Recent human studies suggested that serum osteocalcin is associated with the cross-talk between bone and energy metabolism. The aim of this study was to determine whether serum osteocalcin level is independently associated with the development of type 2 diabetes.
A retrospective cohort study was performed of 1,229 nondiabetic men, aged 25–60 years, who were recruited from the Health Promotion Center, Samsung Medical Center, between January 1997 and December 1997. They were followed regularly at the center on an out-patient basis and during hospitalization for a mean of 8.4 years, and the development of type 2 diabetes was determined.
In the baseline analysis, BMI, body fat percentage, triglyceride, homeostasis model assessment of insulin resistance value, and plasminogen activator inhibitor-1 levels varied inversely with the osteocalcin tertiles, and serum high-density lipoprotein cholesterol levels increased with the osteocalcin tertiles. However, no differences were observed in fasting glucose and glycated hemoglobin levels across the osteocalcin tertiles. Incident type 2 diabetes occurred in 90 (7.3%) of the study subjects. In Cox proportional hazards models, however, no statistical differences in the development of type 2 diabetes across the osteocalcin tertiles were evident after adjustment of other risk factors for incident diabetes.
Despite baseline associations with favorable metabolic parameters, the serum osteocalcin level was not associated with the development of type 2 diabetes in middle-aged males.
Recently, it has been suggested that the osteoblast-specific protein osteocalcin is central in the cross-talk between bone remodeling and energy metabolism (1–3). Osteocalcin signals to both pancreatic β-cells and adipose tissues have been demonstrated; they increase insulin secretion and β-cell proliferation and regulate body fat mass and adiponectin gene expression in a genetically modified mouse model (4). In addition, subsequent studies have shown that osteocalcin may be of value to prevent the development of type 2 diabetes in wild-type mice, which suggests a therapeutic potential for the treatment of metabolic diseases (5,6).
In agreement with these observations, accumulating human evidence has demonstrated that serum osteocalcin levels are inversely associated with blood glucose levels and positively associated with insulin secretion and insulin sensitivity (7–9). In addition, it has been suggested that serum osteocalcin levels are inversely associated with dysmetabolic phenotypes such as atherogenic dyslipidemia, abdominal obesity, metabolic syndrome, and increased brachial-ankle pulse wave velocity and carotid intima-media thickness (10–15). Moreover, in one study, serum osteocalcin levels were significantly lower in the angiographically proven coronary heart disease group than in the normal group and were decreased with the number of vessels involved (14).
However, almost all of the human studies were based on a cross-sectional analysis. Thus, it is still uncertain whether the aforementioned findings are merely correlations or if any osteocalcin actually directly affects energy metabolism in humans, as has been shown in animal- and cell-based studies. Therefore, we performed this study to determine whether serum osteocalcin levels are independently associated with the development of type 2 diabetes over a mean follow-up period of 8.4 years.
RESEARCH DESIGN AND METHODS
Study population
A detailed description of the study design has been published previously (16). In brief, the study subjects were recruited from among those who visited the Health Promotion Center of the Samsung Medical Center between January 1997 and December 1997. A total of 2,435 apparently healthy subjects (1,761 men and 674 women), aged 20–78 years, were enrolled in this study and were followed up regularly at the Health Promotion Center, on an out-patient basis and during hospitalization, for a mean of 8.7 years. Of the 2,435 subjects, we limited the subjects to 1,332 men, aged 25–60 years, who were expected to have minimal bone turnover, to minimize the effects of age, sex, and other factors on serum osteocalcin levels. Among them, 103 subjects had diabetes at baseline and were excluded from the analysis. Finally, 1,229 men who had normal glucose tolerance (n = 622) or impaired fasting glucose (IFG; n = 607) were enrolled for the analyses and followed up for a mean of 8.4 years. This study was approved by the institutional review board and the ethics committee of Samsung Medical Center (IRB no. 2009-10-035) and complied with the Declaration of Helsinki.
Baseline examination
At baseline, a complete physical examination was performed, and personal medical history and lifestyle factors, including cigarette smoking and alcohol consumption, were determined using a standardized questionnaire. Smoking was classified into three groups (current smoker, ex-smoker, and never smoker), and alcohol intake was defined as more than three drinks per day (>30 g ethanol/day). Blood pressure was measured using a mercury sphygmomanometer on the right arm with subjects in a sitting position after a 5-min rest. Weight and height were measured in the morning with subjects wearing light clothing but no shoes, and the BMI was calculated as the weight in kilograms divided by the square of the height in meters (kg/m2). Percentage of body fat was measured by bioelectrical impedance analysis (In Body 3.0; Biospace, Seoul, Korea). All blood samples were obtained during the morning after an overnight fast of 12–14 h. Plasma glucose was measured in duplicate by the hexokinase method using an autoanalyzer (Hitachi, Tokyo, Japan). The interassay coefficient of variation (CV) was 1.6%. Plasma insulin was measured in duplicate using an immunoradiometric assay method (Medgenix, Nivelles, Belgium). Intra- and interassay CV at serum insulin levels of <215 pmol/L were 2.2 and 5.8%, respectively, and at serum insulin levels of ≥215 pmol/L were 3.9 and 4.5%, respectively. Standard liver function tests, uric acid, γ-glutamyl transpeptidase (γGT), and lipid profiles, including serum total cholesterol, triglyceride, HDL cholesterol, and LDL cholesterol, were determined using an autoanalyzer (Hitachi), as were the white blood cell counts (Sysmex, Kobe, Japan). Fibrinogen level was measured by clotting method using a STA coagulation analyzer (Stago S.A., France). Plasminogen activator inhibitor-1 (PAI-1) level was measured by an enzyme-linked immunosorbent assay method using an Asserachrom PAI-1 kit reagent (Stago S.A., France) as specified by the manufacturer’s protocols. Lipoprotein (a) level was measured by the rate nephelometry method. Serum total osteocalcin level was analyzed by Metra TM Osteocalcin (Quidel, Santa Clara, CA) using a Dade Behring ELISA-Processor III (Behring Diagnostics, Somerville, NJ) with intra- and interassay CVs of 4.8–10.0 and 4.8–9.8%, respectively.
Definition of metabolic syndrome and diabetes
Metabolic syndrome was defined using the modified National Cholesterol Education Program Adult Treatment Panel III (NCEP-ATP III) criteria (17). However, we used a BMI cutoff value of 25.0 kg/m2 as the parameter for abdominal obesity instead of waist circumference. In this study, diabetes was defined by any one of the following: fasting plasma glucose (FPG) level ≥7.0 mmol/L, use of diabetes medication, or self-reported physician diagnosis. IFG was defined as glucose levels of 5.6–7.0 mmol/L. At baseline, the subjects were assessed for diabetes, as determined by a standardized questionnaire and the medical records, and followed regularly for the occurrence of new onset diabetes. To ascertain the incidence of type 2 diabetes, all data were collected from the Health Promotion Center on an out-patient basis and during hospitalization. Incident type 1 diabetes was excluded based on clinical parameters, including history of ketoacidosis at initial presentation, essentially requiring insulin therapy for blood glucose control, and low to undetectable serum insulin or C-peptide levels. The homeostasis model assessment (HOMA) was used to estimate the insulin sensitivity and β-cell function from FPG and fasting plasma insulin (FPI) concentrations. Insulin resistance (IR) was estimated using the HOMA-IR, defined as [FPI (μU/mL) × FPG (mmol/L)] ÷ 22.5. HOMA-%B was calculated using (20 × FPI) ÷ (FPG – 3.5) and was used to represent β-cell function.
Statistical methods
All data are presented as the mean ± SD or proportion, except for skewed variables, which are presented as the median (interquartile range, 25–75%). Because the distributions of serum levels of triglyceride, γGT, lipoprotein (a), PAI-1, and HOMA values were skewed, natural logarithmic transformation was applied in the statistical analysis. For clarity, nontransformed median values are presented in the tables and text. One-way ANOVA, followed by Tukey’s post hoc test, was used to compare the means between the tertiles of osteocalcin levels. Linear-by-linear association χ2 test was used for trend analysis between the tertiles. Pearson correlation coefficients were calculated to evaluate the associations between osteocalcin and clinical and laboratory measurements. Cox proportional hazards regression model was used to calculate an adjusted hazard ratio (HR) for incident type 2 diabetes among the subjects with osteocalcin tertiles. The upper osteocalcin tertile was used as a reference, and the results for the analyses are presented as HR with a 95% CI. Analysis was performed using PASW version 18.0 (SPSS, Chicago, IL). P values <0.05 were considered significant.
RESULTS
Baseline characteristics of the study subjects according to serum total osteocalcin tertiles are shown in Table 1. Serum total osteocalcin levels were associated with favorable metabolic profiles. BMI, body fat percentage, and triglyceride levels varied inversely with the osteocalcin tertiles (all P < 0.01), and serum HDL cholesterol levels increased with the osteocalcin tertiles (P = 0.014). In addition, inflammatory markers such as lipoprotein (a) (P = 0.049) and PAI-1 levels (P = 0.004) were different, and the prevalence of National Cholesterol Education Program–defined metabolic syndrome also showed a decreasing tendency across the osteocalcin tertiles (P < 0.001). However, although the levels of IR, represented by HOMA-IR value, tended to decrease across the osteocalcin tertiles (P = 0.025), no differences were observed in FPG and glycated hemoglobin (HbA1c) levels with the osteocalcin tertiles. In line with the results of the Table 1, the correlation analysis showed similar results (Supplementary Table 1).
Incident type 2 diabetes occurred in 90 (7.3%) of the 1,229 total study subjects and 79 (13.0%) of the 607 IFG subjects during the 8.4 years of mean follow-up. When compared with the upper osteocalcin tertile (reference group), no differences were observed in the middle or lower osteocalcin tertiles for the development of type 2 diabetes in Kaplan-Meier survival curve (P = 0.578 across the tertiles, Fig. 1). The results of Cox proportional hazard model analysis are shown in Table 2. Independent variables included in the models were age, BMI, body fat percentage, fasting blood glucose, HbA1c, systolic blood pressure, triglyceride, HDL cholesterol, fibrinogen, PAI-1, smoking status, presence of metabolic syndrome, and serum osteocalcin level by tertiles. FPG (HR per 1-SD 10.79 [95% CI 6.47–17.99]) and HbA1c levels (2.51 [1.82–3.45]) independently predicted future development of type 2 diabetes, and HDL cholesterol was associated with a lower incidence of diabetes (0.61 [0.47–0.79]). However, serum total osteocalcin was not associated with incident type 2 diabetes across the tertiles. Likewise, serum total osteocalcin was not associated with incident type 2 diabetes in the IFG group (Supplementary Fig. 1 and Supplementary Table 2).
CONCLUSIONS
In the current study, we tried to determine whether serum osteocalcin level is independently associated with incident diabetes using a retrospective cohort recruited from the Health Promotion Center. Of the 2,435 apparently healthy subjects (1,761 men and 674 women), we limited the subjects to 1,229 nondiabetic men, aged 25–60 years, who were expected to have minimal bone turnover to minimize the effects of altered bone turnover on serum osteocalcin levels. Previous studies have demonstrated that serum osteocalcin levels showed dynamic changes with age and sex (18). In men, serum osteocalcin levels were highest in the 20–29 year age group, declined and stabilized, then increased again in the seventh decade, and serum osteocalcin levels showed more dynamic changes, and increased abruptly around the time of menopause in women. Therefore, although we could not totally exclude the effects of altered bone turnover on osteocalcin level by measuring bone mineral density or other bone turnover markers, such as bone-specific alkaline phosphatase or other markers of bone resorption, we did attempt to minimize the effects of other confounders by uniformly enrolling middle-aged men, who were expected to have minimal bone turnover. In addition to middle-aged male subjects, we investigated the effect of osteocalcin on incident type 2 diabetes in another study population including total men, total women, women who were likely to be premenopausal (age ≤50 years), and total men and women. In Cox proportional hazard regression analyses, serum osteocalcin was not independently associated with incident type 2 diabetes in all of the aforementioned study groups. Therefore, it appears that serum osteocalcin is not associated with incident type 2 diabetes, not only in middle-aged men but also in other groups (data not shown).
In agreement with previous reports (10–15), serum osteocalcin levels were inversely associated with the presence of obesity, atherogenic dyslipidemia (high triglyceride and low HDL cholesterol levels), inflammatory status (high PAI-1 level), and metabolic syndrome in the baseline cross-sectional analyses (Table 1 and Supplementary Table 1). However, although the levels of HOMA-IR, an indicator of IR, decreased progressively from lower to upper osteocalcin tertiles, no differences were observed in FPG and HbA1c levels across the osteocalcin tertiles. Instead, serum osteocalcin levels were significantly correlated with FPG (r = −0.202, P = 0.041) and HbA1c levels (r = −0.211, P = 0.033) in 103 patients with diabetes at baseline (data not shown). Therefore, it appears that serum osteocalcin is more closely associated with the parameters reflecting insulin resistance including obesity, atherogenic dyslipidemia, and inflammation rather than glucose tolerance itself before the onset of diabetes. Presently, the association between serum osteocalcin and glucose tolerance became evident just after the onset of diabetes. In the current study, serum osteocalcin was not associated with the development of incident diabetes during the mean 8.4-year follow-up both in unadjusted and adjusted models. In addition, when we limited the study subjects to the IFG group, no differences in the progression rate to diabetes were observed (Supplementary Table 2 and Supplementary Fig. 1). Therefore, the role of circulating osteocalcin on glucose metabolism might be limited in human subjects.
To date, almost all human studies have agreed with the results of previous animal studies; that is, serum osteocalcin levels are inversely associated with blood glucose levels and positively associated with insulin secretion and insulin sensitivity. In addition, serum osteocalcin predicts all-cause and cardiovascular disease–related mortality in a prospective cohort of community-dwelling elderly men even though the relationship is U-shaped with men at both ends of the distribution at increased risk (19). However, a few studies have been conducted to determine the association between serum osteocalcin level and glucose tolerance with longitudinal follow-up. Pittas et al. (12) reported that mean osteocalcin concentration during follow-up predicted change of FPG at year 3 with 198 study subjects, independently of baseline FPG, age, sex, physical activity, smoking, education, and change in BMI at year 3. Another study also showed that reduced serum total osteocalcin (odds ratio 0.90; 95% CI 0.81–0.99), but not undercarboxylated osteocalcin levels, was an independent risk factor for development of diabetes after adjustment for baseline FPG, age, BMI, and exercise in 126 subjects within a nested case-control cohort in Thailand during the 10-year follow-up (20). Collectively, contrary to our results, several cross-sectional and some longitudinal studies have consistently demonstrated that circulating osteocalcin plays a favorable role in glucose metabolism in humans. However, previous longitudinal follow-up studies were conducted with a relatively small sample size and short follow-up period. Thus, the association between serum osteocalcin level and glucose tolerance is not yet conclusive in human subjects.
The basis of the discrepancy between the present and previous studies is unclear. Several suggestions can be made. Contrary to animal models, it may be possible that osteocalcin has minimal effects on energy metabolism in humans, particularly for glucose homeostasis. There are data to support this suggestion. First, several studies have shown that antiresorptive agents such as bisphosphonates, raloxifene, strontium ranelate, and the RANK ligand antibody denosumab suppress bone turnover by inhibiting the osteoclasts, and thus decrease circulating total osteocalcin levels (21–25). In addition, it was reported that alendronate treatment significantly decreased serum undercarboxylated osteocalcin level, while serum carboxylated osteocalcin and the undercarboxylated-to-carboxylated osteocalcin ratio were not affected in postmenopausal women (26). These observations indicate that antiresorptive agents have the potential to increase the risk of diabetes in theory. However, no large clinical trials have reported an association between treatment with antiresorptive agents such as bisphosphonates and incident diabetes. Vestergaard (27) showed that the risk of developing type 2 diabetes was reduced with alendronate, etidronate, and raloxifene treatment in a nationwide Danish cohort. In particular, a dose-dependent risk reduction was observed for alendronate; therefore, antiresorptive drugs do not appear to be associated with an increased risk of diabetes. Rather, they might be protective involving the suppression of bone turnover. Second, it was reported that chronic exposure to high glucose inhibited cell growth of a human osteoblast–like cell line in a dose-dependent manner and decreased osteocalcin mRNA levels in mouse osteoblast (28,29). In agreement with this observation, several human studies showed that osteocalcin and other bone turnover markers are lower in patients with established diabetes (30–33). Thus, it remains unclear whether osteocalcin precedes hyperglycemia or vice versa. Third, although almost all human studies have reported the positive association between serum osteocalcin levels and better glycemic control, Aoki et al. (34) showed that serum osteocalcin concentration was increased in Japanese subjects with early-stage type 2 diabetes who had never taken medication. Especially, serum osteocalcin level was positively correlated with postload 2-h glucose level on oral glucose tolerance test. Fourth, protein tyrosine phosphatase OST-PTP, the protein encoded by Esp that regulates osteocalcin carboxylation, is a mouse gene that is not expressed in humans (35). The collective observations highlight the controversy over whether osteocalcin plays a direct role in regulating glucose homeostasis in humans, as is the case from the animal studies. It is possible that previous human studies showing the association between circulating osteocalcin and favorable metabolic profiles were merely correlations.
In addition, it has been reported that about 25% of subjects with prediabetes progress to type 2 diabetes in 5 years (36). In addition, results of several recent prospective studies support the suggestion that the rate of progression to type 2 diabetes may be even higher, averaging 10–12% per year (37–39). However, only 13% (79 of 607) of our subjects with prediabetes progressed to diabetes during the mean 8-year mean follow-up. This may be partly because our study subjects presented for medical health check-ups. They may have been healthier than others in the general population. Consequently, although the rate of incident diabetes showed a tendency to decrease in subjects with upper serum osteocalcin tertile, it could not reach statistical significance.
Apart from its being a retrospective follow-up study in nature, this study had several limitations. First, according to the initial observations of Lee et al. (4), only the uncarboxylated form of osteocalcin has the ability to induce insulin secretion and the expression of genes encoding adiponectin and insulin. As a result, glucose metabolism is improved. In contrast, carboxylated osteocalcin was found to have none of these effects. However, we did not differentiate serum osteocalcin with respect to the γ-carboxylation status and only measured the total form of osteocalcin, instead of directly measuring carboxylated and uncarboxylated osteocalcin. Therefore, we do not know the differential mechanism of both types of osteocalcin to regulate insulin secretion and insulin sensitivity. Second, although we limited the subjects to middle-aged men to minimize the effects of age, sex, and altered bone turnover on serum osteocalcin levels, we could not entirely exclude the effects of age, sex, and other variables in the associations between plasma osteocalcin levels and glucose metabolism. Therefore, the additional measurement of bone resorption markers may further clarify the potential association between osteocalcin and glucose homeostasis. Third, data regarding concurrent medication, which may affect bone turnover and serum osteocalcin levels, including calcium supplements, vitamin D supplements, and osteoporosis medications, were not determined. Fourth, vitamin D deficiency increases bone remodeling and may increase serum osteocalcin levels. However, vitamin D status measured by serum 25-hydroxyvitamin D was not assessed in this study. Finally, we did not use the oral glucose tolerance test or HbA1c for the diagnosis of diabetes. Thus, the incidence of diabetes might be underestimated. However, the development of diabetes was regularly followed not only at the Health Promotion Center but also on an out-patient basis and during hospitalization by medical record review. Thus, we could minimize the number of undiagnosed subjects with diabetes.
Despite these limitations, this is the first study to investigate whether serum osteocalcin is associated with the development of incident diabetes in a relatively large sample size and long-term follow-up period with valid statistical methods. In conclusion, serum osteocalcin level is not associated with the development of type 2 diabetes in middle-aged male Koreans. However, further prospective studies to examine the role of undercarboxylated osteocalcin on the development of diabetes are warranted.
Acknowledgments
This study was supported by Samsung Medical Center Clinical Research Development Program Grant CRS-110-22-1.
No potential conflicts of interest relevant to this article were reported.
Y.-C.H. analyzed and interpreted the data, contributed to the discussion, and wrote the manuscript. J.-H.J., I.-K.J., K.-J.A., and H.Y.C. contributed to discussion and reviewed the manuscript. M.-K.L. designed the study, analyzed and interpreted the data, contributed to discussion, and reviewed and edited the manuscript. M.-K.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
The authors thank Hong-Yup Ahn from the Department of Statistics, Dongguk University-Seoul (Seoul, Korea), for helping with the statistical analysis.