Abstract
Background:
It is uncertain if lipids or lipoproteins are associated with osteoporotic fractures. In this study, incident hip fracture risk according to conventional lipid levels and lipoprotein levels and sizes was examined.
Methods:
We followed 5832 participants aged ≥65 years from the Cardiovascular Health Study for hip fracture for a mean of 13.5 (SD 5.7) years. Standard enzymatic methods were used to determine lipid levels (HDL-c, LDL-c, triglycerides). Nuclear magnetic resonance spectroscopy was used to measure lipoprotein fractions (VLDL-P, LDL-P, HDL-P) in a subset of 1849 participants.
Results:
We documented 755 incident hip fractures among women (1.19 fractures per 100 participant years [95% CI, 1.04, 1.35]) and 197 among men (0.67 fractures per 100 participant years [95% CI, 0.41, 1.10]) over an average follow-up. HDL-c and LDL-c levels had statistically significant non-linear U-shaped relationships with hip fracture risk (HDL-c, p=0.009; LDL-c, p=0.02). Triglyceride levels were not significantly associated with hip fracture risk.
In fully adjusted conjoint models, higher VLDL-P concentration [HR per 1-standard (SD) increment 1.47 (1.13, 1.91)] and size [HR per 1-SD increment 1.24 [1.05, 1.46]) and higher HDL-P size (HR per 1-SD increment 1.81 [1.25, 2.62]) were all associated with higher hip fracture risk.
Conclusions:
Lipids and lipoproteins are associated with hip fracture risk in older adults. The associations are complex. Mechanistic studies are needed to understand these findings.
Keywords: hip fracture, lipid, lipoprotein particle
INTRODUCTION
There are many reasons to posit a direct or indirect association between serum lipid levels, osteoporosis, and hip fracture risk. Laboratory studies show that cholesterol can extend the survival of osteoclast-like cells, contributing to osteoporosis (1). High levels of total cholesterol are associated with low 25(OH) vitamin D levels (required for the absorption of calcium (2)) and with inflammatory cytokines (e.g., TNF-α, IL-1, IL-6) which uncouple bone remodeling (3). Metabolomic studies demonstrate low bone mineral density to be associated with lipids (4). Clinical studies also show an association of lipids with bone disease. Chronic medical conditions associated with dyslipidemia – e.g., metabolic syndrome, obesity, and diabetes – are often associated with increased fracture risk (5–7). Dyslipidemia is a strong risk factor for atherosclerosis. Clinical and subclinical atherosclerosis of large and medium sized blood vessels is associated with osteoporosis and fracture risk (8, 9), as well as with disrupted distal bone capillary blood flow (10). Statin medications increase bone mineral density, suggestive of a causal link between the lowering of lipid levels and improved bone health (11). Last, the skeleton takes up a large proportion of a post-prandial lipoprotein load (12).
Despite these reasons, the reported associations of lipid levels with hip fracture risk and, by extension, with osteoporosis, are mixed (13 – 16). Certain data suggest positive associations, some report no associations, and others negative associations. Most studies are cross-sectional and small; the ethnic compositions of the cohorts vary; and the ages of the cohorts often include young people in whom osteoporotic fractures are uncommon. Prior studies examine lipid fractions individually but do not adjust for the effect of the other lipid fractions on fracture risk. Finally, another factor, heretofore not considered, is the variety of lipoprotein subclasses that carry lipids in the blood.
In this longitudinal study from the Cardiovascular Health Study (CHS) the association of baseline lipids and lipoproteins with incident hip fracture risk is examined in a cohort of adults aged ≥65 years at baseline who were well-phenotyped and followed for hip fracture for up to 20 years.
METHODS
The CHS is a prospective observational study of community dwelling adults, aged ≥65 years at study entry, from four US communities drawn from Medicare lists (17). In 1989–1990, 5201 participants were recruited, followed by an additional 687 predominantly African American participants in 1992–1993. All participants gave informed consent prior to study entry. Institutional review board approval was received at all clinical sites. From 1989–1990 to 1998–1999, participants were seen in clinic annually and had telephone contact midway between clinic visits. Following the 1998–1999 visit, participants continued to be contacted biennially to update hospitalizations, incident diagnoses, and medications. Surveillance for hip fractures ended June 30, 2015.
Analytic Cohorts:
Lipid Level Cohort: Lipid data collected at the baseline 1989–1990 and 1992–1993 visits were combined for lipid analyses. A total of 56 participants had missing lipid analytes and were excluded from the lipid level cohort. Participants with fractures prior to the baseline date, including hip fractures, were not excluded. Lipoprotein Cohort: Plasma collected at baseline from a total of 1622 CHS participants from the 1989–1990 cohort, and 228 African American participants from the 1992–1993 cohort underwent NMR spectroscopy at LipoScience (Raleigh, NC) for determination of lipoprotein subclasses as part of a nested case-cohort study (18). All participants in the lipoprotein cohort were free of clinical cardiovascular disease at baseline. Participants comprised several groups: [1] 249 who were free of all subclinical atherosclerosis at baseline [as defined in CHS (19)]; [2] 492 who were free of incident myocardial infarction or angina through June 30, 1995; [3] 222 with incident angina but not myocardial infarction through June 30, 1995; [4] 213 with incident myocardial infarction through June 30, 1995; [5] 200 with incident stroke through June 30, 1995; and [6] 246 with subclinical cerebral infarcts by cranial magnetic resonance imaging through June 30, 1995. Participants free of incident myocardial infarction or angina and those free of all subclinical disease were sampled randomly from the CHS population. The 228 randomly selected African American participants were free of myocardial infarction or stroke. Participants with fractures prior to study baseline were not excluded.
Study Outcome:
The study outcome was incident hip fracture. Fracture data were obtained through participant report and confirmed through hospital medical records, including discharge summaries, gathered every 6 months from the 1996–1997 visit through June 30, 2015. Data were checked against Medicare claims data to identify any hospitalizations not reported by participants. Hip fracture was defined using the International Classification of Diseases, Ninth Revision (ICD-9) code 820.xx. Pathological fractures (ICD-9 code 773.1x) and motor vehicle accidents (E810.xx-E825.xx) were excluded.
Primary Predictors of Hip Fracture:
Lipid Levels: Standard enzymatic methods were used to determine total cholesterol (TC), HDL-c, and triglycerides, standardized according to the Centers for Disease Control and Prevention guidelines as described previously (20). LDL-c levels were calculated using the Friedewald equation. Lipoprotein Subclasses: Aliquots (0.5 mL) of EDTA plasma stored at −80°C at the CHS central laboratory were shipped on dry ice to LipoMed, Inc, for nuclear resonance resonance (NMR) lipoprotein subclass analysis. Lipoprotein particle diameters were measured with an automated NMR spectroscopic assay. In brief, lipoprotein subclasses emit characteristic lipid methyl group NMR signals, and the signal amplitude reflects particle concentration. The categories examined were very-low-density lipoprotein particles (VLDL-P), low-density lipoprotein particles (LDL-P) and high-density lipoprotein particles (HDL-P). Particle concentrations are in nanomoles per liter for VLDL-P and LDL-P and micromoles per liter for HDL-P. Weighted-average VLDL-P, LDL-P, and HDL-P sizes (in nanometer diameter units) were calculated as the particle size of each subclass multiplied by its relative mass percentage as estimated from the amplitude of its NMR signal.
Covariates:
Analyses were adjusted for baseline factors associated with lipid levels and hip fracture risk. These included: age, sex, race, smoking status (never, past, current), current alcohol intake (none, less than 7 drinks per week, 7 or more drinks per week), presence of self-reported and adjudicated diabetes (DM) and cardiovascular disease (CHD [angina, MI, angioplasty, bypass surgery], stroke), energy expended per week (kcal), estimated GFR based on cystatin-C level, C-reactive protein (CRP) level (20), hypertension (HTN), frailty status (none, pre-frail, frail) based on Fried criteria (21), difficulties with activities of daily living (ADLs) and instrumental ADLs (iADLs), weight and height.
Statistical Analysis:
Due to skewed lower and upper tails of the distributions of the lipids, we winsorized HDL-c, LDL-c, and triglycerides at 2% (30, 63.7, 59 mg/dl, respectively) and at 98% (93, 210.8, 352 mg/dl, respectively). Lipoproteins were similarly winsorized.
Incidence rates of hip fracture, total and by sex, were calculated with quasi-Poisson models with offset to accommodate time at risk.
Multivariable Cox hazards models estimated the hazard ratios (HR) and 95% confidence intervals (CI) of incident hip fracture associated with a standard deviation higher exposure. We used nested models (M) adjusting for factors as: M0: unadjusted; M1: age, sex, race, clinic. M2: M1 + smoking, alcohol, HTN, eGFR, DM, physical activity score, CRP (log base 2), estrogen, weight, height, and prevalent CVD. We included models with a single lipid exposure as well as all three lipids simultaneously in the model to estimate mutually adjusted HRs.
To study linearity in the models, the functional association of lipids with incident hip fracture using generalized additive models (GAMs) with splines was examined, observing U-shaped relationships. Using a permutation approach for significance tests (22), we observed significant non-linear associations with LDL-c and HDL-c. In follow-up analyses, we categorized LDL-c and HDL-c into quintiles and used the middle quintile as a reference category in the above-described model M2 with all 3 exposures included.
Analyses were conducted using R (R Foundation for Statistical Computing, 2019).
RESULTS
Baseline characteristics of the cohort, categorized by sex, are shown in Table 1. Women comprised 57.5% of the cohort. Mean age was ~73 years. Men were more likely to be former smokers and to drink ≥7 alcoholic drinks per week. Slightly less than 50% of men and women had more than 12 years of education. Women had a higher prevalence of difficulties with iADLs than men but both sexes were equivalent in the prevalence of pre-frailty and frailty. Men had a higher prevalence of CHD than women, but renal function was approximately the same. Diabetes prevalence was roughly equivalent. Women were more likely to use calcium supplements than men. There was low use of vitamin D supplements and statins.
Table 1:
Baseline characteristics of the Cardiovascular Health Study cohort categorized by sex.
| WOMEN N= 3351 |
MEN N=2481 |
|
|---|---|---|
| Demographic | ||
| Black race (%) | 16.6 | 13.6 |
| Age (yrs) std dev | 72.5 (5.5) | 73.3 (5.7) |
| Weight (kg) | 67.9 (14.2) | 79.3 (12.6) |
| Height (m) | 1.6 (0.1) | 1.7 (0.1) |
| ≥7 | 8.4 | 17.4 |
| Never | 57.0 | 32.1 |
| Education level ≥12yrs (%) | 40.7 | 47.6 |
| Kcal expended/week (median, IQ range) | 412 (10–1102) | 967 (319–2243) |
| Difficulty ADL (%) | 9.7 | 5.6 |
| Difficulty iADLs (%) | 30.6 | 19.0 |
| Diabetes (%) | 14.2 | 18.9 |
| Frail | 8.0 | 5.1 |
| Prevalent CVD (%) | ||
| CHD* | 15.5 | 24.9 |
| MI | 6.3 | 14.0 |
| CHF | 4.1 | 5.3 |
| Stroke | 3.1 | 5.6 |
| Hypertension | 60.6 | 55.9 |
| Use of antihypertensive medications | 49 | 45.1 |
| Renal function based on cystatin C (eGFR ml/minute/1.73m2) | 79.7 (20.2) | 74.3 (18.8) |
| Medications (%) | ||
| Thiazide | 22.6 | 15.1 |
| Loop diuretic | 7.3 | 6.4 |
| Statin | 2.8 | 1.5 |
| Supplementation Use (%) | ||
| Calcium | 26.4 | 9.4 |
| Vitamin D | 0.3 | 0.1 |
angina, MI, angioplasty, bypass surgery
Supplementary Table 1 provides baseline lipid values of men and women winsorized at the 2nd and 98th percentiles. Women had higher LDL-c and HDL-c levels than men. Triglyceride levels were equivalent.
Mean (SD) follow up was 13.5 (7.1) years (median 13.2 years [IQR, 7.9, 19]). We documented 755 incident hip fractures among women (1.19 fractures per 100 participant years [95% CI, 1.04, 1.35]) and 197 among men (0.67 fractures per 100 participant years [95% CI, 0.41, 1.10]).
Lipid Levels:
The risks for hip fracture associated with winsorized individual lipid analytes sequentially adjusted for covariates were not statistically significant as linear associations, together or for men and women separately, or when adjusted for the association of the other lipid analytes with hip fracture risk (Table 2).
Table 2:
Hazard ratios for hip fracture associated with a one standard deviation increase in standard lipid variables in the Cardiovascular Health Study.
| ALL | WOMEN | MEN | ||||
|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | |
| LDL-c | ||||||
| M0 | 0.99 | 0.92, 1.07 | 0.99 | 0.91, 1.08 | 0.88 | 0.74, 1.03 |
| M1 | 0.96 | 0.90, 1.04 | 0.98 | 0.90, 1.07 | 0.92 | 0.78, 1.09 |
| M2 | 1.00 | 0.92, 1.09 | 1.02 | 0.93, 1.13 | 0.92 | 0.76, 1.11 |
| HDL-c | ||||||
| M0 | 1.13 | 1.05, 1.22 | 1.07 | 0.98, 1.17 | 0.93 | 0.77, 1.12 |
| M1 | 1.07 | 0.98, 1.16 | 1.10 | 1.00, 1.21 | 0.92 | 0.75, 1.11 |
| M2 | 1.06 | 0.96, 1.17 | 1.08 | 0.97, 1.21 | 0.90 | 0.70, 1.15 |
| TRIG | ||||||
| M0 | 0.97 | 0.89, 1.06 | 0.97 | 0.87, 1.07 | 0.95 | 0.80, 1.14 |
| M1 | 0.95 | 0.87, 1.05 | 0.94 | 0.85, 1.05 | 0.99 | 0.83, 1.19 |
| M2 | 0.97 | 0.88. 1.08 | 0.96 | 0.86, 1.09 | 1.03 | 0.84, 1.26 |
| HDL-c M0 | 1.14 | 1.05, 1.24 | 1.06 | 0.96, 1.18 | 0.88 | 0.71. 1.08 |
| LDL-c M0 | 1.00 | 0.93, 1.09 | 1.00 | 0.92, 1.10 | 0.87 | 0.74, 1.03 |
| Trig M0 | 1.00 | 0.90, 1.11 | 0.94 | 0.83, 1.07 | 0.92 | 0.74, 1.13 |
| HDL-c M1 | 1.04 | 0.95, 1.15 | 1.08 | 0.97, 1.21 | 0.89 | 0.72, 1.10 |
| LDL-c M1 | 0.97 | 0.90, 1.05 | 1.00 | 0.91, 1.09 | 0.92 | 0.78, 1.08 |
| Trig M1 | 0.94 | 0.84, 1.05 | 0.92 | 0.81, 1.06 | 0.96 | 0.78, 1.19 |
| HDL-c M2 | 1.05 | 0.94, 1.17 | 1.08 | 0.95, 1.22 | 0.88 | 0.67, 1.15 |
| LDL-c M2 | 1.00 | 0.92, 1.10 | 1.04 | 0.95, 1.15 | 0.92 | 0.76, 1.11 |
| Trig M2 | 0.95 | 0.84, 1.07 | 0.93 | 0.80, 1.07 | 1.00 | 0.79, 1.27 |
TRIG – triglycerides
M0 – unadjusted
M1: age, gender [for “all” only], race adjusted.
M2: M1 + smoking, alcohol, HTN, eGFR, DM, energy expended per week, frailty, CRP, ADLs, iADLs, weight, height, and prevalent CVD.
Generalized additive models of winsorized LDL-c, HDL-c, and triglyceride levels with hip fracture risk, with each lipid fraction adjusted for the effect of the other and for covariates from model 2, are shown in Figure 1. LDL-c and HDL-c levels both had significant non-linear, U-shaped associations with fracture risk, while triglyceride levels showed no significant association with hip fracture (permutation-based p-values: LDL-c, p=0.02; HDL-c, p=0.009; triglyceride, p=0.92). To further examine the potential nonlinear association of LDL-c and HDL-c with fracture risk, both LDL-c and HDL-c were divided into quintiles of distribution with the 3rd quintile serving as the reference for Cox models (Table 3; model 2). The lowest and highest quintiles of HDL-c were positively associated with hip fracture risk (HR 1.48 [1.12, 1.97]; 1.32 [1.02, 1.71], respectively). The pattern was similar but of lower magnitude for LDL-c.
Figure 1:
Generalized additive models with splines for winsorized lipid levels in M2 in a single model. W = winsorized.
Table 3:
Cox Regression estimates for hip fracture risk [Model 2] for the three exposures simultaneously adjusted for each other’s association with hip fracture, per 1 standard deviation higher value. HDL-c and LDL-c values are presented in quintiles; triglyceride values are continuous. The 3rd quintile of HDL-c and LDL-c are the reference group.
| ALL | WOMEN | MEN | ||||
|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | |
| HDL-c | ||||||
| Q1 | 1.48 | 1.12,1.97 | 1.66 | 1.56,2.38 | 1.05 | 0.65,1.7 |
| Q2 | 1.02 | 0.77,1.34 | 1.17 | 0.84,1.61 | 0.70 | 0.42,1.17 |
| Q4 | 1.16 | 0.90,1.49 | 1.21 | 0.91,1.61 | 1.06 | 0.61,184 |
| Q5 | 1.32 | 1.02,1.71 | 1.41 | 1.05,1.87 | 0.67 | 0.3,1.48 |
| LDL-c | ||||||
| Q1 | 1.22 | 0.94,1.57 | 1.17 | 0.86,1.58 | 1.38 | 0.84,2.26 |
| Q2 | 1.03 | 0.80,1.33 | 0.90 | 0.67,1.20 | 1.61 | 0.96,2.69 |
| Q4 | 1.12 | 0.87,1.44 | 1.06 | 0.8,1.41 | 1.37 | 0.79,2.36 |
| Q5 | 1.14 | 0.89,1.45 | 1.10 | 0.83,1.45 | 1.21 | 0.68,2.18 |
| Triglycerides | 0.92 | 0.81,1.04 | 0.88 | 0.76,1.02 | 1.03 | 0.81,1.31 |
Quintile values:
HDL-c (mg/dl): 20th, 41; 40th, 48; 60th, 56; 80th, 66.
LDL-c (mg/dl): 20th 101; 40th 120; 60th 137; 80th 158.
Lipoproteins:
Lipoprotein particle concentrations and sizes are shown in Supplementary Table 2. The top of Table 4 shows fully adjusted hazard ratios (Model 2) for hip fracture risk for 1 standard deviation higher concentrations and sizes of the individual lipoprotein particles for each individual analyte. HDL-P concentration was associated with lower risk (HR per SD 0.80 [0.66. 0.98], while one standard deviation higher HDL-P size was associated with an increased risk (HR 1.31 [1.06, 1.63]). In paired analyses (middle Table 4), higher VLDL particle size had a borderline increased risk of fracture, while higher HDL size had a statistically significant association. When all six variables were included in a single Cox model (bottom Table 4), one standard deviation higher VLDL-P concentration and size and higher HDL-P sizes were all associated with statistically significant higher risk of fracture. Figure 2 illustrates these associations in generalized additive models.
Table 4:
Cox regression models for the association of a one standard deviation increase of three lipoprotein particle concentrations and particle size, individually, in pairs, and in one model, with hip fracture risk in the Cardiovascular Health Study. Bolded numbers are statistically significant at the p <0.05 level. Values are adjusted for variables in model 2, Table 1.
| ALL | WOMEN | MEN | ||||
|---|---|---|---|---|---|---|
| HR | 95% CI | HR | 95% CI | HR | 95% CI | |
| VLDL - P conc | 1.09 | 0.86, 1.38 | 1.18 | 0.93, 1.51 | 0.79 | 0.45, 1.38 |
| HDL – P conc | 0.80 | 0.66, 0.98 | 0.78 | 0.63, 0.98 | 0.84 | 0.54, 1.30 |
| LDL – P conc | 0.94 | 0.78, 1.13 | 1.00 | 0.81, 1.24 | 0.75 | 0.51, 1.13 |
| VLDL P size | 1.17 | 0.97, 1.42 | 1.19 | 0.95, 1.49 | 1.14 | 0.85, 1.52 |
| HDL P size | 1.31 | 1.06, 1.63 | 1.20 | 0.94, 1.54 | 1.85 | 1.20, 2.86 |
| LDL P size | 1.04 | 0.85, 1.28 | 0.95 | 0.73, 1.22 | 1.35 | 0.97, 1.88 |
| VLDL – P size | 1.22 | 1.00, 1.48 | 1.27 | 1.00, 1.60 | 1.08 | 0.80, 1.47 |
| LDL – P conc | 0.94 | 0.72, 1.21 | 0.93 | 0.67, 1.29 | 0.88 | 0.56, 1.39 |
| LDL – P size | 1.00 | 0.75, 1.32 | 0.89 | 0.61, 1.32 | 1.26 | 0.88, 1.80 |
| HDL – P conc | 0.83 | 0.67, 1.02 | 0.80 | 0.63, 1.00 | 0.92 | 0.60, 1.42 |
| HDL – P size | 1.27 | 1.02, 1.58 | 1.16 | 0.92, 1.48 | 1.82 | 1.14, 2.93 |
| VLDL - P conc | 1.47 | 1.13, 1.91 | 1.57 | 1.19, 2.09 | 1.31 | 0.75, 2.29 |
| VLDL – P size | 1.24 | 1.05, 1.46 | 1.28 | 1.06, 1.54 | 1.21 | 0.91, 1.60 |
| LDL - P conc | 0.96 | 0.74, 1.24 | 0.96 | 0.71, 1.30 | 1.00 | 0.63, 1.59 |
| LDL - P size | 0.85 | 0.62, 1.17 | 0.83 | 0.56, 1.24 | 0.91 | 0.55, 1.50 |
| HDL – P conc | 0.84 | 0.68, 1.05 | 0.81 | 0.64, 1.03 | 0.91 | 0.60, 1.39 |
| HDL – P size | 1.81 | 1.25, 2.62 | 1.82 | 1.23, 2.71 | 2.24 | 1.13, 4.45 |
Conc = concentration
Figure 2:
Generalized additive models with splines for winsorized lipoprotein particle levels and sizes in M2 in a single model.
VLDLP – VLDL particle concentration. VZ – VLDL particle size. LDLP – LDL particle concentration. LZ - LDL particle size. HDLP – HDL particle concentration. HZ – HDL particle size. W- winsorized.
DISCUSSION
In this cohort of older adults, there were two key findings. First, HDL-c and LDL-c levels were significantly associated with hip fracture risk; triglyceride levels were not. The associations of HDL-c and LDL-c with hip fracture were non-linear, with HDL-c and LDL-c levels of 48 – 56 mg/dl and 120 – 137 mg/dl, respectively, associated with the lowest risk. Second, we observed novel associations of lipoproteins with hip fracture risk, including positive associations of HDL-P size and VLDL-P concentration and size.
The question arises how to understand these findings. The CHS is an epidemiological study; it does not contain mechanistic data. Hence, explanations are perforce speculative. Nonetheless, several of our findings have precedence. First, the finding that low HDL-c levels are associated with elevated hip fracture risk is not surprising since reduced HDL-c levels are associated with adverse health effects, including low bone mineral density (23). However, our finding that elevated HDL-P concentrations and HDL-P size are associated with increased hip fracture risk appear to be incongruent with their known health benefits. Reports from Denmark and the United Kingdom suggest that the association of HDL-c levels and mortality is U-shaped, with both high and low concentrations associated with elevated all-cause mortality (24, 25). In a meta-analysis of 12 studies of lipid levels in association with osteopenia and osteoporosis, HDL-c levels were elevated in people with osteoporosis in cross-sectional studies (26). Another earlier meta-analysis reported similar findings (27). Potential factors that may explain the mechanism by which HDL-c and HDL-P could impact bone physiology include adipokines, genetic factors, inflammation, and regulators of lipoprotein metabolism (28, 29).
Second, as to the association of LDL-c levels with hip fracture risk, a Mendelian randomization study demonstrated a significant negative association between LDL-c and bone mineral density (30). Elevated LDL-c levels are associated with atherosclerosis, a potent risk factor for osteoporosis (3). LDL-c, when oxidized, in associated with inflammation factors (31). The association of low LDL-c and hip fracture risk could reflect the decline in LDL-c levels with advanced aging, although our results persisted with adjustment for frailty. The significant non-linear association of LDL-c levels with hip fracture was generally lesser in magnitude than the association of HDL-c with hip fracture risk and could therefore also represent the play of chance. Alternatively, low LDL-c levels in older adults could be related to the presence of chronic illnesses which can increase fracture risk.
Finally, regarding our VLDL-P findings, VLDL-P consist mostly of triglycerides. The availability of triglycerides is the primary determinant of the rate of VLDL synthesis (32); VLDL production also requires intact hepatic function. Elevated triglyceride levels are associated with high fat diets, insulin resistance, obesity, hepatic steatosis, metabolic syndrome, and inflammation, all of which negatively affect the bone environment and osteoblast function (33, 34). Two clinical studies have reported increased risk of fracture in women with elevated triglyceride levels (35, 36), though no such associations were found here. In one study (35) the average age was much younger than in CHS; in the other (36) only vertebral fractures were examined.
This study has several strengths. First, we adjusted the associations of lipids and lipoprotein fractions with hip fracture risk for the effects of the associations of other lipid and lipoprotein subfractions with hip fracture risk. Such an approach has not been done previously and may explain in part why our results differ from other studies. We believe our approach represents an accurate representation of the associations of lipids and lipoproteins with hip fracture risk. Second, we measured lipids in two complementary ways, including standard clinical values and more novel NMR-based lipoproteins; our results suggest that the latter may be useful for understanding the relationships of lipids with fracture risk. Third, we focused on hip fractures, which are reliably ascertained with data sources like those available in CHS. We documented nearly 1,000 hip fractures, providing for precision and the ability to control simultaneously for covariates. Fourth, all the women were menopausal, reducing variability in lipid levels and hip fracture associated with perimenopause. Last, the potentially complex effect of statin medication use was limited in this cohort since much of the follow up time was in the era prior to its widespread use.
There are also important limitations in this study. Not all participants underwent lipoprotein measurement, although we were able to re-weight our results to reflect the larger cohort. Nonetheless, confidence intervals for some lipoprotein measurements were wide. As in all observational studies, unmeasured factors that were not captured could have influenced our results. Although we included adjustment for many covariates, the progressive clinical and subclinical comorbidity that accompanies older age could have influenced our results. CHS included few Asian or Latino participants, and we cannot generalize our results to these ethnicities nor to younger adults. Causal factors for fractures are not available in CHS. Finally, only a subset of CHS had bone density testing. This was done 6–7 years after the baseline examination, far removed from the time of baseline lipid collection.
In conclusion, we observed positive associations of HDL-c and LDL-c levels, and of VLDL-P number and size, and HDL-P size with hip fracture risk. These findings should be confirmed, and mechanistic studies are needed to understand them. These findings highlight the value of detailed phenotyping to understand the physiological determinants of bone health in older adults. They suggest that a possible unforeseen benefit of LDL-c lowering in those with high LDL-c levels may be fewer incident hip fractures.
Supplementary Material
GRANT SUPPORT
This research was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, 75N92021D00006, and grants U01HL080295 and U01HL130114 from the National Heart, Lung, and Blood Institute (NHLBI), with additional contribution from the National Institute of Neurological Disorders and Stroke (NINDS). Additional support was provided by R01AG023629 from the National Institute on Aging (NIA). A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
None of the authors declares a financial or intellectual conflict of interest.
All the authors have read and concur with the final version of the paper.
REFERENCES
- 1.Luegmayr E, Glantschnig H, Wesolowski GA, Gentile MA, Fisher JE, Rodan GA, Reszka AA. Osteoclast formation, survival and morphology are highly dependent on exogenous cholesterol/lipoproteins. Cell Death Differ. 2004;11(Suppl 1): S108–118. [DOI] [PubMed] [Google Scholar]
- 2.Chung JY, Hong SH. Vitamin D status and its association with cardiometabolic risk factors in Korean adults based on a 2008–2010 Korean National Health and Nutrition Examination Survey. Nutr Res Pract. 2013; 7:495–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kawai M, de Paula FJ, Rosen CJ. New insights into osteoporosis: the bone-fat connection. J Intern Med. 2012; 272:317–329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Zhao Q, Shen H, Su KJ, Zhang JG, Tian Q, Zhao LJ, Qiu C, Zhang Q, Garrett TJ, Liu J, Deng HW. Metabolomic profiles associated with bone mineral density in US Caucasian women Nutr Metab (Lond). 2018. Aug 10; 15:57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.El Maghraoui A, Rezqi A, El Mrahi S, Sadni S, Ghozlani I, Mounach A. Osteoporosis, vertebral fractures and metabolic syndrome in postmenopausal women. BMC Endocr Disord. 2014; 14:93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Nielson CM, Srikanth P, Orwoll ES. Obesity and fracture in men and women: an epidemiologic perspective. J Bone Miner Res. 2012; 27:1–10 [DOI] [PubMed] [Google Scholar]
- 7.Heath H, Melton LJ, Chu CP. Diabetes mellitus and risk of skeletal fracture. N Engl J Med. 1981; 304:115–116. [DOI] [PubMed] [Google Scholar]
- 8.Sennerby U, Melhus H, Gedeborg R, et al. Cardiovascular diseases and risk of hip fracture. JAMA. 2009; 302:1666–1673. [DOI] [PubMed] [Google Scholar]
- 9.Barzilay J, Buzkova P, Cauley JA, Robbins JA, Fink HA, Mukamal KJ. The associations of subclinical atherosclerotic cardiovascular disease with hip fracture risk and bone mineral density in elderly adults; the Cardiovascular Health Study. Osteoporos Int. 2018; 29 (10): 2219–2230. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Campos-Staffico AM, Freitas WM, Carvalho LSF, Coelho-Filho OR, Nadruz W Jr, Oliveira RB, Sposito AC; Brasilia Study on Healthy Aging and Brasilia Heart Study. Lower bone mass is associated with subclinical atherosclerosis, endothelial dysfunction and carotid thickness in the very elderly. Atherosclerosis. 2020; 292:70–74. [DOI] [PubMed] [Google Scholar]
- 11.Wang Z, Li Y, Zhou F, Piao Z, Hao J. Effects of Statins on Bone Mineral Density and Fracture Risk: A PRISMA-compliant Systematic Review and Meta-Analysis. Medicine (Baltimore). 2016. May;95(22): e3042. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Niemeier A, Niedzielska D, Secer R, Schilling A, Merkel M, Enrich C, Rensen PCN, Heeren J. Uptake of postprandial lipoproteins into bone in vivo: impact on osteoblast function. Bone. 2008;43(2):230–237. [DOI] [PubMed] [Google Scholar]
- 13.Alekos NS, Moorer MC, Riddle RC. Dual effects of lipid metabolism on osteoblast function (2020). Front Endocrinol; 11: 578194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Tintut Y, Demer LL. Effects of bioactive lipids and lipoproteins on bone. Trends Endocrinol Metabol 2014; 25 (2): 53–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Tian L, Yu X Lipid metabolism disorders and bone dysfunction--interrelated and mutually regulated (review). Mol Med Rep. 2015;12(1):783–794. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Ghorabi S, Shab-Bidar S, Sadeghi O, Nasiri M, Khatibi SR, Djafarian K. Lipid Profile and Risk of Bone Fracture: A Systematic Review and Meta-Analysis of Observational Studies. Endocr Res. 2019;44(4):168–184. [DOI] [PubMed] [Google Scholar]
- 17.Fried LP, Borhani NO, Enright P, Furberg CD, Gardin JM, Kronmal RA, Kuller LH, Manolio TA, Mittelmark MB, Newman A, et al. The Cardiovascular Health Study: design and rationale. Ann Epidemiol. 1991; 1(3):263–276. [DOI] [PubMed] [Google Scholar]
- 18.Kuller L, Arnold A, Tracy R, Otvos J, Burke G, Psaty B, Siscovick D, Freedman DS, Kronmal R. Nuclear magnetic resonance spectroscopy of lipoproteins and risk of coronary heart disease in the cardiovascular health study. Arterioscler Thromb Vasc Biol. 2002;22(7):1175–1180. [DOI] [PubMed] [Google Scholar]
- 19.Kuller L, Borhani NO, Furberg G, Gardin JM, Manolio T, O’Leary D, Psaty B, Robbins J: Prevalence of subclinical atherosclerosis and cardiovascular disease and association with risk factors in the Cardiovascular Health Study. Am J Epidemiol 1994; 139:1164–1179. [DOI] [PubMed] [Google Scholar]
- 20.Cushman M, Cornell ES, Howard PR, Bovill EG, Tracy RP. Laboratory methods and quality assurance in the Cardiovascular Health Study. Clin Chem 1995; 41:264–270. [PubMed] [Google Scholar]
- 21.Fried LP, Tangen CM, Walston J, Newman AB, Hirsch C, Gottdiener J, Seeman T, Tracy R, Kop WJ, Burke G, McBurnie MA; Cardiovascular Health Study Collaborative Research Group. Frailty in older adults: evidence for a phenotype. Gerontol A Biol Sci Med Sci. 2001;56(3):M146–56. [DOI] [PubMed] [Google Scholar]
- 22.Proschan MA, Dodd LE. Re-randomization tests in clinical trials. Stat Med. 2019; 38(12):2292–2302. [DOI] [PubMed] [Google Scholar]
- 23.McFarlane SI, Muniyappa R, Shin JJ, Bahtiyar G, Sowers JR. Osteoporosis and cardiovascular disease: brittle bones and boned arteries, is there a link? Endocrine. 2004; 23(1): 1–10. [DOI] [PubMed] [Google Scholar]
- 24.Madsen CM, Varbo A, Nordestgaard BG. Extreme high high-density lipoprotein cholesterol is paradoxically associated with high mortality in men and women: two prospective cohort studies. Eur Heart J. 2017; 38(32): 2478–2486. [DOI] [PubMed] [Google Scholar]
- 25.Hamer M, O’Donovan G, Stamatakis E. High-Density Lipoprotein Cholesterol and Mortality: Too Much of a Good Thing? Arterioscler Thromb Vasc Biol. 2018;38(3):669–672. [DOI] [PubMed] [Google Scholar]
- 26.Zhao H, Li Y, Zhang M, Qi L, Tang Y. Blood lipid levels in patients with osteopenia and osteoporosis: a systematic review and meta-analysis. J Bone Miner Metab. 2021; 39(3):510–520. [DOI] [PubMed] [Google Scholar]
- 27.Chen YY, Wang WW, Yang L, Cheng WW, Zhang HX. Association between lipid profiles and osteoporosis in postmopausal women: a meta-analysis. Eur Rev Med Pharmacol Sci 2018; 22: 1–9. [DOI] [PubMed] [Google Scholar]
- 28.Papachristou NI, Blair HC, Kypreos KE, Papachristou DJ. High-density lipoprotein (HDL) metabolism and bone mass. J Endocrinol. 2017;233(2): R95–R107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ackert-Bicknell CL. HDL cholesterol and bone mineral density: is there a genetic link? Bone 2012; 50: 525–533 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Zheng J, Brion MJ, Kemp JP, Warrington NM, Borges MC, Hemani G, Richardson TG, Rasheed H, Qiao Z, Haycock P, Ala-Korpela M, Davey Smith G, Tobias JH, Evans DM. The Effect of Plasma Lipids and Lipid-Lowering Interventions on Bone Mineral Density: A Mendelian Randomization Study. J Bone Miner Res. 2020;35(7):1224–1235. [DOI] [PubMed] [Google Scholar]
- 31.Engelen SE, Robinson AJB, Zurke YX, Monaco C. Therapeutic strategies targeting inflammation and immunity in atherosclerosis: how to proceed? Nat Rev Cardiol. 2022. Jan 31:1–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Gotto AM. Interrelationship of triglycerides with lipoproteins and high-density lipoproteins. Am J Cardiol. 1990; 66(6):20A–23A. [DOI] [PubMed] [Google Scholar]
- 33.Ahmed LA, Schirmer H, Berntsen GK, Fonnebo V, Joakimsen RM. Features of the metabolic syndrome and the risk of non-vertebral fractures: the Tromso study. Osteoporos Int. 2006; 17:426–432. [DOI] [PubMed] [Google Scholar]
- 34.Lee SH, Baek S, Ahn SH, et al. Association between metabolic syndrome and incident fractures in Korean men: a 3-year follow-up observational study using national health insurance claims data. J Clin Endocrinol Metab. 2014; 99:1615–1622. [DOI] [PubMed] [Google Scholar]
- 35.Chang P-Y, Gold EB, Cauley JA, Johnson WO, Karvonen-Gutierrez C, Jackson EA, Ruppert KM, Lee JS. Triglyceride Levels and Fracture Risk in Midlife Women: Study of Women’s Health Across the Nation (SWAN). J Clin Endocrinol Metabol 2016; 101(9): 3297–3305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Johnell O, Kanis JA, Black DM, et al. Associations between baseline risk factors and vertebral fracture risk in the Multiple Outcomes of Raloxifene Evaluation (MORE) Study. J Bone Mineral Res. 2004; 19:764–772. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.