compared with nonosteoporotic controls, to determine their utility as potential predictors for osteoporosis. Early prediction of those at risk for osteoporosis can enable early intervention before the irreversible loss of critical bone mass.
Methods
Serum samples were obtained from 20 women given the diagnosis of osteoporosis after age 46 years and 20 age-matched women with normal bone mineral density from 4 time points in their life (ages 25–31, 32–38, 39–45, and 46–60 years). Serum levels of bone turnover markers (propeptide of type I collagen, parathyroid hormone, bone-specific alkaline phosphatase, osteocalcin, C-terminal telopeptide of type I collagen, sclerostin, osteoprotegerin, osteopontin, and 25-OH vitamin D) were measured using commercially available arrays and kits. We used logistic regression to assess these individual serum markers as potential predictors of osteoporosis, and mixed-effects modeling to assess the change in bone turnover markers between osteoporotic and control groups over time, then performed fivefold cross-validation to assess the classification ability of the models.
Results
Markers of bone turnover, bone-specific alkaline phosphatase, C-terminal telopeptide of type I collagen, sclerostin, and osteocalcin were all independent predictors at multiple time points; osteopontin was an independent predictor in the 39- to 45-year age group. Receiver operating characteristic analyses demonstrated moderately strong classification ability at all time points. Sclerostin levels among groups diverged over time and were higher in the control group than the osteoporotic group, with significant differences observed at time points 3 and 4.
Conclusions
Serum markers of bone turnover may be used to estimate the likelihood of osteoporosis development in individuals over time. Although prospective validation is necessary before recommending widespread clinical use, this information may be used to identify patients at risk for developing low bone mineral density long before traditional screening would ostensibly take place.
Osteoporosis and low bone mass currently affect over 54 million people in the United States. There are over 2 million associated fragility fractures per year, which is expected to continue to rise.
The current diagnosis of osteoporosis is centered on identifying patients with decreased bone mineral density (BMD) through dual-energy x-ray absorptiometry (DEXA) scans. Dual-energy x-ray absorptiometry screening is normally initiated at age 65 years; however, bone loss may begin in early adulthood, and several authors have reported on fragility fractures in premenopausal women, which suggests that manifestations of the condition can begin much earlier than the recommend age for screening.
Wrist fracture as a predictor of future fractures in younger versus older postmenopausal women: results from the National Osteoporosis Risk Assessment (NORA).
Although fragility fractures themselves are not often fatal, they are temporally associated with a substantial risk for mortality, as much as 39% in women within 5 years of sustaining the first such fracture.
Despite the serious nature of the condition, prior studies demonstrated that only 30% of women and 15% of men subsequently undergo diagnostic testing or treatment after sustaining a fragility fracture.
Degraded bone microarchitecture in the distal radius was demonstrated in premenopausal women with distal radius fractures (DRF) compared with nonfractured control subjects, even without significantly different (P > .05) DEXA scores between the 2 groups.
This demonstrates a need to develop new diagnostic strategies to facilitate early detection of those at risk for osteoporosis and to create opportunities for early intervention.
The use of serum markers of bone turnover as a method to identify individuals with an increased rate of bone loss is a promising direction. Postmenopausal women with high levels of bone turnover markers demonstrated an increased rate of bone loss and increased risk for fracture, and bone turnover markers were shown to be elevated in premenopausal and post-menopausal women with fragility fractures.
This demonstrates that clinically relevant bone loss and its sequelae can manifest at a young age, and that biochemical evidence of increased bone turnover is already detectable in this population. However, a limitation of much previous research in this area was the evaluation of serum markers of bone turnover using a single serum sample, which was drawn after the patient had already demonstrated clinically relevant bone loss by sustaining a fragility fracture. Table 1 provides a brief description of bone turnover markers and their function.
A more useful clinical tool would detect patients with increased rates of bone loss early in life. The objective of this study was to compare patients with osteoporosis and matched controls without osteoporosis with regard to serum markers of bone turnover across multiple decades of the adult life span. We hypothesized that women with osteoporosis, as diagnosed on DEXA scan, would have different biomarker profiles compared with nonosteoporotic controls.
Materials and Methods
Study design and patient populations
This study was approved by the Uniformed Services University Institutional Review Board. We used the Department of Defense Serum Repository (DoDSR) to obtain patient serum samples. The DoDSR was initially established in 1985 and contains stored frozen serum samples from service members obtained at entry into the military before and after deployment, and at regular 2-year intervals throughout military service.
Because samples are collected routinely throughout the career of military service members, including upon their entry into the military, this represents an opportunity to probe serum samples across multiple decades of the adult life span of an individual of interest. In a case control design, we queried the Military Health System Management and Reporting Tool (M2) database for all female patients with a Current Procedural Terminology code indicating a DEXA scan obtained between ages 46 and 60 years to ensure the availability of serum in the DoDSR during the age range of interest. The individual electronic medical records of these patients were examined in the Armed Forces Health Longitudinal Technology Application, the military-wide electronic medical record system. We defined 2 groups: patients who met criteria for osteoporosis (DEXA T score less than –2.5 at one or more sites) were categorized as osteoporotic, and patients with normal BMD (DEXA T score greater than –1.0, nonosteoporotic, and nonosteopenic) were categorized as controls. Patients were excluded as potential subjects if any of the following criteria were present: known endocrinopathies; metabolic bone disease; or use of glucocorticoids, hormone replacement therapy, aromatase inhibitors, selective estrogen receptor modulators, teriparatide, denosumab, or bisphosphonates.
Eligible subjects in these groups were given to the DoDSR for the selection of a convenience sample of 20 osteoporotic patients and 20 control patients, matched by age and body mass index (BMI), and based on the availability of serum samples within each of the 4 age ranges of 25 to 31, 32 to 38, 39 to 45, and greater than 46 years (between age 46 and the time of the DEXA scan), thus indicating the subject was on active duty over those 4 age groups. We refer to these age ranges as time points 1 to 4, respectively. Although subjects were matched for age and BMI, because of deidentification protocols for serum samples, detailed patient demographic and clinical information was not returned with the serum samples.
Serum analysis
The DoDSR consists of approximately 60 million serum samples originally collected for mandatory HIV testing in armed forces personnel.
After collection, samples were stored at –30°C. Deidentified serum samples for the study subjects were transported from the DoDSR to Uniformed Services University in a sealed shipping container containing dry ice. A single aliquot of a 0.5-mL serum sample was obtained for each subject at each time point. Each serum specimen was further aliquoted into 5 tubes of 0.1 mL each and stored at –80oC until the time of analysis and were not subjected to additional freeze–thaw cycles. Parathyroid hormone (PTH), osteocalcin (OC), sclerostin (SOST), osteoprotegerin, and osteopontin (OPN) were analyzed in duplicate using a commercial multiplex Luminex array (HBNMAG-51K-13, MILLIPLEX MAP Human Bone Magnetic Bead Panel, Millipore Sigma). The multiplex assay was carried out according to the manufacturer’s instructions. Median fluorescence intensities were collected on a Bio-Plex 200 System (BioRad), using Milliplex Analyst analysis software (version 5.1). Analyte concentrations were calculated from the appropriate standard curve using 5-point curve fitting to transform mean fluorescence intensity into concentration. In addition, each sample specimen was analyzed in duplicate for N-terminal extension propeptide of type I collagen (MBS9314633, MyBioSource Inc), bone-specific alkaline phosphatase (MBS724100, MyBioSource, Inc), C-terminal telopeptide of type I collagen (CTX-1) (NBP2-69073, Novus Biologicals), and 25-OH vitamin D (RDKAP1971, R&D Systems, Inc) using commercially available enzyme-linked immunosorbent assays, run according to the manufacturer's instructions. Samples were analyzed in batch to minimize variability. Absorbance was measured using a microplate reader (Tecan Infinite 200 Pro, Tecan Group Ltd). Standard curves were generated for these analytes using the premixed lyophilized standards provided in the kits. Analyte concentrations in each sample were calculated from the appropriate standard curve using 5-point curve fitting to transform mean fluorescence intensities into concentrations.
Statistical analysis
For this pilot study, we selected a convenience sample size of 20 subjects/group based on previous between-group comparisons performed on similar analytes in fragility fracture populations.
All values were log-transformed to approximate normality better. Multivariable modeling was performed for samples at each time point using a logistic regression model. The outcome of interest was osteoporosis, and the predictor variables were the biomarkers described earlier. We selected analytes for the model at each time point based on P value filtering (P < .05) after first applying the false discovery rate method. The model was internally validated using fivefold cross-validation, a standard technique used to assess predictive models using resampling on a limited dataset to reduce the risk for bias and overfitting. Calibration of each model was assessed using calibration curves. Receiver operating characteristic analysis was performed for the model at each time point to assess its discriminatory performance. Finally, to evaluate individual serum markers among groups along with the change over time, a linear mixed-effects model using the fixed effects of study group and age and a random effect of participant was used. Planned group comparisons were performed at individual time points using t tests.
Results
Multivariable models
Bone-specific alkaline phosphatase, CTX-1, SOST, and OC were all independent predictors at multiple time points; OPN was an independent predictor at time point 3 (Table 2). Calibration curves for each model are demonstrated in Figure 1. Receiver operator characteristic analyses demonstrated moderately strong performance at time points 1 to 4 (area under the curve [95% confidence interval]) = 0.82 (0.69–0.95), 0.78 (0.63–0.92), 0.91 (0.82–1.0), and 0.87 (0.76–0.99), respectively) (Fig. 2).
Table 2Results of Multivariable Logistic Regression
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
(P < .05)
0.87 (0.76–0.99)
∗ Independent predictors of osteoporosis. Area under the curve value indicates the performance of the model developed for each time point on receiver operating characteristic analysis.
Figure 1Calibration curves for the model developed for each time point, demonstrating the comparison of observed (black line + shaded 95% CI) to predicted outcomes (dashed line).
Several serum markers of bone turnover demonstrated statistically significant changes over time (CTX-1, P < .05; SOST, P < .05; OC, P < .05; PTH, P < .05; vitamin D, P < .05; OPN, P < .05). However, SOST was the only biomarker that demonstrated between-group differences (P < .05), with generally higher levels in control patients than in osteoporotic patients. It appeared that this trend of larger SOST values in control patients emerged over time, and that osteoporotic subjects had a smaller increase in SOST levels with age. Time points 3 (P < .05) and 4 (P < .05) had significantly different levels of SOST between groups (Fig. 3). Figure E1 and Table E1 (available on the Journal’s Web site at www.jhandsurg.org) illustrate comparisons among groups over time for all markers. In addition, results from t test comparisons at each time point for each marker are shown in Table 1 and Table E1.
Figure 3SOST levels over time between subjects with and without an eventual diagnosis of osteoporosis. (P < .05 between groups overall; P < .05 for Time Point 3 [Age 39-45]; and P < .05 for Time Point 4 [Age 46-60]. ∗ Group difference at time point).
The goal of this study was to explore changes in circulating serum markers of bone turnover in women at risk for osteoporosis and the potential of using biomarker signatures as possible early indicators of osteoporosis. We found a divergent trend in SOST levels between osteoporotic and nonosteoporotic women over time, with significant differences between groups at ages 39 to 45 years and greater than 46 years. Although we found significant changes in other markers over time, distinct trends between osteoporotic and nonosteoporotic women were not demonstrated. This was a preliminary analysis in small groups with a high prevalence of the outcome of interest; nevertheless, using multiple markers of bone turnover, we identified independent predictors and generated models with good classification ability. Our models demonstrated strong discriminatory ability and clinical utility for the prediction of osteoporosis in patients as young as age 25.
This study had several limitations. The sample size was limited and increased the risk for overfitting, the process of modeling noise within the training data. This study was designed to test the early diagnostic capability of these biomarker profiles as they relate to osteoporosis; subsequent investigations must be performed with a larger number of subjects, and they should go through a validation process before clinical use. Acquiring samples from the DoDSR allowed us to perform this study in a longitudinal nature on the same subjects over multiple decades; however, this study design had some limitations. Each biomarker exhibits different degrees of stability both in vivo and in vitro, and analysis of frozen serum creates the potential for some sample degradation over time. Nonetheless, as reported by Hassis et al,
human plasma proteins are stable for multiple decades in frozen storage, and the main risk to sample integrity is prolonged time to processing and repeated freeze–thaw cycles. The samples in this study were procured, processed, and archived under stringent Department of Defense protocols and not subject to repeated freeze–thaw cycles. Therefore, we do not suspect sample quality was degraded to a degree that would affect the results. In addition, the DoDSR collection protocol provides for less control over specific conditions (ie, seasonality, time of day, specific patient demographics) under which serum samples were provided. Variations in hormonal changes throughout the menstrual cycle and transition to menopause have an intricate relation with bone physiology and biochemistry, and represent another factor that could potentially influence our results.
Because of the small exploratory design of this pilot study and these limitations, we do not recommend clinical decision-making based on these data. However, the models presented here show preliminary evidence that serum biomarkers of bone turnover may be associated with risk for osteoporosis development, and showcase a potential avenue of investigation for the future development of clinical tools based on these markers.
Several authors previously reported associations between the serum markers in this investigation and the presence of fragility fractures and/or low BMD. Ting et al
studied premenopausal women with DRF compared with age-matched controls and found independent associations between DRF and both CTX-1 and OC, whereas differences in vitamin D and PTH were not observed. A similar investigation in postmenopausal women with DRF found propeptide of type I collagen to be independently associated with DRF, and also found no differences in vitamin D or PTH levels.
We similarly found CTX-1 and OC to be independent predictors at multiple time points, along with bone-specific alkaline phosphatase and SOST, and did not find vitamin D or PTH to contribute to any of the models. Although the studies performed thus far were limited in sample size, it may suggest that despite the current widespread availability of vitamin D and PTH laboratory studies, other markers may be better suited to predict osteoporosis or fracture.
Sclerostin inhibits the Wnt/β-catenin signaling pathway, acting as a negative regulator of osteoblasts and inhibitor of bone formation. Sclerostin was shown to increase with age and correlate positively with BMD.
In this investigation, we found increases in SOST levels with age; however, the levels of SOST diverged over time. The osteoporotic group increased at a slower rate than did the controls and significant differences in groups were evident at time points 3 and 4. Despite multiple studies demonstrating a positive correlation between SOST and BMD, Arasu et al
demonstrated an association between elevated SOST levels and hip fractures in elderly women despite demonstrating a positive correlation between SOST and BMD in the same subjects; this indicated a possible effect of SOST on fracture independent of BMD. The authors posited that in the setting of decreased bone mass, an increase in local mechanical strain may lead to the downregulation of SOST in an effort to favor bone formation over resorption, and that dysregulation of this feedback mechanism may predispose some individuals to fracture. The counterintuitive increased SOST levels in women with higher BMD may also be reflect the fact that SOST is produced by osteocytes themselves, and therefore may serve as an indirect quantification of osteocyte number and therefore bone mass.
Low BMD is a serious health issue that affects predominately women; however, it remains underdiagnosed and undertreated. Low rates of referral for BMD combined with its potentially limited sensitivity for the detection of early bone loss and degradation of bone microarchitecture highlight the need to explore different screening modalities for BMD loss.
Serum biomarker testing has the potential to serve as a screening tool that detects biochemical evidence of increased bone turnover at an age young enough to intervene meaningfully and prevent critical loss of bone mass. Serum testing also obviates the need for a separate referral for radiographic screening, which likely contributes to the low rate at which DEXA scans are currently obtained.
The contents of this publication are the sole responsibility of the author(s) and do not necessarily reflect the views, opinion, or policies of Uniformed Services University of the Health Sciences, the Department of Defense, or the Departments of the Army, Navy, or Air Force. Mention of trade names, commercial products, or organizations does not imply endorsement by the US Government. This study was funded by a Basic Science Grant from the American Foundation for Surgery of the Hand.
Appendix A
FIGURE E1Levels over time among subjects for each marker with and without an eventual diagnosis of osteoporosis. VIT D, vitamin D.
Data are shown as P values. Results from t test comparisons at each time point for each individual marker, comparing subjects with and without an eventual diagnosis of osteoporosis.
N-Terminal extension propeptide of type 1 collagen
.25
.87
.72
.60
BSAP
.08
.75
.08
.13
CTX-1
.95
.33
.30
.14
Vitamin D
.63
.64
.59
.64
∗ Data are shown as P values. Results from t test comparisons at each time point for each individual marker, comparing subjects with and without an eventual diagnosis of osteoporosis.
Wrist fracture as a predictor of future fractures in younger versus older postmenopausal women: results from the National Osteoporosis Risk Assessment (NORA).
00128-3&id=gr1.jpg){kind=link}
00128-3&id=gr2.jpg){kind=link}
00128-3&id=gr3.jpg){kind=link}
00128-3&id=fx1.jpg){kind=link}