The authors state that they have no conflicts of interest.
Published online on December 29, 2008
Epidemiological observations support a positive relationship between cardiovascular diseases (CVD) and osteoporosis, where cholesterol has been indicated to be a possible link. Only a few studies have investigated the relation between lipids and BMD, but the association remains unclear. We studied the relationship between serum lipids and BMD of the calcaneus. A cross-sectional population-based study was performed, based on data from the Longitudinal Aging Study Amsterdam, including 620 men and 635 women, 65–88 yr of age. BMD was measured by quantitative ultrasound (QUS), velocity of sound (VOS; m/s), and broadband ultrasound attenuation (BUA; dB/MHz). Models were adjusted for age, body mass index, physical activity, smoking, alcohol, diabetes mellitus, hypertension, testosterone, and 25-hydroxyvitamin D. No association was found between total cholesterol (TC) and QUS. Men and women in the highest quartile of high-density lipoprotein cholesterol (HDL-c) had a significantly lower QUS (men—VOS: β = −20.8, p = 0.00; BUA: β = −5.2, p = 0.02; women—VOS: β = −18.6, p = 0.00) compared with men and women in the lowest quartile. An even stronger positive association was seen between TC/HDL-c ratio and QUS (men—VOS: β = 21.8, p = 0.00; BUA: β = 5.5, p = 0.01; women—VOS: β = 19.2, p = 0.00; BUA: β = 3.6, p = 0.05). Our analysis shows that the lipid profile that is favorable in the prevention of CVD (i.e., high levels of HDL-c and low TC/HDL-c ratio) is unfavorable for QUS. These results indicate that HDL-c levels do not explain the association between osteoporosis and CVD.
Cardiovascular disease (CVD) and osteoporosis are major causes of impaired quality of life, morbidity, and mortality among elderly men and women. Several epidemiological studies found a relationship between the two diseases,(1–4) which implies that there might be some common mechanism underlying the pathogenesis of both CVD and osteoporosis. However, the mechanism of this association has still not been fully elucidated. Different corresponding risk factors have been suggested, such as estrogen deficiency, smoking, hypertension, oxidative stress, inflammation, malnutrition, and hyperhomocystinemia,(5,6) but none of them could clarify the relationship between CVD and osteoporosis completely.
Serum lipid levels have a prominent role in the pathogenesis of CVD. Some(7–9) but not all(10–12) observational studies found that treatment of CVD with statins, lipid-lowering drugs, was associated with a significantly reduced fracture risk and increased BMD. Therefore, the question arises whether lipids are associated with the pathogenesis of osteoporosis as well and whether they could be the link between CVD and osteoporosis.
Only little is known about the relationship between lipids and BMD in white men and women. Despite the fact that, of all lipids, high-density lipoprotein cholesterol (HDL-c) is associated most strongly with CVD, strikingly most research has been devoted to the relation between total cholesterol (TC) and BMD. Only a few studies include the relation between HDL-c and BMD.(13–20) Their results are contradictory because both positive and negative relationships have been found. Recently, much progress has been made in the understanding of the mechanisms behind the protective effect of HDL-c on CVD. HDL-c could have a more important role in bone metabolism than has been assumed thus far.
The aim of this study was to investigate the relationship between lipid levels and quantitative ultrasound (QUS). Data were used from the Longitudinal Aging Study Amsterdam (LASA) to examine the relationship between different lipids (TC, HDL-c, and TC/HDL-c ratio) and BMD of the calcaneus (velocity of sound [VOS] and broadband ultrasound attenuation [BUA] measured by QUS).
MATERIALS AND METHODS
Data for this study were collected within the LASA. LASA is an ongoing cohort study of the predictors and consequences of changes in autonomy and well being in older persons in the Netherlands (predominantly white). At the baseline examination (in 1992–1993), a sample of 3805 older men and women (55–84 yr of age), stratified for age, sex, and expected 5-yr mortality, was drawn from the population registry of 11 municipalities in areas in the west, northeast, and south of The Netherlands. The sample represents the older Dutch population with respect to geographic region and degree of urbanization. The sampling and data collection procedures have been described elsewhere.(21)
Of the 3805 elderly persons who were initially approached, 3107 (62.3%) took part in the baseline examination (1992/1993). Nonresponse was related to age (p < 0.001), the oldest persons being less likely to participate. For this study, the first follow-up (1995/1996) was used as a baseline. Loss to follow-up between the first and second cycle was mainly because of death. This study was performed within a subgroup of the LASA sample, which consisted of participants who participated in both the main and the medical interview of the second data collection and who were born between 1908 and 1930. Of the 1720 eligible respondents, 1509 (87.8%) took part in the medical interview. Exclusion criteria were use of bisphosphonates (n = 18), estrogens (n = 10), and statins (n = 3). We excluded persons with missing values for TC (n = 201) and/or QUS (n = 223). As a result, our final study sample consisted of 1255 participants. Informed consent was obtained from all respondents. The study was approved by the medical Ethics Committee (VUMC) and conducted according to the principles of the Helsinki Declaration.
We examined the influence of lipids (HDL-c, TC/HDL-c ratio, and TC) versus the BMD of the calcaneus, measured with QUS (VOS [m/s] and BUA [dB/MHz]). To determine whether lipids can be the link between CVD and osteoporosis, we also studied the relationships between CVD (stroke, heart diseases, and peripheral arterial diseases) and QUS and between CVD and serum lipids. To explore the impact of low lipid values on QUS, we compared the QUS of the lowest quartile of TC/HDL-c ratio and TC with the QUS of the other quartiles together. Models were adjusted for age (quartiles), BMI (<25, 25–30, >30 kg/m2), physical activity (quartiles), smoking (no, former, yes), alcohol consumption (Garretsen index: does not drink, light, moderate, and excessive drinking), diabetes mellitus (self-reported and/or use of antidiabetic medicine), hypertension (diastolic > 100, systolic >160 and/ or use of antihypertensives), testosterone (quartiles), and 25-hydroxyvitamin D [25(OH)D] (quartiles).
Blood was collected in 1995 and 1996 after an overnight fast. Serum TC was measured with an automated analyzer. A few years after the original measurements, frozen serum was used to determine HDL-c levels. TC/HDL-c ratio was calculated from TC and HDL-c (TC/HDL-c).
QUS data were obtained using the CUBA Clinical instrument (McCue Ultrasonics, Winchester, UK). BUA (dB/MHz) and VOS (m/s) were measured twice in both the right and left calcaneus. Mean BUA and VOS values were calculated from these four measurements. The short-term precision of VOS and BUA was reported as CVs of 1.4% and 3.4%, respectively, whereas long-term precision of VOS and BUA was reported as CVs of 1.3% and 4.9%, respectively.
Chronic diseases were assessed with a detailed questionnaire on self-reported chronic diseases from which we used the information on stroke, heart diseases, and peripheral arterial diseases. To be classified as CVD, suffering from either one was taken as sufficient condition in the analysis.
The baseline information on age and sex was derived from the municipal registries. Body weight was measured without clothes and without shoes using a calibrated bathroom scale. Height was measured using a stadiometer. BMI was calculated as weight (kg) divided by the square of height (m2). Smoking status (never smoker, former smoker, or current smoker), alcohol use (number of drinks per week), physical activity (housekeeping activities, sports and leisure activities during the previous 2 wk), and the presence of chronic diseases like diabetes and hypertension were assessed in a face-to face interview. To obtain information on medication use, the respondent had to show all medication he/she used at the moment of the interview. The interviewer recorded the names, types, and doses of these medications. Blood pressure (mmHg) was measured after 5 min of rest at the upper left arm with subjects in a seated position, using an oscillometric blood pressure monitor (model HEM-706; Omron, Tokyo, Japan).
Serum testosterone concentrations were only measured in men by radioimmunoassay (Coat-A-Count; DPC, Los Angeles, CA, USA), with a CV of 11% at 2·6 nM.
25(OH)D was determined using a competitive protein binding assay (Nichols Diagnostics, San Juan Capistrano, CA, USA). The interassay CV was 10%.
Data analysis was conducted with SPSS version 14.0. While the association between lipids and QUS turned out to be a nonlinear one, quartiles were made of each lipid parameter. In linear regression models, the lowest quartile was used as reference value. Associations between lipids and CVD and between QUS and CVD were assessed by logistic regression models. All models were controlled for BMI, age, smoking, physical activity, alcohol consumption, hypertension, diabetes mellitus, vitamin D, and, in men, testosterone. The accepted level of statistical significance was p < 0.05. Men and women were analyzed separately because of the difference in bone strength in older men and women.(22)
Table 1 shows subjects' descriptive characteristics. The mean age of the participants in this study was 75 yr. Average serum TC levels in men and women were slightly elevated, whereas serum HDL-c levels and TC/HDL-c ratios were within the normal range. The prevalence of CVD was 43% in men and 30% in women.
Table Table 1.. Descriptive Characteristics of the Study Population
Relation between lipids and CVD
In accordance with the literature, this study showed that, of all lipids, HDL-c was most strongly associated with CVD. Men and women in the highest quartile of HDL-c had a significantly lower risk of CVD compared with men and women in the lowest quartile (men: OR = 0.27; 95% CI = 0.16–0.46; women: OR = 0.55; 95% CI = 0.32–0.95; Table 2). Between TC/HDL-c ratio and CVD, a positive association was observed in men (OR = 2.4; 95% CI = 1.4–3.9), but not in women. In this study, no significant association was found between TC levels and CVD.
Table Table 2.. Association Between HDL-c and CVD in Men and Women
Relation between CVD and QUS
After adjusting for confounding, a significant inverse association was found between QUS (VOS) and peripheral arterial diseases (men: OR = 0.30; 95% CI = 0.13–0.71; women: OR = 0.30; 95% CI = 0.12–0.73; Table 3). No significant association was found between CVD and QUS.
Table Table 3.. Association Between QUS (VOS) and PAD in Men and Women
Relation between lipids and QUS
Serum levels of HDL-c were significantly inversely associated with QUS, and the association was even stronger (positively) for the TC/HDL-c ratio: men in the highest quartile of HDL-c had a significantly lower QUS (VOS: β = −20.8; p = 0.00; BUA: β = −5.2; p = 0.02) compared with men in the lowest quartile of HDL-c. A comparable positive association was seen between TC/HDL-c ratio and QUS (VOS: β = 21.8; p = 0.00; BUA: β = 5.5; p = 0.01; Fig. 1). Among women, a similar negative association between HDL-c and VOS (β =−18.7; p = 0.00) and positive association between TC/HDL-c ratio and QUS (VOS: β = 19.2; p = 0.00; BUA: β = 3.6; p = 0.05) were found (Fig. 2). These associations remained significant after adjusting for age, BMI, smoking, physical activity, alcohol consumption, hypertension, and diabetes mellitus. Only in an unadjusted analysis was a significant positive association seen between TC and VOS in women.
To examine the influence of a low TC/HDL-c ratio on QUS, the QUS of participants of the lowest quartile of TC/HDL-c ratio were compared with the QUS of the combined upper three quartiles. After adjusting for confounding, men and women with a TC/HDL-c ratio in the lowest quartile had significantly lower VOS (men: β = 13.6; p = 0.00; women: β = 15.3; p = 0.00) compared with those from the combined upper three quartiles. Similar analyses were performed to evaluate the influence of low TC on QUS. Univariate models showed that men of the lowest quartile of TC had significantly lower VOS compared with the rest of the population (data not shown). After adjusting for confounding, this relationship did not remain significant.
Our analysis showed that the lipid profile that is favorable in the prevention of CVD is at the same time unfavorable for QUS. A high HDL-c level, often called “good” cholesterol, was found to be related to low QUS. Moreover, the relation was even stronger (inversely) when high serum HDL-c levels were combined with low TC levels (that is low TC/HDL-c ratio). The results remained significant after correcting for age, BMI, physical activity, smoking, alcohol use, diabetes mellitus, hypertension, testosterone, and vitamin D. Univariate models also showed that very low TC levels were detrimental to QUS (VOS). However, after adjusting for confounding, the trend was no longer significant. These findings indicate that HDL-c, the lipid most strongly related to CVD according to both literature and our study, is also most strongly related to QUS.
The influence of HDL-c on BMD and bone metabolism is still unclear. A few studies that examined the relation between HDL-c and BMD, measured by DXA, showed different results. A number of studies showed a negative relationship between HDL-c and BMD,(14–16,19,20,23) supporting our results. However, most of these studies were performed in small patient groups, and only few or no confounders were added to the models. Two larger studies(17,23) showed a negative relation as well; however, after adjusting for body fat percentage but not waist-hip ratio in one study(23) and BMI, age, and sex in the other study,(17) the relations disappeared, supporting the idea that adiposity may confound the relations between lipids and BMD. In our study, the relations between HDL-c and QUS were still significant after adjusting for BMI. In the cited studies, the relation between BMD and adiposity was remarkable high. In a small subgroup (n = 480) that included participants from Amsterdam who had undergone a DXA scan, we found DXA measurements to be more subject to BMI than QUS measurements. In this subgroup, univariate models also showed an inverse relation between HDL-c and DXA of the hip and lumbar spine. After adjusting for BMI and age, the relationships partly disappeared; only the relationship among men between HDL-c and BMD of the hip remained significant (data not shown).
In another large study(13) among 1176 elderly women, univariate models showed a relation between lipids and BMD as well. After adjusting for the severity of aorta calcification, the relations were no longer significant, which led them to conclude that the relationship is an indirect one, through promotion of atherosclerosis. However, the cited study failed to show significant relations between lipids and aorta calcifications.
Epidemiological studies have shown that low HDL-c levels in humans with type 2 diabetes are accompanied by an increased BMD.(24–27) Moreover, the Tromsø study(25) showed that low HDL-c levels provide protection from nonvertebral fractures. The HDL-c level is a powerful, independent indicator for the risk of CVD. In addition to reverse cholesterol transport and cellular cholesterol efflux, recent studies found a range of atheroprotective properties ascribed to HDL-c, including several anti-inflammatory, antioxidative, and antithrombotic functions.(28–30) Recent years have seen a growing interest in the development of HDL-c–raising medication. A combination of these drugs and cholesterol-lowering ones could provide a good treatment of CVD.(28,29) However, the question arises whether the protection from nonvertebral fractures by low levels of HDL-c could be nullified by HDL-c–raising drugs. Given that osteoporosis and related bone fractures are an increasing health risk in our aging societies, elucidation of the relation between HDL-c and BMD before the introduction of HDL-c–raising medication is of great importance. The mechanism behind the negative relationship between HDL-c and BMD is yet unknown. Several studies on mice have examined the influence of cholesterol, oxysterols (oxygenated derivatives of cholesterol) in particular, on the differentiation of mesenchymal stem cells (MSCs). MSCs are the progenitors of osteoblasts, chondrocytes, myocytes, adipocytes, and fibroblasts.(31–34) For several types of osteoporosis, it was found that, in bone marrow tissue, adipogenic differentiation of MSCs was favored at the cost of osteogenic differentiation. A study about the factors and mechanisms that drive the differentiation of MSCs showed that specific oxysterols stimulate the osteogenic differentiation of MSCs while inhibiting their adipogenic differentiation.(32) Given that HDL-c is capable of removing oxysterols from peripheral tissues, a high HDL-c level would be detrimental to osteogenic differentiation. This could explain the observed negative relation between HDL-c and QUS. In this line of thinking, the combination with a low TC could lead to even lower oxysterol levels in bone cells, possibly explaining the strong relation that was found between TC/HDL-c ratio and QUS.
Initially, elevated TC was supposed to be the related factor to both CVD and osteoporosis. Therefore, several studies examined the relation between TC and BMD. Their results are contradictory because absent,(18,35) positive,(14,36) and negative(37) relations have been found. Some studies also showed that lipid-lowering statins increase BMD values and reduce the risk of bone fracture.(7,9) However, a large case-control study(38) showed that this is not an effect of the lowered cholesterol, but of the pleiotropic effect applied to lipophilic statins only. In this study, only among women was a positive relation between TC levels and QUS found in univariate models. After adjusting for confounding, this relationship did not remain significant. Our findings indicate that not TC but HDL-c is most strongly related to BMD.
The results of this study are in agreement with some studies cited above. However, in this study, QUS was used as a parameter for BMD. Our study was carried out among a large population (n = 1255) and took several confounding variables into account. We excluded subjects taking bisphosphonates, estrogens, and statins, all known to affect bone metabolism. Besides the relationship between lipids and QUS, also the relationships between lipids and CVD and between CVD and QUS were studied. In agreement with the literature, our study population showed HDL-c to be most strongly related to CVD. Also, a significant inverse relationship between QUS and peripheral arterial diseases was found, which remained significant after adjustment for lipids (HDL-c, TC/HDL-c ratio, and TC). To our knowledge, this is the first report on the relation between TC/HDL-c ratio and QUS. We found the TC/HDL-c ratio to have a stronger correlation with BMD than HDL-c alone. Our analysis is restricted to men and women with serum HDL-c levels and TC/HDL-c ratios within the normal range and just slightly elevated serum TC levels. Therefore, findings are not generalizable to men and women with dyslipidemia.
In this study, we found inverse associations between QUS and HDL-c(39,40) and between peripheral arterial diseases and QUS.(41,42) Whereas low levels of testosterone are related to both low QUS and high HDL-c levels and vitamin D insufficiency [low serum 25(OH)D] with both low QUS and a higher prevalence of peripheral arterial diseases, we controlled all models for these confounders. After adjustment, all results remained significant, and the associations became even slightly stronger.
There are some limitations to our study. First, we examined primarily QUS as a measurement for BMD. Second, serum levels of triglyceride (TG) were not determined in the LASA study. Therefore, we could not study the relationships between TG, LDL-c, and QUS.
In conclusion, our analysis showed that the lipid profile that is favorable in the prevention of CVD (i.e., high levels of HDL-c and a low TC/HDL-c ratio) is at the same time unfavorable for BMD. These results indicate that HDL-c levels do not explain the association between osteoporosis and CVD. Because osteoporosis and related bone fractures pose an important health risk, further elucidation of the relation between lipids and QUS is of great importance.
We thank Els Lommerse, Ellen Tromp, Saskia Pluijm, Mariëtte Westendorp-de Serière, Marleen van der Horst, and Jan Poppelaars for help in collecting and processing the data; our colleagues from the Laboratory of Clinical Chemistry for the biochemical estimations; and our colleagues from the Department of Nuclear Medicine for performing the BMD measurements. This study is based on data collected in the context of the Longitudinal Aging Study Amsterdam (LASA) and was funded by the Ministry of Health, Welfare, and Sports of the Netherlands and by Prevention Fund (currently ZonMw), The Hague (Grant 28-2551).