Relationship Between Lipids and Bone Mass in 2 Cohorts of Healthy Women and Men
- First Online:
- Cite this article as:
- Adami, S., Braga, V., Zamboni, M. et al. Calcif Tissue Int (2004) 74: 136. doi:10.1007/s00223-003-0050-4
- 241 Views
A number of recent findings seem to indicate that fat and bone metabolism are strictly connected. We investigated the relationship between lipid profile and bone mineral density (BMD) in 236 either pre- or postmenopausal women, aged 35–81 years, attending our osteoporosis center (“clinic group”). In order to verify the consistency of the results, 265 men and 481 women aged 68–75, participating in a population-based epidemiological investigation (“community cohort”), were also studied. Lumbar spine, femoral neck, total hip and total body BMD, total body fat, % fat mass and lean mass were measured using dual energy X-ray absorptiometry (DXA). In the clinic group, lumbar spine and hip BMD Z score values were both strongly related to all measured serum lipids: the relationship was negative for HDL cholesterol (P < 0.05) and Apo A lipoprotein (P < 0.000) and positive for LDL cholesterol (P < 0.05), Apo B lipoprotein (P < 0.001) and triglycerides (P < 0.05). When BMD values were adjusted for body weight and BMI, most relationships remained statistically significant. In the community cohort, total body and hip BMD values were strongly related in both men and women to age, body weight, height, BMI, fat mass, lean mass, % fat mass. Total body and hip BMD were significantly related to serum lipids in both women and men. The relationship was negative for HDL cholesterol and positive for total cholesterol, triglycerides and LDL cholesterol. Most of these relationships (triglycerides, HDL cholesterol, LDL/HDL cholesterol ratio in women, and all measured lipids in men) remained statistically significant (P values ranging from 0.000 to 0.03) when the BMD values were adjusted also for anthropometric measures (body weight, height, fat mass). This study demonstrates for the first time that the lipid profile is strictly related to bone mass in both men and women. The interpretation of this association remains hypothetical but it might open new perspectives for understanding the mechanisms controlling bone metabolism.
KeywordsSerum lipidsBone massOsteoporosisBone metabolismStatins
Osteoporosis is a major cause of morbidity and mortality among elderly persons but the pathogenesis of the condition has not been fully elucidated. In the last few years a strong relationship between fat tissue and bone metabolism has become apparent. Bone loss is associated with an expansion of adipose tissue in the marrow [1, 2, 3, 4, 5, 6] and osteoblasts and adipocytes share a common progenitor arising from the stromal cells in the marrow [7, 8, 9, 10]. Products of lipoprotein oxidation and an atherogenic diet inhibit preosteoblast differentiation [6, 11] and result in reduced bone mineralization. A mutation in the low-density lipoprotein (LDL) receptor-related protein 5 (LRP5) [13, 14] and polymorphisms of apolipoprotein E genotype  are associated with alteration in bone mineralization. In a number of studies it has been shown that statins, widely used as lipid-lowering agents, may stimulate bone formation and are often of benefit in the preservation of bone mass and prevention of osteoporotic fractures [16, 17, 18]. Hormone replacement therapy exerts strong effects on both bone and lipid metabolism . Aminobisphosphonates have been shown to interfere with the metabolic pathway leading to cholesterol synthesis  and their intravenous administration is associated with beneficial effects on lipid profile .
Altogether these observations argue for a relationship between lipid and bone metabolism, and might suggest a negative effect of an atherogenic lipid profile on bone formation. We then decided to investigate the relationship between lipid profile and bone mass. The first analysis was carried out in individuals referred for assessing osteoporosis risk. Since the results were somewhat unexpected in both men  and women (the worse the lipid profile the better the bone mass), we extended our analysis to the data of a cohort of elderly men and women participating in an epidemiological investigation on the risk factors for disability that included some bone mass and serum lipid measurements.
Material and Methods
The initial study group included all pre- and postmenopausal women, referred by their general practitioners (GPs) to our osteoporosis center from September to December 1999, for screening and assessment of osteoporosis risk from the now called “clinic group.” Subjects all signed an informed consent previously submitted and approved by our Ethics Committee.
The second cohort included both women and men aged 68–75 years participating in a population-based, longitudinal, epidemiological investigation (called “community cohort”). The main aim of this study was the identification of risk factors for disability and mortality in subjects homogeneous for age and ethnicity. The participants were randomly selected from the Health Registry of Verona, with the collaboration of their GPs who referred all selected individuals to our Epidemiololgical Centre. Seventy-seven percent of the women and 67% of the men agreed to participate in the study and to sign the informed consent. The most frequent reason for nonparticipating was the lack of time or because some persons preferred to be followed by their own specialists. The results reported here are those of the baseline assessment.
Exclusion criteria for both groups of subjects were present or past endocrine, metabolic bone diseases and rheumatoid arthritis, treatment with lipid lowering or anti-diabetic agents, bisphosphonates (within 5 years), estrogens or calcitonin or corticosteroids (within the last 12 months). Overall, 78 subjects had to be excluded, 18 of the “clinic group” and 60 of the “community cohort.” None of the subjects had malignancies, severe chronic liver, heart or renal diseases and they were all able to attend our out-patients clinic. The “clinic group” eventually included 195 postmenopausal and 41 premenopausal women (Table 1). The “community cohort” included 481 women and 265 men (Table 2).
Principal characteristics of the women of the “clinic group” (no. 236)
HDL chol (mmol/l)
Apo A (mmol/l)
LDL chol (mmol/l)
Apo B (mmol/l)
Hip BMD (Z-score)
Spine BMD (Z-score)
Principal characteristics (mean and standard deviation, SD) for women and men of the “community cohort”
Women (n = 481)
Men (n = 265)
HDL chol. (mmol/l)
LDL chol. (mmol/l)
% Fat mass
Total body BMD (g/cm2)
Hip BMD (g/cm2)
In all subjects background information included lifestyle (consumption of dairy products, calcium and protein intake, outdoor activity, smoking habit) and medical history. The assessment procedures for all these risk factors are those adopted for the ESOPO study .
The patients were asked whether they had been confined to bed (dichotomous variable) for more than 2 months at ages <25 years, between 25 years of age and last year, and within the last year. The time spent on daily walking or biking was categorized for never, less than 30 min, between 30 to 60 min and more than 60 min. The mean daily consumption of dairy products was specifically recalled for milk, yoghurt, soft and hard cheese. From these values the individual mean dairy calcium intake was calculated. The dairy product intake together with the mean intake of meat and legumes was used to calculate the daily protein intake. Smoking habit was assessed for the present and the past and also the number of cigarettes smoked daily. Daily and total alcohol intake of alcoholic beverages were calculated. Coffee or tea consumption was calculated in terms of cups per day.
All subjects underwent physical examination. Body weight and height were measured using a balance beam scale and a Harpenden stadiometer, respectively. Height and weight were used to calculate body mass index (BMI, kg/cm2).
All subjects had a routine biochemical assessment at the time of the study or within the previous 3 months that included serum calcium, phosphate, albumin, creatinine, serum protein profile, alkaline phosphatase, and eritrocyte sedimentation rate.
For the “clinic group” the relationship between BMD and lipid profile was the primary endpoint of the study. Thus, the investigations concentrated on measurements of BMD at both the spine and the hip and on a comprehensive evaluation of lipid profile (see Results). In the “community cohort” the primary end point was different and thus per protocol investigations included body composition by DXA but not BMD at the spine and a somewhat simplified lipid profile not including serum lipoproteins.
Lumbar spine, total hip, total body BMD, total body fat, total fat-free mass tissue (lean mass) and percent total fat (% fat) were measured using dual energy X-ray absorptiometry (DXA) (Hologic QDR 4500, Waltham, MA, USA). In the “community cohort” with a narrow age range the BMD values were expressed in absolute values (g/cm2), whereas for the “clinic group,” including both pre- and postmenopausal women in whom the age range was wide, the BMD values were expressed in terms of Z score, adjusted also for menopausal state. Calibration was performed daily and a lumbar spine phantom was scanned at least twice a week. The coefficients of variation (CV) for DXA were assessed by triple measurements in 30 subjects and they were less than 1.3% at all skeletal sites and less than 2% for body composition.
Fasting blood and spot urine samples were collected between 7 and 9.30 a.m. and stored in aliquots at −80°C while awaiting analysis. Apolipoprotein A-I and B measurements were performed on Behring Nephelometer Analyzer (BNA) (Behringwerke, Marburg AG, Germany) employing anti-apo A-I and apo B nephelometric antisera (Behringwerke). Results were evaluated by logit-log function of light scattering intensities versus respective concentrations. Intra-assay and interassay CVs of apo A-I and apo B lipoproteins were less than 3% and 5%, respectively. Serum cholesterol was measured enzymatically using the Cobas Integra Roche analyser (F. Hoffman-La Roche Ltd. Basilea, Switzerland). Analytical imprecision was 0.95% at a level of 3.87 mmol/l and 0.96% at 9.49 mmol/l. Serum triglycerides were assayed using the glyceryl dehydrogenase reaction following enzymatic hydrolysis of the glycerides on the Cobas Integra Roche analyzer (F. Hoffman-La Roche Ltd. Basel, Switzerland). Analytical imprecision was 1.4% at a level of 1.49 mmol/l and 1.5% at 5.19 mmol/l. High-density lipoprotein (HDL) cholesterol was measured after precipitation of low-density lipoproteins (LDL) and very low-density lipoproteins with polyanions and phosphotungstic acid/magnesium chloride. The supernatant was assayed enzymatically on the Cobas Integra Roche analyzer (F. Hoffman-La Roche Ltd. Basel, Switzerland) and had an analytical imprecision of 3.8% at a concentration of 1.03 mmol/l and 3.75% at 3.98 mmol/l.
LDL cholesterol was calculated using the Friedwald formula . In none of the specimens were the triglyceride values greater than 3.33 mmol/l and chilomicrons were not present. In our laboratory the total error for LDL cholesterol measurement is <7.5%.
Serum 25-hydroxyvitamin D (25OHD) was measured by radioimmunoassay (RIA) using a commercial kit (detection limit 1.5 ng/ml; DiaSorin, Italy); serum intact PTH 1-84 was measured using an immunoradiometric method (DiaSorin, Italy) with a sensitivity of 0.7 pg/ml. The inter-assay coefficients of variation were 10.1 and 10.9% for 25OHD and PTH, respectively.
All statistical procedures were carried out with SPSS for Windows version 10.0 (SPSSC, Inc., Chicago). The association between each serum lipid and all other anagraphic and anthropometric variables (bone mass, body composition, age, years since menopause) were assessed by simple linear regression. Since both serum lipids and bone mass values were often significantly related to age, body weight, BMI, or fat mass, the relationship between bone mass and serum lipids was re-assessed by adjusting (multivariate and partial correlation analysis) bone mass values for these variables.
Overall, 39 subjects had either calcium or vitamin D insufficiency, but in none was this associated with secondary hyperparathyroidism (serum PTH >80 pg/ml). Serum 25OH vitamin D was measured only in subjects in whom vitamin D deficiency was suspected (house-bound subjects and those having a very low animal fat diet), allowing the identification of 28 subjects with 25OHD concentrations lower than 12 ng/ml. A dietary calcium intake lower than 600 mg/day was found in 11 subjects.
The women of the “clinic group” ranged in age from 35 to 81 years and their mean BMD values were somewhat lower than those expected in a control population, reflecting possibly their recruitment source (Table 1). For this reason and for the wide distribution of age the BMD values are reported as Z score by using the Hologic and the NHANES reference values for spine and hip BMD, respectively. In a survey carried out in Northern Italy the normative range for spine and hip BMD has been found to be basically superimposable to those adopted here . Lumbar spine and hip BMD Z score values were strongly related to all anthropometric variables (data not shown) but in a “step-wise” multivariate regression model, a significant positive relationship persisted only for body weight. Spine and hip BMD Z score values were both strongly related to all measured serum lipids (triglycerides, HDL and LDL cholesterol, Apo A-I and Apo B lipoproteins) (Table 3). The relationship was negative for HDL cholesterol and APO A-I and positive for LDL cholesterol, Apo B and triglycerides. When BMD values were adjusted for body weight by multivariate and partial correlation analysis most relationships remained statistically significant with the only exception of LDL cholesterol with hip BMD. As an example, Figure 1 illustrates the relationship between HDL cholesterol and Z score BMD values, both adjusted for body weight.
Correlation coefficients with P values (in brackets) for serum lipids (mmol/l) in women of the clinic group as related to body weight and spine and hip BMD
Body weight (kg)
Lumbar spine BMD (z-score)
Lumbar spine BMD adjusted for body weight
Hip BMD adjusted for body weight
None of these relationship were affected by covariates such as smoking, physical activity, calcium or protein intake. The exclusion of subjects with vitamin D insufficiency or premenopausal women modified neither the direction nor the statistical significance of all relationships.
In the subjects of the “community cohort” (Table 2) the age range per protocol was very narrow and the mean BMD values were similar to those found in the reference population. Total body and hip BMD values were significantly related in both men and women to age, body weight, height, BMI, fat mass, lean mass, % fat mass. A weak association was found between BMD values and smoking, physical activity score, and calcium intake  in both men and women. In a multivariate regression model, BMD values were significantly related to age, body weight, lean mass and BMI in women and to age, lean mass, % fat mass or BMI in men (Tables 4 and 5).
Correlation coefficients (r) and significance level (p) of the linear relationship between BMD values and some characteristics of the women of the “community cohort”
Total body BMD
P (Multivariate model)
P (Multivariate model)
Fat mass (g)
Lean mass (g)
% Fat mass
Correlation coefficients (r) and significance level (P) of the linear relationship between BMD values and other characteristics of the men of the “community cohort”
Total body BMD
P (Multivariate model)
P (Multivariate model)
Fat mass (g)
Lean mass (g)
% Fat mass
Measured serum lipids (mostly triglycerides and HDL cholesterol) were significantly related to age and body weight or fat mass (results not shown) in both genders. In men, total body and hip BMD were significantly related to serum lipids: negatively for HDL cholesterol and positively for triglycerides and LDL cholesterol (Table 7). In women these relationships were statistically significant for triglycerides, HDL cholesterol and LDL/HDL cholesterol ratio (Table 6). Most relationship remained statistically significant even though to somewhat a lesser degree (Tables 6 and 7) when both BMD and lipid values were adjusted for the body mass and composition indices to which they were significantly related in the multivariate function (Tables 4 and 5). Figures 2 and 3 show four examples of such relationships in men and women, respectively. Smoking habit, physical activity, calcium and protein intake also had no effect in this cohort of subjects on the relationship between serum lipids and BMD values.
Correlation coefficients (r) with P values (in brackets) for serum lipids (mmol/l) in men of the “community cohort”
Total body BMD
Total body BMD adjusted for age, lean mass, % fat mass
Hip BMD adjusted for age and body weight
Correlation coefficients (r) with P values (in brackets) for serum lipids (mmol/1) in women of the “community cohort”
Total body BMD
Total body BMD adjusted for age, body weight, lean mass, BMI
Hip BMD adjusted for age and body weight
This study demonstrates for the first time that the lipid profile is moderately related to BMD both in men and women. We have been able to confirm it in different cohorts of subjects. We first found these relationships in men  and here in a group of women recruited from our osteoporosis center. In order to exclude an influence of age on both lipid profile and BMD we extended our observation to an additional cohort of men and women with a very narrow age range and representative of the general population, yielding the same results. Our data indicate that the worse the lipid profile (lower HDL cholesterol and higher LDL cholesterol or triglycerides) the higher the bone mass. This association is only partly driven by the relationship of body weight and fat mass with both lipid profile and BMD, since it remained highly significant after adjusting BMD values for all indices of body mass and body composition. The relationship was not influenced by all known determinants of bone mass such as calcium intake, physical activity, smoking, or parity. Our findings are somewhat in contradiction with some epidemiological studies showing a relationship in postmenopausal women between cardiovascular mortality and low BMD, or osteoporotic fracture risk [26, 27, 28]. From our observation it appears that cardiovascular chronic diseases are responsible for increased risk of osteoporosis rather than the opposite. In addition, there are evidences that hyperlipemia and minimally oxidized low-density lipoprotein exert negative effects on bone metabolism [6, 8]. These findings suggest an association between lipid profile and bone mass opposite to those reported here. On the other hand it has been recently reported that the cholesterol biosynthetic pathway plays a permissive role in osteoblastic differentiation . We had been aware of the contradictory value of our results since the first analysis of the data on postmenopausal women attending our out-patient clinic for osteoporosis assessment. For this reason we investigated the males of our out-patients clinic  and then two cohort of elderly persons, representative of the general population and then less likely to be affected by observational biases. From a thorough search in the literature we found 2 other reports also showing a negative association between BMD and HDL cholesterol even though in a limited number of women [29, 30].
Due to its cross-sectional nature we are not able to clarify whether lipid profile is associated with peak bone mass acquisition or to age-related bone loss. An additional limitation of the study is that the subjects with an unfavorable lipid profile were likely aware of it and this might have driven them to a lower calorie diet, potentially responsible for some bone losses. However, the associations we found were controlled for dairy product, total calories and protein intake, and vitamin D levels.
At the moment we are unable to provide a documented explanation of our results and clearly, other longitudinal studies of the rate of bone loss, such as bone markers and serum steroidal hormones, are warranted. However, our results may help to reinterpret the association between statin use and lower risk of osteoporotic fracture. In fact, it is possible that the effect of statin is only apparent, and at least partly driven by the more favorable bone mass and less favorable lipid profile of the statin users.
In conclusion the present clinical study indicates that lipid profile is consistently associated with bone mass. The interpretation of this association remains hypothetical but it might open new perspectives for understanding the mechanism controlling bone metabolism.