The American Journal of Medicine (2005) 118, 1414.e1-1414.e5 CLINICAL RESEARCH STUDY Lipid levels and bone mineral density Daniel H. Solomon, MD, MPH, a,b Jerry Avorn, MD, a Claire F. Canning, MA, a Philip S. Wang, MD, DrPH a,c a Division of Pharmacoepidemiology, b Division of Rheumatology, Department of Medicine, c Department of Psychiatry, Brigham and Women s Hospital, Harvard Medical School, Boston, Mass. KEYWORDS: Osteoporosis; Hyperlipidemia; Epidemiology ABSTRACT PURPOSE: There has been considerable debate about the potential relationship between the use of statin lipid-lowering drugs and fracture risk; several observational studies suggest a protective effect but no randomized controlled trials have confirmed such a benefit. Because statins are given preferentially to persons with hyperlipidemia, if lipid levels were associated with bone mineral density, this could explain the discrepancy between epidemiological observations and randomized controlled trials. The aim of this study was to examine the relationship between lipid levels and bone mineral density. SUBJECTS AND METHODS: We included the 13 592 participants in the National Health and Nutritional Examination Survey (NHANES) III who had bone mineral density and lipid levels measured; participants who reported the use of a lipid-lowering therapy were excluded. We examined the unadjusted bone mineral density across quintiles of total cholesterol, low-density lipoprotein (LDL), and high-density lipoprotein (HDL). We then constructed multivariable models, including age, sex, body mass index, and other potential confounders. RESULTS: In crude analyses, higher total cholesterol and LDL levels were associated with lower bone mineral densities (both P values for trend.001), whereas higher HDL levels were associated with higher bone mineral densities (P value for trend.001). However, in fully adjusted models, there was no significant relationship between total cholesterol, LDL, or HDL levels and bone mineral density (all P values for trend.1). CONCLUSIONS: These results do not support a relationship between lipid levels and bone mineral density. 2005 Elsevier Inc. All rights reserved. Old age is characterized by hardening of the arteries and softening of the bones. Anonymous Although many have theorized a connection between bone metabolism and atherosclerosis, the precise nature of this relationship remains unknown. Several receptors reside on both osteocytes and endovascular lining cells, and cytokines including osteoprotegrin, receptor activator of NF-?B (RANK) and receptor activator of NF-?B ligand (RANKL) appear to play Dr. Solomon receives salary support from NIH AR-48616, the Arthritis Foundation, and the Engalitcheff Arthritis Outcomes Initiative. Requests for reprints should be addressed to Daniel H. Solomon, MD, MPH, Division of Pharmacoepidemiology, Brigham and Women s Hospital, 1620 Tremont Street, Suite 3030, Boston, MA 02120. E-mail address: dhsolomon@partners.org. important roles in both systems. 1 At a clinical level, some have noted that low bone mineral density correlates with an increased risk of cardiovascular morbidity and mortality. 2,3 There has also been interest in whether medicines used to lower blood lipids, the HMG co-a reductase inhibitors (statins), may also increase bone mineral density. Statins, like bisphosphonates, inhibit steps in the mevalonate pathway and have been shown to increase bone mineral density when applied directly to the rat calvaria. 4 Several large observational epidemiologic studies suggest that persons taking statins are at a significantly reduced risk of fractures and have an increased bone mineral density, 5-9 but other epidemiologic studies have found no relationship. More importantly, randomized controlled clinical trials of lipid lowering for cardiovascular disease have not observed any reduction in fracture risk associated with statins. 10,11 0002-9343/$ -see front matter 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2005.07.031
1414.e2 The American Journal of Medicine, Vol 118, No 12, December 2005 One potential explanation of the discrepancy between epidemiologic studies and randomized controlled clinical trials is unobserved confounding. If an unmeasured factor, such as hyperlipidemia in observational datasets, is associated with patients exposure status (receipt of statins) and is also associated with the outcome of interest (fracture reduction), it may explain the observed relationship between statins and reduced fractures. Such a factor is known as a confounder. 12 Such a confounder in this case would be associated with the receipt of statins and would also be related to reduced fractures, perhaps through an improved bone mineral density, a reduced risk of falling, or some other factor. One recent study suggested that hyperlipidemia might be such a factor. 13 Among 236 community-based women, there appeared to be a significant and direct correlation between elevation in low-density lipoprotein (LDL) and bone mineral density at the hip. We assessed the relationship between lipid parameters and bone mineral density in the National Health and Nutritional Examination Survey (NHANES) III. Methods Study population NHANES III was conducted between 1988 and 1994 and enrolled men, women, and children from the United States. Participants were sampled so as to reflect the distribution of age, sex, and racial/ethnic mix of the United States. From the sample of patients aged over 17 years who were available for the home interview (n 20 050), we required that persons undergo a bone mineral density test and have a laboratory examination for lipids. From the remaining 13 909 participants, we excluded those who were current users of lipidlowering therapy at the time of their interview (n 317) for a final sample size of 13 592. In secondary analyses, we included this group. Data collection As previously described, bone mineral density of the hip was estimated from the total proximal femoral region using dual-energy x-ray absorptiometry. Hologic QDR 1000 densitometer was used with appropriate quality control measures. 14 Lipid levels and serum 25 (OH) vitamin D were measured after an overnight fast in most participants using previously described methods. 15,16 The 25 (OH) vitamin D assay used a two-step radio-immunoassay procedure (INCSTAR, Stillwater Minn). 17 Physical examinations were conducted using a structured format. Height and weight were assessed allowing for calculation of a body mass index. Survey information included standardized questionnaires regarding physical activity, diet, and medical history. Medication and supplement information was also collected by asking participants to bring in all currently used products. The identity of each and duration of Table 1 Distribution of patient characteristics by quintile of total cholesterol Quintile of total cholesterol (mg/dl) 168 169-190 191-211 212-239 240 n 2723 2651 2664 2850 2704 Age, mean (SD) 39 (18) 43 (18) 48 (18) 52 (18) 57 (17) Female sex 1368 (50) 1328 (50) 1321 (50) 1443 (51) 1518 (56) Postmenopausal status* 218 (16) 390 (29) 577 (44) 858 (59) 1168 (77) White 1746 (64) 1707 (64) 1859 (70) 2066 (72) 1976 (73) BMI, mean (SD) (kg/m 2 ) 26 (6) 27 (6) 27 (6) 28 (6) 28 (5) Prior hip or wrist fracture 197 (7) 195 (7) 221 (8) 233 (8) 257 (9) Alcohol use (drinks/month), mean (SD) 30 (47) 34 (56) 34 (68) 32 (49) 34 (53) Current smoker 840 (31) 761 (29) 673 (25) 701 (25) 603 (22) Leisure activity (METs/month), mean (SD) 93 (132) 89 (119) 82 (103) 79 (107) 76 (106) Limitations in IADLs 196 (12) 224 (14) 274 (16) 355 (19) 405 (21) Fair or poor self-reported health 573 (21) 571 (22) 637 (24) 734 (26) 752 (28) 25 (OH) vitamin D (ng/ml), mean (SD) 25 (11) 25 (11) 26 (11) 26 (11) 26 (11) Calcium intake (mg), mean (SD) 925 (264) 929 (257) 934 (249) 940 (238) 941 (236) Thiazide diuretics 94 (3) 110 (4) 170 (6) 215 (8) 320 (12) Oral glucocorticoids 17 (0.6) 24 (0.9) 27 (1) 34 (1) 41 (2) Hormone replacement therapy* 24 (11) 46 (12) 101 (18) 96 (11) 143 (12)
Solomon et al Bone mineral density and lipids 1414.e3 Table 2 Distribution of patient characteristics by quintile of low density lipoprotein Quintile of low density lipoprotein (mg/dl) 11-95 96-115 116-135 136-155 156 n 2719 2613 2789 2194 2720 Age, mean (SD) 40 (18) 44 (19) 48 (18) 52 (18) 56 (17) Female sex 1431 (53) 1393 (53) 1416 (51) 1081 (49) 1423 (52) Postmenopausal status* 326 (23) 451 (32) 629 (44) 624 (58) 1040 (73) Caucasian 1695 (62) 1757 (67) 1943 (70) 1604 (73) 1915 (70) BMI, mean (SD) (kg/m 2 ) 25 (6) 27 (6) 27 (6) 28 (5) 28 (5) Prior hip or wrist fracture 198 (7) 215 (8) 224 (8) 183 (8) 240 (9) Alcohol use (drinks/month), mean (SD) 38 (69) 32 (51) 31 (49) 31 (45) 30 (51) Current smoker 883 (32) 690 (26) 707 (25) 493 (22) 651 (24) Leisure activity (METs/month), mean (SD) 95 (129) 88 (119) 80 (107) 80 (108) 77 (105) Limitations in IADLs 222 (13) 245 (15) 281 (16) 252 (17) 378 (20) Fair or poor self-reported health 608 (22) 573 (22) 658 (24) 549 (25) 714 (26) 25 (OH) vitamin D (ng/ml), mean (SD) 25 (11) 26 (11) 26 (11) 26 (11) 26 (11) Calcium intake (mg), mean (SD) 925 (263) 935 (247) 932 (253) 943 (231) 933 (250) Thiazide diuretics 122 (4) 130 (5) 169 (6) 163 (7) 278 (10) Oral glucocorticoids 20 (0.7) 27 (1) 32 (1) 23 (1) 37 (1) Hormone replacement therapy* 65 (20) 77 (17) 80 (13) 74 (12) 95 (9) Table 3 Distribution of patient characteristics by quintile of high density lipoprotein Quintile of high density lipoprotein (mg/dl) 8-38 39-45 46-52 53-62 63 n 2796 2731 2671 2783 2611 Age, mean (SD) 49 (18) 48 (19) 46 (19) 47 (19) 49 (19) Female sex 886 (32) 1165 (43) 1436 (54) 1678 (60) 1813 (69) Postmenopausal status* 407 (46) 524 (45) 618 (43) 729 (43) 933 (51) Caucasian 2121 (76) 1994 (73) 1804 (68) 1851 (67) 1584 (61) BMI, mean (SD) (kg/m 2 ) 29 (5) 28 (5) 27 (6) 26 (5) 25 (5) Prior hip or wrist fracture 231 (8) 227 (8) 200 (7) 214 (8) 231 (9) Alcohol use (drinks/month), mean (SD) 29 (49) 30 (51) 28 (42) 32 (45) 43 (76) Current smoker 859 (31) 714 (26) 668 (25) 669 (24) 668 (26) Leisure activity (METs/month), mean (SD) 77 (106) 80 (111) 84 (115) 85 (117) 94 (119) Limitations in IADLs 309 (17) 281 (16) 271 (16) 282 (16) 311 (18) Fair or poor self-reported health 791 (28) 666 (24) 614 (24) 614 (22) 582 (22) 25 (OH) vitamin D (ng/ml), mean (SD) 26 (10) 26 (11) 26 (11) 26 (11) 26 (12) Calcium intake (mg), mean (SD) 928 (258) 935 (247) 939 (239) 934 (248) 932 (252) Thiazide diuretics 204 (7) 190 (7) 174 (7) 161 (6) 180 (7) Oral glucocorticoids 20 (0.7) 25 (0.9) 28 (1) 30 (1) 40 (2) Hormone replacement therapy* 18 (4) 36 (7) 67 (11) 103 (14) 186 (20)
1414.e4 The American Journal of Medicine, Vol 118, No 12, December 2005 Table 4 Total hip bone mineral density by quintile of lipid parameter in NHANES III Quintiles of total cholesterol (mg/dl) 169 170-189 190-212 213-239 240 BMD (g / cm 2 ) Unadjusted 0.97 0.96 0.94 0.91 0.98 Age, sex, and BMI 0.96 0.95 0.95 0.95 0.95 Fully adjusted* 0.97 0.98 0.97 0.98 0.97 Quintiles of low-density lipoprotein (mg/dl) 11-95 96-115 116-135 136-155 156 Unadjusted 0.97 0.96 0.94 0.92 0.97 Age, sex, and BMI 0.95 0.95 0.95 0.95 0.96 Fully adjusted* 0.97 0.98 0.98 0.98 0.97 Quintiles of high-density lipoprotein (mg/dl) 8-38 39-45 46-52 53-62 63 Unadjusted 0.99 0.97 0.96 0.94 0.90 Age, sex, and BMI 0.95 0.95 0.95 0.96 0.96 Fully adjusted* 0.97 0.97 0.99 0.98 0.97 BMI body mass index; BMD bone mineral density. *Fully adjusted models include menopausal status, race, prior fracture, thiazide use, smoking status, HRT use, limitation in activities of daily living, poor self-reported health, alcohol consumption, 25 (OH) vitamin D levels, calcium consumption, glucocorticoid use, and physical activity. The P value for the test of trend across quintiles of total cholesterol.9. The P value for the test of trend across quintiles of low density lipoprotein.9. The P value for the test of trend across quintiles of high density lipoprotein.15. therapy were noted. Total calcium intake was calculated by combining the supplemental and dietary calcium intake. Other covariates included the number of chronic medical conditions, tobacco use (current vs past or never), alcohol intake (number of drinks in the prior month), self-reported health (fair or poor vs good or excellent), history of hip or wrist fracture, history of fall (0, 1, 2 ), menopausal status, and any limitation in an instrumental activity 18 Leisuretime activity level was calculated as the number of metabolic equivalents expended in the past month. 19 Statistical analysis We first created quintiles of each lipid parameter based on the distribution of values. Covariates were assessed across these quintiles. Mean bone mineral density was calculated for each quintile and then adjusted for age, sex, and body mass index using the LSMEANS option in SAS (SAS Institute, Cary, NC). Other covariates were then tested one by one in models that included age, sex, and body mass index. A fully adjusted model including all covariates was then assessed. We conducted secondary analyses stratified on menopausal status and on whether a patient was using a lipid-lowering treatment. diuretic use, and current smoking varied in a predictable fashion across quintiles of each of the lipid parameters. However, several other covariates were distributed in an unexpected manner: prior fractures increased in frequency with higher total cholesterol and LDL levels but did not vary much across HDL levels, and glucocorticoid and hormone replacement therapy use increased across quintiles of all three parameters. Calcium intake and 25 (OH) vitamin D levels did not vary across quintiles of lipid parameter. The remaining covariates functional status, exercise level, and alcohol use did not vary in any pattern. In unadjusted analyses, bone mineral density varied across all lipid quintiles: decreasing across quintiles of total cholesterol and LDL and increasing across HDL quintiles (Table 4). However, after adjusting for age, sex, and body mass index, these trends were attenuated and there were no clinically meaningful or statistically significant differences in bone mineral density across quintiles. Fully adjusted models also showed no significant relationship between lipid parameters and bone mineral density. In separate regression models, we examined for potential effect modification by menopausal status and the use of lipid-lowering therapy. Neither of these stratified models produced results different from the main analysis. Results The distribution of covariates ranged substantially across lipid level quintiles (Tables 1-3). Age, body mass index, thiazide Discussion We assessed the relationship between lipid levels and bone mineral density among 13 592 participants in the NHANES III
Solomon et al Bone mineral density and lipids 1414.e5 survey who were not taking a lipid-lowering treatment. Unadjusted analyses suggested that higher total cholesterol and LDL levels were associated with lower bone mineral density, and lower HDL levels were associated with lower bone mineral density. Yet, models that considered all relevant covariates measured in NHANES yielded no significant relationship between any of the lipid measures and bone mineral density. This topic is of interest for two reasons. First, prior studies have suggested that cardiovascular morbidity and mortality is inversely related to bone mineral density, but this relationship has not been well defined. Second, several but not all epidemiologic studies suggest that statin use is associated with both elevated bone mineral density and a reduced risk of fractures. Because randomized controlled trials have found no relationship between statins and fractures, one possible explanation for the epidemiologic studies findings is the presence of an unmeasured confounder, such as hyperlipidemia. If hyperlipidemia were such a confounder, it would be associated with higher bone mineral density. However, our null finding suggests that it is unlikely that a relationship between hyperlipidemia and bone mineral density is the missing link between atherosclerosis and osteoporosis. Therefore, it is also unlikely that hyperlipidemia is the unmeasured confounder that explains the discrepancy between observational studies and randomized controlled clinical trials. One prior study of this relationship found that lipids and bone mineral density were related. 13 However, this study included only several hundred patients at one referral center, and important covariates such as functional status, thiazide use, alcohol intake, and tobacco were not included in multivariable models. Exclusion of such variables may have produced a finding distorted by confounding. Our findings should be viewed in light of several methodologic limitations. We used data from NHANES, a large crosssectional study conducted in the United States. A more powerful study design would include longitudinal data to assess changes over time in lipids and bone mineral density. However, without any suggestion of a cross-sectional association in our data, a longitudinal relationship seems unlikely. The discrepancy between our findings and those from an earlier study conducted in Europe suggests that the potential relationship between lipids and bone mineral density could be modified by genetic heritage. In conclusion, we found no relationship between lipid parameters and bone mineral density. Lipid measurements and bone mineral density are only two variables associated with atherosclerosis and osteoporosis. The relationship between other parameters warrants investigation. As well, the observed relationship between statin use and fracture reduction requires further examination with other clinical and epidemiologic studies. References 1. Hofbauer LC, Schoppet M. Clinical implications of the osteoprotegrin/ RANKL/ RANK system for bone and vascular diseases. JAMA. 2004; 292:490-495. 2. Samelson EJ, Kiel DP, Broe KE, et al. Metacarpal cortical area and risk of coronary heart disease: the Framingham Study. Am J Epidemiol. 2004;159:589-595. 3. Varosy PD, Shlipak, Vittinghoff E, et al. Fracture and risk of coronary events in women with heart disease. Am J Med. 2003; 115:196-202. 4. Mundy G, Garrett R, Harris S, et al. Stimulation of bone formation in vitro and in rodents by statins. Science. 1999;286:1946-1949. 5. Wang PS, Solomon DH, Mogun H, Avorn J. HMG-CoA reductase inhibitors and the risk of hip fractures in elderly patients. JAMA. 2000;283:3211-3216. 6. Chan KA, Andrade SE, Boles M, et al. Inhibitors of hydroxymethylglutaryl-coenzyme A reductase and risk of fracture among older women. Lancet. 2000;355:2185-2188. 7. Meier CR, Schlienger RG, Kraenzlin ME, Schlegel B, Jick H. HMG- CoA reductase inhibitors and the risk of fractures. JAMA. 2000;283: 3205-3210. 8. Edwards CJ, Hart DJ, Spector TD. Oral statins and increased bonemineral density in postmenopausal women. Lancet. 2000;355:2218-2219. 9. Schoofs MW, Sturkeboom MC, van der Klift M, Hofman A, Pols HA, Stricker BH. HMG-coA reductase inhibitors and the risk of vertebral fractures. J Bone Miner Res. 2004;19:1525-1530. 10. Reid IR, Hague W, Emberson J, et al. Effect of pravastatin on frequency of fracture in the LIPID study: secondary analysis of a randomised controlled trial. Lancet. 2001;357:509-512. 11. Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of antioxidant vitamin supplementation in 20 536 high-risk individuals: a randomised placebo-controlled trial. Lancet. 2002;360: 23-33. 12. Rothman KJ, Greenland S. Modern Epidemiology, 2nd edition. Philadelphia, Pennsylvania: Lippincott Williams and Wilkins; 1998. 13. Adami S, Braga V, Zamboni M, et al. Relationship between lipids and bone mass in 2 cohorts of healthy women and men. Osteoporos Int. 2004;74:136-142. 14. Wahner HW, Looker A, Dunn WL, Walters LC, Hauser MF, Novak C. Quality control of bone densitometry in a national health survey (NHANES III) using three mobile examination centers. J Bone Miner Res. 1994;9:951-960. 15. Johnson CL, Rifkind BM, Sempos CT, et al. Declining serum total cholesterol levels among US adults: the National Health and Nutrition Examination Surveys. JAMA. 1993;269:3002-3008. 16. Nesby-O Dell S, Scanlon KS, Cogswell ME, et al. Hypovitaminosis D prevalence and determinants among African American and white women of reproductive age: third National Health and Nutrition Examination Survey, 1988 1994. Am J Clin Nutr. 2002;76:187-192. 17. Gunter EW, Lewis BG, Koncikowski SM. Laboratory procedures used for the Third National Health and Nutrition Examination Survey (NHANES III), 1988-1994. Available at: http://www.cdc.gov/nchs/ data/nhanes/nhanes3/cdrom/nchs/manuals/labman.pdf. Accessed October 3, 2005. 18. Spector WD. Functional disability scales. In: Spilker B, ed. Quality of Life Assessment for Clinical Trials. New York, New York: Raven Press; 1990. 19. Nelson KM, Reiber PG, Boyko MEJ. Diet and exercise among adults with type 2 diabetes. Diabetes Care. 2002;25:1722-1728.