Introduction

Osteoporosis is a preventable and treatable disease which may not become clinically evident until a fracture occurs [1]. Many factors other than low bone mineral density (BMD) have been suggested as predictors of risk of future fractures [2]. Risk factors for osteoporosis and osteoporotic fracture have been determined and used to identify the need for further evaluation or preventive and therapeutic regimens [1]. Vertebrae fracture is the most common type of osteoporotic fracture and occur earlier in the natural history of the disease. The future risk of fracture, even ending with mortality, is greatly increased among patients with prevalence vertebrae fracture [3].

Osteoporosis is described as being associated epidemiologically with atherosclerosis and hyperlipidemia [4]. There is a growing evidence that bone and fat metabolisms are related to each other, but data are limited and contradictory. Several receptors reside on both osteocytes and endovascular lining cells, and cytokines appear to play important roles in both the systems [5]. Also in previous clinical studies, it is reported that the lipid profile is related to the bone mass [6, 7].

This study aims to investigate the effect of the serum lipid levels on BMD and vertebral fractures.

Materials and methods

From 138 records examined, 107 were included in the study. These women, with an age range of 45–79 years, had been postmenopausal for more than 12 months and had not used any medications which could affect the lipid profile or the bone metabolism for at least 6 months before the clinical evaluation (hormone replacement therapy, lipid lowering agents or antiresorptives). Patients who had diabetes mellitus, thyroid disorder, parathyroid disorder, metabolic bone diseases, rheumatoid arthritis, ankylosing spondylitis, or chronic obstructive pulmonary disease were not included. All the patients were questioned about smoking habits, sun exposure, physical activity (ladder use, number of going out of home, regular exercise 2–3 days/week, carrying weight), fracture history, family fracture history, number of pregnancy, daily calcium intake (milk and milk products), regular read meat, saturated lipid, alcohol and coffee consumption.

Body mass index (BMI) was calculated as weight divided by height squared (kg/m2) [8]. Blood samples were taken after an overnight fast. In addition to routine tests, total cholesterol (TC), triglyceride (TG), high density lipoprotein cholesterol (HDL-C) levels were measured by auto-analyzer. Low density lipoprotein cholesterol (LDL-C) levels were calculated using the Friedewald equation (LDL-C = TC − HDL-C − TG/5). VLDL cholesterol was equal to TG/5. Parathyroid hormone (PTH) was measured by radioimmunoassay.

Lateral dorsal and lateral lumbar spine radiographies, which were taken to assess existing vertebral fractures, were evaluated by a radiologist and the 89 radiographies with good technical properties were chosen, then scored by the Kleerekoper method [9]. Anterior, posterior, mid-vertebral heights of the vertebrae corpora were measured and a vertebrae deformity score (VDS) was calculated for each vertebra. All vertebrae from thoracic 4 to lumbar 5 were scored on inspection with the naked eye and compared with the vertebrae below and above, then scored as follows: 0 (normal shape and dimensions); 1 (only end plate deformity: middle height < 85%); 2 (anterior wedge deformity: anterior height < 85%); 3 (compression deformity: all three heights < 85%). We also accepted decreases of >4 mm at middle height, anterior height and all heights as criteria to determine VDS 1, 2, 3, respectively. Patients with VDS = 0 for all the vertebrae evaluated were included in the subgroup without vertebrae fractures and with VDS = 1, VDS = 2, VDS = 3 were included in the subgroup with vertebrae fractures.

BMD values were measured by dual-X ray absorptiometry (DXA) (Hologic QDR 2000 Inc., Woltham, MA, USA) at lumbar spine (L1–4 vertebrae), right hip (femoral neck and trochanter) and radius [1/3 distal, mid-distal (MD), ultra distal (UD)]. BMD values were given as g/cm2 and the results were reported as T scores (standard deviation above or below values for young healthy population). World Health Organization definitions were used for the T score assessment. We accepted T scores more than −1.0 as normal, between −1.0 and −2.5 as osteopenia, and less than −2.5 as osteoporosis [10]. The women who had a T score ≤−2.5 at lumbar and/or femoral region were included in the patient group with osteoporosis and women who had a T score ≥−2.5 at lumbar and/or femoral region were included in the group without osteoporosis.

We used χ 2 test for the comparison of the categorical data. The independent sample t test, Mann–Whitney U were used for the comparison of the data that distributing normally and non-normally, respectively. Since some factors such as age, menopausal duration, menarche age and BMI are related to lipid profile and bone mass, we adjusted the results by ANCOVA. Pearson correlation analysis was used for the associations between the lipid profile and the bone mass. We reassessed the relationship between them by partial correlation analysis after adjustment for the possible confounders. We performed multivariate binary logistic regression analysis to determine the risk factors for osteoporosis and the vertebrae fracture. SPSS 11.5 for windows was used for statistical analysis and P < 0.05 was considered as significant.

Results

The demographic characteristics of the study population are shown in Table 1. The mean age of the entire study group was 60.07 ± 6.31 years (range 45–79 years). The patients recruited for the study had middle level of income. The diet profile was largely consisting of carbohydrate and the use of saturated lipid was prevalent. The red meat was consumed one or three times in a month nearly by the half of them. The milk products were more preferred instead of milk. The alcohol and the caffeine consuming were not usual. Most of them were not smoking (Table 1).

Table 1 Demographic characteristics in the study population
Full size table

After the evaluation of the radiographies, we determined vertebrae fractures in 63 women and no fracture in 26 women. When we compared age, menarche age, menopause age, menopause duration, BMI between the patients with and without vertebral fractures, there were no significant differences in age, menopause age, menopause duration, BMI between the two groups. However the mean menarche age of the group with vertebrae fractures (14.4 ± 1.7 years) was significantly older than the group without vertebrae fractures (13.5 ± 1.3 years). When we evaluated two groups for the diet profile, regular habits (smoking, alcohol, caffeine), physical activity, fracture history, family fracture history, number of pregnancy, we found no differences between the two groups.

When we compared the lipid profiles of the patients with and without fractures, we found that the mean serum TC, TG and LDL-C levels of the patients with vertebrae fractures (214 ± 31.3, 139.4 ± 60.6, 132.7 ± 31.2 mg/dl, respectively) were significantly lower than the patients without fractures (236.8 ± 39.3, 170.3 ± 57.1, 151.6 ± 39.6 mg/dl, respectively). There was no difference in HDL-C and VLDL-C. After the adjustment for age, menopause duration, menarche age, BMI the differences remained significant in TC, TG and LDL-C levels. The results and P values are summarized in Table 2.

Table 2 The comparison of the serum lipid values in the patients with vertebrae fractures and without vertebrae fractures
Full size table

We used multivariate binary logistic regression test to determine the possible risk factors for the vertebrae fracture existence. Total cholesterol level was the most prominent factor affecting the vertebral fracture existence. An increase of 1 mg/dl total cholesterol decreased the risk of vertebrae fracture by 2.2% (P = 0.009) and the delay of the menarche for 1 year increased the risk of vertebrae fracture by 39% (P = 0.052).

Next, we divided 107 postmenopausal women into two groups according to the T scores obtained by the BMD measurements to analyze the relationship between the BMD and the lipid profile. We determined osteoporosis in 36 (33.6%) women at the lumbar region and/or the hip, while no osteoporosis was found in 71 women (66.3%). The age range was 52–79 (62.8 ± 6.3 years) for the group with osteoporosis and 45–71 (58.6 ± 5.8 years) for the group without osteoporosis. When the two groups were compared statistically in terms of age, BMI, menopause duration, menarche age, the mean age of the osteoporosis group was significantly older than the other group (P < 0.01). The mean menopause duration of the osteoporosis group was significantly longer and the mean BMI value of the osteoporosis group was significantly lower than the values of the non-osteoporosis group (P < 0.001, P < 0.01, respectively). There was no difference between the two groups in the menarche age, the diet properties, the lifestyle properties, the sun exposure, the fracture history, the family fracture history.

The comparison of the laboratory values of the osteoporosis group and the non-osteoporosis group was insignificant (P > 0.05). The mean values of the serum TC, TG, HDL-C, LDL-C, VLDL-C in the patient group with osteoporosis were as follows: TC, 229 ± 46 mg/dl; TG, 155.6 ± 132.9 mg/dl; HDL-C, 55.1 ± 16.7 mg/dl; LDL-C, 141.7 ± 40.5 mg/dl; VLDL-C, 32.1 ± 27.9 mg/dl. The results of the group without osteoporosis were as follows: TC, 220.7 ± 33.7 mg/dl; TG, 158.7 ± 70.6 mg/dl; HDL-C, 51 ± 12.9 mg/dl; LDL-C, 136.7 ± 34.1 mg/dl; VLDL-C, 31.7 ± 14.2 mg/dl. The difference between the lipid profiles of the two groups was insignificant (P > 0.05). After the adjustment of the values of the lipid profile for the possible confounders (age, BMI, menopause duration), we also found no differences between the two groups. The comparison of age, menopause duration, BMI, and laboratory parameters between the two groups are summarized in Table 3.

Table 3 The comparison of the mean age, menopause duration, BMI, menarche age and laboratory parameters in the patients with osteoporosis and without osteoporosis
Full size table

We performed correlation analyses (Pearson correlation test) between the BMD values and the serum lipids (TC, TG, HDL-C, LDL-C, VLDL-C) for the entire group. There was no correlation between the lumbar, femoral neck, femoral total, femoral Ward’s, radius 1/3 distal, radius MID, radius UD BMD measurements and TC, TG, HDL-C, LDL-C, VLDL-C through the entire group (P > 0.05). After the adjustment for age, menopause duration and BMI, only in the forearm UD region there were weak associations for TC and LDL-C (r = 0.188, P = 0.05; r = 0.206, P = 0.038, respectively). We did not find any associations for the remaining BMD values and the lipid profile parameters even after the adjustment. When a multivariate binary logistic regression analysis was performed with the presence of osteoporosis as dependent variable and each lipid level as independent variables, none of the lipid parameters were found to be associated with the existence of osteoporosis.

Discussion

The relationship between lipids and bone metabolism has become a topic of interest in recent years and there is increasing evidence showing a connection between these two systems. This study shows that serum lipids have impact on the vertebrae fracture existence, rather than the BMD alterations. The TC, TG and LDL-C levels were lower in the postmenopausal women who had at least one vertebrae fracture. Total cholesterol level was the strongest factor affecting the vertebral fracture existence. And an increase of 1 mg/dl TC decreased the risk of vertebrae fracture by 2.2%. The existence of osteoporosis due to T score was not influencing the lipid values, even after the adjustment for the confounding factors. There were only weak associations between TC, LDL-C and the BMD at the forearm UD region in the adjusted model.

In this report we would like to point to the lower levels of TC, TG, and LDL-C in the patients with vertebrae fractures in comparison with the patients without vertebrae fractures. There are two issues that mention the relation of the vertebrae fracture existence and the serum lipids [7, 11]. In these studies that performed in postmenopausal women the TG levels of the vertebra fracture groups were found to be lower as our patients. Our study has some superiority to the study performed by Yamaguchi et al. and Bagger et al. First, we were able to evaluate the lifestyle and diet properties, and we showed the equality of them in the patients with and without vertebrae fractures. Secondly, the factors such as age and menopause duration were not influencing our results, which were not taken into account as confounding factors for the lipid values by them. And thirdly, the lumbar BMD of which the lower levels would be a cause of vertebrae fracture itself was not differing between our groups opposite to their results. Beside these factors, in our vertebrae fracture evaluation method which is a reliable one, a decrease of 15% height is enough to accept the existence of a vertebrae fracture. However, in these studies the amount of the decrease in the vertebrae corpus was needed to be higher (20% for mid-height and 25% for anterior height in the Yamaguchi’s; 20% in the Bagger’s). So our vertebrae fracture ratios are higher from these studies. Of course the geographic, racial and the dietary differences cannot be ruled out to match the whole results.

In our study, only the menarche age was differing between the groups with and without vertebrae fractures. The mean menarche age was older; accordingly the estrogen exposure time was shorter in the group with vertebrae fractures. But this data cannot be influencing our results, because the lower levels of TC, TG and LDL-C in the patients with vertebrae fractures remained significant after the adjustment for the menarche age.

Our results suggest that the lower levels of serum lipids are associated with vertebrae fracture presence. This may be related with the some previous knowledge about the relation of estrogen with either bone metabolism or lipid metabolism. The menopausal bone loss is a composite of loss caused by estrogen deprivation and age per se for the hip and the total body, but is caused by estrogen deprivation alone for the spine [12]. Estrogen is a steroid hormone which is synthesized from cholesterol [13]. The esterified forms of these water insoluble steroid hormones are stored in lipoproteins in the fat tissue and are also transported by the lipoproteins [14]. So the lower levels of LDL-C would be associated with the lower levels of stored estrogen. This could be an explanation to the fracture of the most estrogen responsive bone ‘vertebra’ in the patients with lower levels of LDL-C. Also in a community-based large study, estrone, the principle estrogen in the postmenopausal women, is reported to be strongly associated with serum TC and TG in a group of postmenopausal women with carotid atherosclerosis [15]. The lower levels of TC, TG and LDL-C of the women with vertebrae fractures in our study may be associated with the possible lower levels of estrone.

The mechanism may be directly related with the contribution of the lipids to the structure of the bone. Despite some experimental studies which report negative effects of hyperlipidemia on bone [4, 16], Parhami et al. indicate that a baseline level of cholesterol synthesis is necessary for the osteoblastic differentiation of marrow stromal cells [17]. Also, more recently, the oxysterols which are the products of cholesterol oxidation have been shown to have pro-osteogenic effects on pluripotent mesenchimal stem cells [18]. The contributions of the cholesterol to the structure of the vertebrae would be impaired by the lower levels of TC in our patients with vertebrae fractures. The underlying mechanism should be investigated experimentally.

The mean BMD of our patients with vertebrae fractures was not significantly different from the patients without vertebrae fractures. And none of the lipid parameters were found to be associated with BMD of the lumbar spine. The BMD and the bone quality are the determiners of the fracture risk [2]. The vertebrae fracture is the only tool that can be clinically used to asses the bone quality [7]. As the BMD of the vertebrae fracture group were not significantly different from the other group we can suggest that the effect of the serum lipids may be on the bone quality.

We did not find a strict relationship between the lipid profile and the BMD of our patients. Also our multiple regression analyses could not prove a significant association between the serum lipids and the BMD at any skeletal site. These results are similar to the large population-based observational study the NHAEMS III survey including 13,592 participants [5] and also a recent observational study including 1,176 subjects [11]. Although there are small cross-sectional clinical studies that showed associations between lipid and BMD, they were not able to evaluate the subjects for the common risk factors that would affect the both systems or the study population was not homogenous as ours [6, 7]. A follow up study which showed that the relation at the beginning of the menopause disappears after the 8-year period, stated that this was a result of estrogen deprivation influencing the both systems [12].

There was a possible limitation of the current study. The total body fat mass was not investigated in this study, although it has been described to be strongly associated with serum lipids [19]. We preferred to use BMI which is much more available.

In conclusion we found that the serum TC, TG, LDL-C levels of the patients who had vertebrae fractures were lower than the patients without vertebrae fractures. The TC level was the most important factor affecting the vertebral fracture existence. And an increase of 1 mg/dl total cholesterol decreased the risk of vertebrae fracture by 2.2%. As a result the serum lipids have impact on vertebrae fracture existence rather than major BMD alterations.