Association Between Bone Mineral Density, Bone Turnover Markers, and Serum Cholesterol Levels in Type 2 Diabetes

Purpose: The association between bone mineral density (BMD), bone turnover markers, and serum cholesterol in healthy population has already been proved. However, in patients with type 2 diabetes mellitus (T2D), it has not been adequately analyzed. In this study, we investigated the correlation between BMD, bone turnover markers, and serum cholesterol levels in people with T2D. Methods: We enrolled 1,040 men and 735 women with T2D from Zhongshan Hospital between October 2009 and January 2013. Their general condition, history of diseases and medication, serum markers, and BMD data were collected. We used logistic regression analysis to identify the association between serum cholesterol levels and BMD as well as bone turnover markers. Results: In multivariate regression analysis, we observed that in men with T2D, high high-density lipoprotein-cholesterol and total cholesterol levels were significantly associated with low total lumbar, femur neck, and total hip BMD, while low-density lipoprotein-cholesterol level was only inversely associated with total lumbar and femur neck BMD. Total cholesterol and low-density lipoprotein-cholesterol levels were also negatively associated with osteocalcin, procollagen type I N-terminal propeptide, and β-crosslaps. In women with T2D, high-density lipoprotein-cholesterol level was observed to be negatively correlated with total lumbar, femur neck, and total hip BMD, while total cholesterol and low-density lipoprotein-cholesterol levels were only associated with BMD at the total lumbar. Furthermore, total cholesterol was also negatively associated with osteocalcin, procollagen type I N-terminal propeptide, and β-crosslaps; high-density lipoprotein-cholesterol was only related to osteocalcin and parathyroid hormone, while low-density lipoprotein-cholesterol was only related to β-crosslaps in women. Conclusion: Our study suggests a significantly negative correlation between serum cholesterol levels and BMD in both men and women with T2D. The associations between serum cholesterol levels and bone turnover markers were also observed in T2D patients.


INTRODUCTION
The prevalence of osteoporosis and low-energy fractures in the aging population is increasing, which may lead to disability, poor living quality, and even death (1). In patients with type 2 diabetes mellitus (T2D), the fracture risk is usually higher than that in the general population (2)(3)(4) and varies even after adjusting for age, duration of diabetes, antidiabetic drug usage, and body mass index (BMI) (5,6). The risk factors for osteoporosis in the general population have been reported and analyzed in previous studies (7). However, because of metabolic disorders, the risk factors for osteoporosis in people with T2D may be different, and it is important to discuss them. According to current studies, several factors are regarded as risk factors for osteoporosis in people with diabetes. Afshinnia et al. (8) reported that in patients with diabetes, old age, low body weight, low serum calcium, and low-density lipoprotein cholesterol (LDL-C) levels were independently associated with lumbar spine osteoporosis (8). Chen et al. (9) observed that abnormal blood lipid, abnormal adipokine levels, and elevated inflammatory factor levels were independent risk factors for osteoporosis in patients with T2D (9).
The associations between serum cholesterol and bone turnover markers were also widely studied. Most studies found that TC, HDL-C, and LDL-C levels were negatively correlated with osteocalcin (OCN) in the general population (28,29), while in patients with T2D, no association was reported (30,31). Both procollagen type I N-terminal propeptide (PINP) and β-crosslaps (β-cTX) showed no relationship with serum cholesterol in healthy postmenopausal women as well as in T2D patients (12,30). According to Ponda et al. in vitamin D repletion group, LDL-C was inversely correlated with serum parathyroid hormone (PTH) (32). However, another study found no association between PTH and serum cholesterol (33).
From the above, the association between BMD, bone turnover markers, and serum cholesterol remain controversial, and little research in T2D patients has been conducted. Therefore, we investigated the Chinese population with T2D, and aimed to clarify the association between serum cholesterol levels, including TC, HDL-C, and LDL-C, and BMD, at total lumbar, femur neck, and total hip. We analyzed men and women with T2D separately, as well as the linear and non-linear correlation.

Study Population
We conducted a hospital-based cross-sectional study. All participants were selected consecutively from the endocrinology department of Zhongshan Hospital between October 2009 and January 2013. All selected participants were >18 years old with T2D. T2D was diagnosed based on the Standards of Medical Care in Diabetes by the American Diabetes Association as follows: (a) hemoglobin A1c (HbA1c) ≥6.5%; or (b) fasting blood glucose (FBG) ≥7.0 mmol/L (no caloric intake for 8 h at least); or (c) 2-h blood glucose ≥11.1 mmol/L by oral glucose tolerance test (which uses glucose load containing the equivalent of 75 g anhydrous glucose dissolved in water); or (d) random blood glucose ≥11.1 mmol/L in patients with typical hyperglycemia symptoms or hyperglycemia crisis, which occurs in the absence of unequivocal hyperglycemia, results should be confirmed by repeat testing) (34). The exclusion criteria included (a) diagnosis of malignant tumor and severe heart, liver, or kidney diseases; (b) diagnosis of pituitary, thyroid, parathyroid, adrenal, and gonadal diseases; (c) long-term bedridden patients; (d) long history of using calcium, vitamin D, or other drugs that affect bone metabolism; and (e) patients lacking available information. Our final study sample comprised 1,776 diabetic patients (1,040 men and 735 women). The personal history and past medical history of each participant was obtained through individual questionnaires. This study was approved by the ethics committee of Zhongshan Hospital, Fudan University. All patients included in this retrospective observational study signed informed consent.

Clinical Assessment and Health History
Trained doctors conducted overall physical examinations for each participant. Body weight and height were measured with light clothing and without shoes, and the minimum unit of measurement was 0.5 kg and 0.01 m, respectively. We calculated BMI by dividing weight (kg) by the square of height (m 2 ). Blood pressure (systolic and diastolic) was measured in the sitting position and after at least 30 min of rest. We measured the blood pressure of each patient twice and obtained an average value to reduce errors. All other background information was voluntarily provided by the patients themselves or by their family members. Duration of diabetes was calculated in years, from the time the patient was diagnosed with T2D according to the patient's medical record, to the time we conducted blood tests and BMD measurements. The treatment of diabetes was classified into four categories: no treatment, insulin, oral medicine, and insulin and oral medicine. Smoking and drinking history were defined as never or ever. We also obtained the participants history of other diseases, family history, as well as history of trauma and operation.

Biochemical Parameters
Serum samples were collected at 6 a.m. after overnight fasting (8 h at least) and stored at room temperature for no more than 4 h before assay. Standard laboratory techniques were used to test glucose metabolism indices, including FBG and HbA1c; serum lipid metabolism indices, including TC, HDL-C, and LDL-C levels; and bone metabolism markers, including PTH, OCN, PINP, β-cTX, and 25-hydroxy-vitamin D [25(OH)D]. In addition, other laboratory markers such as hypersensitive Creactive protein (hsCRP), blood creatinine (Cr), blood urea nitrogen (BUN), calcium (Ca), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and alkaline phosphatase (ALP) levels were recorded.

BMD Measurement
We used dual-energy X-ray absorptiometry (Hologic-Discovery, USA) to detect the BMD of each patient at three sites: total lumbar, femur neck, and total hip.

Statistical Analysis
We separately performed analyses in men and women. We used mean ± standard deviation (SD) to describe continuous variables and used the number and proportion to describe categorical variables. We used chi-square tests for categorical variables, one-way ANOVA for normally distributed continuous variables, and the Kruskal-Wallis test for skewed continuous variables.
We calculated the regression coefficient and corresponding 95% confidence intervals (CI) using unadjusted and multivariate adjusted logistic regression analyses for the associations between per SD in serum cholesterol level and BMD. The crude model was adjusted for no variables. The multivariate-adjusted model 1 was adjusted for age; diabetes duration (y); treatment of DM; smoking; drinking; and BMI. The multivariate-adjusted model 2 was further adjusted for age; diabetes duration (y); treatment of DM; smoking; drinking; BMI; cerebrovascular disease; kidney disease; family history of DM; diastolic blood pressure; FBG, mmol/L; Cr, µmol/L; BUN, mmol/L; Ca, mmol/L; ALT, U/L; AST, U/L; and ALP, U/L. The multivariate-adjusted model 3 was adjusted for age; diabetes duration (y); treatment of DM; smoking; drinking; BMI; cerebrovascular disease; kidney disease; family history of DM; diastolic blood pressure; FBG, mmol/L; Cr, µmol/L; BUN, mmol/L; Ca, mmol/L; ALT, U/L; AST, U/L; ALP, U/L; HbA1c, %; and hsCRP, mg/L. A two-sided P-value of <0.05 was considered to be statistically significant.
To examine the non-linear association between serum cholesterol level and osteoporotic fracture (logOR), we further applied a two-piecewise linear regression model using a smoothing function after adjusting for age; diabetes duration (y); treatment of DM; smoking; drinking; BMI; cerebrovascular disease; kidney disease; family history of DM; diastolic blood pressure; FBG, mmol/L; Cr, µmol/L; BUN, mmol/L; Ca, mmol/L; ALT, U/L; AST, U/L; ALP, U/L; PTH, pg/mL; OCN, ng/mL; PINP, ng/mL; β-cTX, pg/mL; and 25(OH)D, nmol/L. In addition, we conducted a log likelihood ratio test comparing the one-line linear regression model with the two-piecewise linear model.
Association Between Serum Cholesterol Levels and Total hip BMD Table 4 and Figure 3 present the association between serum cholesterol levels and total hip BMD by the logistic regression model and two-piecewise linear regression model. In the crude model with a 1-SD increase in HDL-C and LDL-C levels, total hip BMD decreases of 0.033 g/cm 2 (P < 0.0001, 95% CI = 0.023-0.042) in men and 0.030 g/cm 2 (P < 0.0001, 95% CI = 0.04-0.019) in women were observed, while a 1-SD increase in LDL-C level was only found to be associated with 0.011-g/cm 2 total hip BMD decrease in men (P = 0.0271, 95% CI = 0.001-0.020).
For comparison, a longitudinal cohort study reported that in the late perimenopausal women, BMD declined substantially with an average loss of 0.018 and 0.010 g/cm 2 /yr from the spine and hip, respectively, and in the postmenopausal women, rates of loss from the spine and hip were 0.022 and 0.013 g/cm 2 /yr, respectively (P < 0.001 for all) (35). Therefore, there are clinical implications to the effects of serum cholesterol levels on BMD in our study.

As shown in
However, in women with T2D, only TC level was found to be related with PINP. When TC was >1.21 mmol/L, 1-SD increase in TC was correlated with 6.60 ng/ml decrease in PINP (P = 0.0241, 95% CI = 0.88-12.31).

PTH
In Table 8, in multivariate regression analysis, we found positive association between TC, HDL-C level and PTH in men with T2D. When TC was <0.45 mmol/L, a 1-SD increase in TC was associated with 1.80 pg/mL (P = 0.0385, 95% CI = 0.10-3.50) decrease in PTH, and when HDL-C was >0.77 mmol/L, per mmol/L increase in HDL-C was associated with 3.71 pg/mL (P = 0.0348, 95% CI = 0.27-7.15) decrease in PTH.

DISCUSSION
To the best of our knowledge, our study was the first to report an inverse correlation between serum cholesterol levels and BMD in an Asian population with T2D. After multivariate-adjusted analyses, we observed a significantly negative association between HDL-C level and BMD at the total lumbar, femur neck, and total hip in both male and female patients with T2D. Moreover, in men with T2D, TC level was significantly negatively correlated with total lumbar BMD, femur neck BMD, and total hip BMD, while LDL-C level was only inversely related to the total lumbar BMD and the femur neck BMD. In women with T2D, TC, and LDL-C levels were found to be negatively related to the total lumbar BMD only.
Although controversial, the association between serum cholesterol and BMD has been widely studied in the general population; however, no such studies have been conducted on people with T2D. Only one study conducted in the American population with T2D reported that low LDL level was independently associated with lumbar spine osteoporosis, after adjusting for sex; race; ethnicity; and use of statins, plasma glucose, and other lipoproteins. Several studies have reported that the correlation between serum cholesterol and BMD may be affected by race. Therefore, our analysis of the effect of serum cholesterol on BMD in Asian T2D patients is of great significance.
Our study observed a significantly negative association between HDL-C level and BMD in patients with T2D. Makovey et al. (11) reported that BMD values were significantly lower in postmenopausal women with higher HDL-C levels in the general population, while Kuipers et al. (16) and Choi et al. (36) reported that a higher HDL-C level was associated with lower BMD in men. Adami et al. (13) and Buizert et al. (15) reported a negative relationship between HDL-C level and BMD in both healthy men and women. However, in studies conducted in both men (36) and women (26), HDL-C level was reported to be positively correlated with BMD, and several studies have reported no association between HDL-C level and BMD in postmenopausal women (27). In our study, TC level was inversely associated with BMD at three sites in men with T2D, and in women, TC level was also inversely associated with total lumbar and femur neck BMD. According to Cui et al. (10), Makovey et al. (11), and Choi et al. (36), serum TC levels were inversely correlated with BMD in both pre-and postmenopausal women in the general population, and Garg et al. (37) observed that TC was weakly negatively associated with BMD in the Indian population in both sexes. In addition, TC was considered an independent risk factor for osteoporosis in some studies (38), which also suggested negative correlation with BMD. However, there are some controversies. Adami et al. (13) revealed that the relationship between TC and BMD was positive in healthy men and women. In addition, some studies reported no association between TC and BMD in postmenopausal women (26,27) and the general population in both sexes (15,23,24).
Furthermore, our study demonstrated that LDL-C level was inversely correlated with BMD at total lumbar and femur neck in men with T2D, and in women, it was negatively associated with total lumber BMD. Similar inverse correlation between LDL-C level and BMD was reported in postmenopausal women (10)(11)(12), as well as in the general Asian population (36,37) without T2D. However, in postmenopausal women, some studies observed no relationship between LDL-C level and BMD (26,27). Moreover, Kuipers et al. (16) and Hernandez et al. (21) reported that LDL-C  level was positively associated with BMD in African and Spanish men. Adami et al. (13) reported a positive relationship between LDL-C level and BMD in both healthy men and women. An association was also found between serum cholesterol and bone turnover markers in T2D patients in our study. In men with T2D, when above the inflection point, TC and LDL-C were negatively associated with OCN, PINP, and β-CTX, and HDL level was positively associated with OCN and PTH. In women with T2D, hen above the inflection point, TC was negatively associated with OCN and β-CTX; LDL-C was inversely related to β-CTX only, while HDL-C was positively related to OCN and negatively related to PTH.
Similar findings of serum cholesterol and bone turnover markers were also reported in previous studies. Zhou et al. (28) and Chen et al. (29) found negative relationships between TC, HDL-C, LDL-C levels and OCN (28,29). Ponda et al. reported that LDL-C was inversely correlated with serum PTH in a vitamin D repletion group (32). However, other studies found no association between serum cholesterol and bone turnover markers (12,30,31,33).
The potential mechanisms through which serum cholesterol may affect bone metabolism is illustrated in Figure 4. Adipocytes and osteoblasts were reported to share a common progenitor, mesenchymal stem cells (MSCs) (39,40), and the expansion of adipose tissue in the marrow was associated with bone loss (41)(42)(43). Kha et al. (40) reported that the osteogenic differentiation of MSCs could be stimulated by specific oxysterols. Therefore, a high HDL-C level, which is able to remove oxysterols from peripheral tissues, demonstrates negative effects on osteogenic differentiation (40). Not only HDL-C level but also LDL oxidation products were reported to inhibit the differentiation of osteoblasts and direct progenitor MSCs to undergo adipogenic differentiation instead of osteogenic differentiation; thus, this reduced bone formation (44). Moreover, oxidized LDL was reported to induce receptor-activated NFκB ligand (RANKL)dependent osteoclastic differentiation of mouse marrow preosteoclasts (45). The formation and survival of osteoclast and osteoclast activity were highly dependent on cholesterol (46,47). LDL receptor-related protein 5 (LRP5) exerts an anabolic action on bone through the Wnt-signal pathway (48), and LRP5-deficient mice were observed to have both high cholesterol levels and low bone mass (49). In humans, LDL receptor-related protein 6 (LRP6) mutation was related to high serum LDL-C levels complicated by severe osteoporosis (50).
The effect of serum cholesterol levels on BMD may differ between cortical and trabecular bones. Experiments in   mice have indicated that hyperlipidemia more prominently blunted the anabolic action of PTH in the cortical bone than in the trabecular bone (51,52); however, Kuipers et al. (16) observed a cross-sectional association between HDL-C level and trabecular BMD but not between HDL-C level and cortical BMD or integral BMD at the hip or whole body measured by dual-energy X-ray absorptiometry (16). According to our study, the associations between serum cholesterol and bone metabolism differed between men and women groups. We considered that these differences in results might be due to the following reasons. Firstly, the basic characteristics of the men and women groups differ. According to Table 1, men were significantly younger than women were and the diabetes durations were shorter. The mean HbA1c was higher in men, whereas, the mean β-cTX was higher in women. The proportion of smoking or drinking history was significantly higher in men than in women; and the treatment of T2D was different. These differences in characteristics between men and women groups might have led to the different BMD losses associated with increases in lipids. Secondly, in the women group, we did not consider the menopausal status or estrogen level, which might also have some effects on our study results.
Our study has several strengths. Firstly, this is the first study to report an inverse association between serum cholesterol levels and BMD in an Asian population with T2D. Secondly, our sample size was large. Thirdly, we collected complete data and adjusted for various possible confounding factors. However, our study also has several limitations. Firstly, the causal relationship between serum cholesterol levels and BMD is difficult to assess in this cross-sectional study, and our retrospective study relied on previous data. Some key statistics could not be further measured, which may affect the selection of controls. Secondly, we only collected the serum samples once from all participants, and BMD at each anatomical site was detected once; this may cause deviations in cholesterol levels and BMD values. In addition, our study used dual-energy X-ray absorptiometry to measure BMD, which may cause measurement errors and cannot distinguish between BMD of the trabecular bone and of cortical bone. Thirdly, we might have omitted some confounding variables, which might also have had some effects on our study results, such as the menopausal status, estrogen level, dietary habits, physical activity, antilipemic medication, and previous fractures.

CONCLUSION
Our study suggests a significantly negative correlation between serum cholesterol levels and BMD in patients with T2D in multivariate regression analysis. The associations between serum cholesterol levels and bone turnover markers were also observed. However, further studies are required to confirm these findings.

AUTHOR CONTRIBUTIONS
XQ and MY: designed the research. YiY, JiL, CZ, and XQ: conducted the research. GX, XY, YZ, YaY, JiL, CZ, and MY: provided the essential reagents or the essential materials. YiY, GL, and XQ: analyzed the data or performed the statistical analyses. YiY and GL: wrote the manuscript. GL, JuL, CM, XQ and MY: critical revised the manuscript.