To study effects of statins on human bone, 82 postmenopausal women were randomized to 1-year treatment with simvastatin 40 mg/day or placebo. The study showed no effect of simvastatin on biochemical bone markers or on BMD at the hip or spine. Thus, our results do not support a general beneficial effect of simvastatin on bone.
Introduction: Statins have been reported to cause bone anabolic as well as antiresorptive effects, and therefore statins have been suggested as potential agents in treatment of osteoporosis.
Materials and Methods: In a double-blinded design, 82 healthy postmenopausal women with osteopenia were randomized to 1-year simvastatin treatment 40 mg/day or placebo. BMD and plasma levels of cholesterol, parathyroid hormone (PTH), and biochemical bone markers were measured at baseline, after 1 year of treatment (week 52), and 26 weeks after withdrawal of treatment (week 78). Calcium supplements (400 mg/day) were administrated during the entire 1.5-year study period.
Results: Seventy-eight women completed the 1-year treatment. After 1 year, simvastatin but not placebo caused reduced plasma cholesterol (−27% versus +1%, p < 0.001) and low-density lipoprotein (LDL) levels (−43% versus +1%, p < 0.001). After withdrawal of treatment, cholesterol and LDL levels returned to baseline levels and no longer differed from the placebo group. However, plasma levels of PTH and biochemical bone markers did not differ between groups at week 52 or 78. Compared with placebo, simvastatin caused no changes in BMD at the lumbar spine, total hip, femoral neck, or whole body at week 52 or 78. However, a significant increase in BMD was found in response to simvastatin at the forearm. Within the simvastatin group, changes in cholesterol levels did not correlate to BMD changes at any site.
Conclusions: Our results do not support a general beneficial effect of simvastatin on bone.
Osteoporosis is a systemic skeletal disease characterized by a low bone mass and microarchitectural deterioration of bone tissue, with a consequent increased bone fragility and susceptibility to fracture.(1) Most drugs currently approved for osteoporosis treatment mainly work by an antiresorptive mechanism of action; that is, they slow down bone turnover, resulting in a reduced bone loss and a relatively reduced fracture risk. However, in severe osteoporosis, drugs that are capable to rebuild bone are needed.
Recent studies have suggested that 3-hydroxy-3methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins), in addition to their cholesterol-lowering effect, may exert antiresorptive and bone anabolic effects. In vitro studies with statins have shown osteoclast apoptosis and enhanced osteoblastic synthesis of bone morphogenetic protein 2 (BMP-2), a growth factor that causes osteoblastic proliferation.(2) In rodents, orally administered statins have been shown to increase bone volume in some,(2,3) but not all, studies.(4) Similarly, in humans, treatment with statins has been associated with an increased BMD(5–7) and a decreased fracture risk,(8–11) although other investigators have been unable to show such effects of statins on BMD and fracture risk.(11–17) In general, all these studies have been nonrandomized, and the subjects studied have been treated with statins because of hypercholesterolemic disorders. Thus, the effects of statins in non-hypercholesterolemic subjects with low BMD remain to be studied.
The aim of this study was to investigate the effect of 1 year of statin treatment in women with low bone mass in a randomized controlled design.
MATERIALS AND METHODS
Eighty-two healthy white women were included in this double-blinded, randomized, placebo-controlled trial. To be included, participants were required to be more than 12 months postmenopause, to be <76 years of age, to be healthy as assessed by a standard biochemical screening program (Table 1), and to have osteopenia at the lumbar spine or total hip, that is, a BMD 1 SD below peak bone mass (T-score < −1).
Table Table 1.. Baseline Characteristics by Study Group (Mean ± SEM)
Exclusion criteria included hysterectomized women who were <55 years of age, and those who had previous or present malignant disease, hyper- or hypothyroidism within the last 2 years, granulomatous disease, impaired renal (plasma creatinine > 120 μM) or hepatic (plasma alanine aminotransferase [ALAT] > 80 U/liter) function, drug or alcohol abuse of more than 14 U of alcohol a week within the last 2 years, diuretic treatment within the last 6 months, or treatment within the last 2 years with systemic gluco/mineralocorticoids, anticonvulsant, bisphosphonates, estrogen, raloxifene, and/or statins. In addition, as a safety measure, subjects were excluded if they had low plasma cholesterol levels (total cholesterol < 4.0 mM and/or low-density lipoprotein [LDL] < 2.5 mM) or if they received treatment with drugs with known adverse drugs interactions with statins (e.g., fibrates, nicotinic acid, bile acid-binding resins, erythromycin, or oral anticoagulants). None of the studied subjects had known hyperlipidemia before study start.
Recruitment of studied subjects
Participants were recruited from the general population. Through direct mailings, we invited 2138 randomly selected women to participate in the study (Fig. 1). Subjects who responded that they were willing to participate were asked to fill in a short questionnaire concerning menopausal status and medical conditions. According to their answers, 612 women were excluded, because they did not fulfill the participation criteria mentioned above. The remaining 219 women were screened at our outpatient clinic, and 82 women were included in the study (Fig. 1).
The study was carried out in accordance with the Declaration of Helsinki II. It was approved by the regional Ethical Committee (Aarhus County 2000/0223) and the Danish National Board of Health. Each individual gave verbal and written informed consent before the study. The Good Clinical Practice (GCP) Unit at the University Hospital of Aarhus, Denmark, monitored the study.
Study treatment and procedures
Participants were randomly and double-blindedly allocated to 1-year (52 ± 4 weeks) of treatment with either simvastatin (Zocor; Merck, Sharp & Dohme) 40 mg/day (n = 41) or placebo (n = 41). In addition, all participants received 400 mg/day of elementary calcium (Unikalk; Pharma-Vinci A/S, Frederiksvaerk, Denmark) during the entire 1.5-year study period. Study medication and calcium were administrated at bedtime. Calcium and simvastatin tablets were obtained commercially. Because it was infeasible to obtain placebo tablets similar to Zocor tablets, the Central Pharmacy at the Country of Copenhagen, Herlev, Denmark, manufactured placebo tablets for the study. To obtain double-blinding of treatment, both Zocor and placebo tablets were embedded in a hard gelatin capsule (Capsugel; Warner-Lambert, Bornem, Belgium), causing a similar appearance of active medicine and placebo.(18) Encapsulation of tablets and randomization were performed by The Pharmacy Department at Aarhus University Hospital, Denmark. Furthermore, to ensure double blinding of the trial, the investigator responsible for the clinic visits (LR) was unaware of the results of laboratory measurements (e.g., plasma cholesterol levels) during the trial period. As a safety precaution, liver and muscle function was measured (ALAT and creatine phosphokinase) at each clinic visit.
The primary efficacy endpoint was the percent change at week 52 from baseline in BMD at the lumbar spine and total hip. Secondary efficacy criteria were the changes in BMD at the femoral neck, forearm, and whole body as well as changes in biochemical indices of calcium homeostasis and bone metabolism.
Clinic visits and measurements
After randomization, clinic visits (at 7:30-10:30 a.m.) occurred at months 1, 6, 12, and 18. At all clinic visits, adverse events were recorded by questioning the patient in a general, nondirective manner, and the number of capsules returned by the patient was counted to assess compliance.
At baseline, after 1 year of treatment and 6 months after withdrawal of treatment (week 78), BMC (g), BMD (g/cm2), and projected bone area (cm2) were measured with DXA at the lumbar spine (L1-L4), hip, forearm, and whole body using a Hologic QDR 2000 densitometer (software version V4.55; Hologic, Waltham MA, USA). Additionally, body composition (BMC, fat mass, and lean tissue mass) was determined employing the same scanner. The CV for the actual QDR scanner was 1.5% for lumbar spine BMD, and long-term stability has been shown to be high, with changes of <0.2%/year.
Blood samples were drawn between 7:00 and 10:30 a.m. after an overnight fast. For urine samples, studied subjects were advised to empty their bladders early in the morning. A sample given within 2 h later was used for urine analysis.
Plasma levels of calcium, creatinine, albumin, thyrotropin (thyroid-stimulating hormone [TSH]), ALAT, creatine kinase (CK), total cholesterol, LDLs and high-density lipoproteins (HDLs), and triglycerides, as well as urinary calcium and creatinine, were determined by standard laboratory methods. Total plasma calcium was corrected for individual variations in albumin according to the following formula: adjusted plasma calcium (mM) = plasma calciumtotal (mM) − 0.00086 × (650 − plasma albumin [μM]).
Plasma intact parathyroid hormone (PTH) was measured by an electrochemiluminescence immunoassay using an Elecsys analyzer (Roche Diagnostics GmbH, Mannheim, Germany). The total CV was <6%.
Plasma osteocalcin and C-terminal telopeptide of type I collagen (β-CrossLaps, CTx) were measured by ELISA using an automated instrument (Elecsys, 2010 immunoassay analyzer; Roche Diagnostics). Antibodies that recognize both the intact (1-49) and the amino-terminal fragment (N-Mid-osteocalcin; 1-43) were used.(19) For both osteocalcin and CTx, the total CV was <6%. Alkaline phosphatase (ALP) and bone-specific ALP (bone-ALP) were measured spectrophotometrically using an automated instrument (Hitachi 917; Roche Diagnostics). Bone-ALP was measured after lectin precipitation (Boehringer Mannheim). The total CV was <8%. Because of a shortage in analytic kits for bone-ALP measurements (the manufacturer has stopped producing them), bone-ALP was only measured at baseline and at weeks 52 and 78. To reduce analytical variation, PTH and bone markers from each patient were analyzed in the same run. Blood samples were kept frozen at −80°C until the time of assay.
Prestudy sample size calculations
Calculations on sample size revealed that 27 participants were required in each group to detect a minimum between-group difference of 2.5% in BMD at the lumbar spine (5% significance level, 80% statistical power). Based on these calculations and to allow for withdrawal, the intention was to recruit 40 subjects per treatment arm (80 subjects in total). In addition, if a drop-out occurred within 4 weeks after randomization, the subject was substituted with a new participant.
At baseline, differences between study groups were assessed using χ2 tests for categorical variables and a two-sample t-test or Mann-Whitney U-test for continuous variables as appropriate.
Data were analyzed using an intention-to-treat approach, comprising all randomized subjects with at least one postbaseline measurement. For participants who withdrew before the end of study, the last obtained measurement before withdrawal was carried forward. Serial changes were studied using ANOVA for repeated measurements, with treatment group as the independent variable (effect of time by group). Assumptions for repeated measures ANOVA were checked by Mauchly's test of sphericity, and accordingly, adjustment in the degrees of freedom was made (Huynh-Feldt ϵ). In case of a significant between-group difference by repeated measures ANOVA, differences between groups were analyzed at each time point of measurements by a posterior analysis using a two-sample test. BMD was adjusted for differences in body weight using a general linear regression model.
All results are given as mean ± SEM unless otherwise stated. Statistical analysis was performed using Statistical Package for Social Sciences (SPSS 8.0) for Windows.
Women included in the study had a median age of 64 years (range, 53-74 years) and had been postmenopausal for a median of 19 years (range, 5-44 years). Baseline characteristics were not statistically different between groups, except that women randomized to placebo were heavier (by chance) and therefore had a higher body mass index (BMI) than subjects randomized to simvastatin treatment (Table 1). Measurements of body composition showed that the weight differences were caused by a higher fat mass in the placebo group than in the simvastatin group (Table 2).
Table Table 2.. BMD, Body Composition, and Body Weight at Baseline, Week 52, and Week 78 (Mean ± SEM)
In the simvastatin group as well as in the placebo group, 39 women completed the 1-year treatment (Fig. 1). The rate of compliance for those who completed the 1-year treatment, as measured by pill count, was a median of 98% (range, 55-103%) in the placebo group and 97% (range, 76-100%) in the simvastatin group (p = 0.26). A high degree of compliance was further supported by changes in plasma cholesterol levels (Fig. 2). In the simvastatin group at week 52, total cholesterol decreased by 27% (p < 0.001 compared with the placebo group) and LDL-cholesterol decreased by 43% (p < 0.001). After withdrawal of treatment, plasma cholesterol returned to baseline levels in the simvastatin group and no longer differed from the placebo group at week 78 (Fig. 2).
Osteodensitometry and body composition
Table 2 shows baseline BMD values as well as changes in BMD by study group. No statistically significant differences were found between the placebo and the simvastatin group on primary efficacy endpoint. At week 52, BMD at the lumbar spine was increased by 1.1 ± 0.5% in the placebo group and by 0.5 ± 0.5% in the simvastatin group (p = 0.46). At the total hip, BMD decreased by 0.2 ± 0.4% in the placebo group and increased by 0.2 ± 0.4% in the simvastatin group during the 52 weeks of treatment (p = 0.52). Posthoc power calculations showed that the study had a 92% statistical power to detect a 2.5% difference and a 77% power to detect a 2.0% difference in BMD at the lumbar spine. Similarly, at the total hip, the statistical power was 98% and 92% to detect a difference in BMD of 2.5% or 2.0%, respectively.
Similar to the results at week 52, BMD did not differ between groups at week 78 at the lumbar spine or total hip. Nor did whole body BMD differ between groups at week 52 or 6 months after withdrawal of treatment (Table 2). Similarly, no significant between-group differences were found in bone area or BMC at week 52 or 78 at the lumbar spine, hip region, or whole body (data not shown). However, forearm BMD was increased at weeks 52 and 78 in the simvastatin group compared with the placebo group (Table 2). Total forearm BMD was increased by 1.1% in the simvastatin group at week 52 (versus −0.3% in the placebo group, p = 0.01) and by 1.8% at week 78 (versus +0.4% in the placebo group, p = 0.02). Ultradistal forearm BMD was increased by 2.9% in the simvastatin group at week 52 (versus +0.1% in the placebo group, p = 0.02) and by 3.4% at week 78 (versus +1.1% in the placebo group, p = 0.04).
At weeks 52 and 78, changes from baseline in fat mass, lean tissue mass, and body weight did not differ significantly between groups (Table 2).
Bivariate correlation analysis showed no significant correlations between changes in BMD (at any measurement site) and the concomitant changes in total- and LDL-cholesterol levels at week 52.
Correlations between body weight and BMD
Because baseline body weight differed significantly between groups, we studied whether BMD differed between groups after weight adjustment. However, after correction for body weight, baseline BMD did not differ significantly between the placebo and simvastatin group at the lumbar spine (0.819 versus 0.822 g/cm2; p = 0.88), total hip (0.774 versus 0.778 g/cm2; p = 0.85), total forearm (0.424 versus 0.430 g/cm2; p = 0.63), or whole body (0.933 versus 0.945 g/cm2; p = 0.50).
In addition, using weight-adjusted BMD instead of raw BMD data, changes from baseline at weeks 52 and 78 did not differ between the placebo and the simvastatin group at the lumbar spine, total hip, or whole body. Changes from baseline in weight-adjusted BMD differed between groups at week 52 (p = 0.02) and week 78 (p = 0.002) only at the forearm, with a larger increase in the simvastatin group (week 52: +0.8%; week 78: +1.5%) than in the placebo group (week 52: +0.1%; week 78: +0.7%).
Although plasma PTH levels and biochemical markers of bone formation (osteocalcin and bone-ALP) and bone resorption (CTx) tended to be higher in the simvastatin than in the placebo group, changes from baseline did not differ significantly between groups (Table 3). Similarly, plasma total calcium, albumin-adjusted calcium, and urinary calcium-creatinine ratio did not differ between groups at week 52 or 78 (data nor shown). Moreover, regression analyses showed no effects of scale body weight on changes in biochemical indices in the two study groups.
Table Table 3.. Plasma Levels of PTH and Biochemical Markers of Bone Turnover (Mean ± SEM)
The number of adverse events was comparable between groups. Only one serious adverse event was recorded; in the placebo group, one participant underwent surgery after she had sustained a distal radius fracture.
In a randomized, controlled trial in postmenopausal women with low BMD, we assessed the effect of simvastatin on BMD and bone turnover. Despite a high compliance to study medication, we found no effect of 1 year of simvastatin treatment on BMD at the lumbar spine, total hip, femoral neck, or whole body. Our poststudy power calculations showed a good statistical power to detect a 2.5% difference between the placebo and the simvastatin group in BMD at the lumbar spine and total hip. Moreover, simvastatin did not affect bone turnover as assessed by plasma levels of biochemical bone markers. Only at the forearm did BMD differ between groups at weeks 52 and 78, with a larger increase in the simvastatin group than in the placebo group. However, changes in forearm BMD did not correlate with changes in cholesterol levels.
In the Scandinavian Simvastatin Survival Study (4S), simvastatin treatment produced changes in plasma total cholesterol, LDL, and HDL levels of −25%, −35%, and +8%, respectively.(20) In the present study, plasma total cholesterol, LDL, and HDL levels changed by −25%, −43%, and +1%, respectively. Thus, because our findings of changes in cholesterol levels are very similar to the findings in the 4S study, it is unlikely that the encapsulation of simvastatin tablets into a hard gelatin capsule affected the bioavailability of simvastatin.
Previous studies on the effects of statins on BMD have revealed conflicting results. As in our findings, several nonrandomized studies have been unable to show an effect of statins on BMD in hypercholesterolemic patients.(11,13,14,17) Most recently, in an analysis from the Women's Health Initiative (WHI) Observational Study, BMD at the lumbar spine, total hip, or whole body did not differ between 422 statin users and 6020 nonusers, after adjustment for multiple confounders.(17) However, in several other studies, treatment with statins has been associated with an increased BMD.(5–7) In a population-based cohort of 1003 women living in the United Kingdom, Edwards et al.(5) identified 41 women who had been treated with a statin for a median of 48 months before assessment of their BMD. Compared with 100 age- and sex-matched controls, statin users had a significantly higher BMD at the spine and at the femoral neck, even after adjustment for age, height, weight, hormone replacement therapy (HRT) use, and smoking status. Similarly, in a retrospective review of medical records, Chung et al.(6) reported an increased BMD in the hip region of 14 male, but not female, patients in response to statin treatment. Finally, comparing 30 postmenopausal hypercholesterolemic women treated for 1 year with simvastatin 40 mg/day with 30 non-hypercholesterolemic (i.e., untreated) postmenopausal women, Montagnani et al.(7) found simvastatin treatment to be associated with a significantly increased BMD at the lumbar spine and femoral neck. The main difference between these three studies showing a positive effect of statin treatment on BMD and our study is that we included women with low BMD who did not have known hypercholesterolemia in a randomized study design. Thus, we cannot exclude that statins may affect bone differently in hypercholesterolemic and in non-hypercholesterolemic individuals. However, the most likely explanation for our findings is the randomized design.
Randomization procedures may be especially important when studying preventive medicine. In epidemiological research, the “healthy volunteer effect” is a well-known selection bias, that is, volunteers for epidemiological research have lower mortality rates than nonvolunteers.(21,22) Similarly, a “healthy drug user effect” may apply to users of preventive medicine.(23) Therefore, effects observed in nonrandomized statin trials may merely be because of factors associated with “lipid lowering treatment in general” than a specific effect of statin treatment per se. Similarly, the finding by some investigators of a reduced fracture risk in statin-treated patients could be caused by selection biases, (i.e., the result of a “healthy drug user effect”).(8–11)
Despite our findings, statin drugs may exert effects on bone cells. In vitro studies on bone cells have suggested that statins may inhibit bone-resorbing cells and concomitantly stimulate bone-forming cells.(2) A possible explanation for the lack of an effect on bone metabolism in our study despite the in vitro findings of a direct effect of statins on bone cells is a low in vivo drug concentration in bone tissue. In vivo, statins are recycled in the enterohepatic circulation. Thereby, the lipid-lowering effect of statins is high, whereas their effective concentration in extrahepatic tissues is low. The findings from in vitro studies may encourages experimental studies in which statin drugs are administrated in a manner that bypasses the liver (e.g., administration through dermal patches.)
Apparently, an antiresorptive effect has occurred in studied subjects with a decreased bone turnover as biochemical markers of bone turnover decreased. Concomitantly, plasma PTH levels increased. The increased PTH levels may be caused by the decreased bone turnover, causing a reduced calcium release form bone with a subsequent increase in PTH. These changes occurred as an effect of time, with no differences between groups. The reason for these changes is not obvious. One explanation could be that studied subjects have changed their lifestyle because at the start of the study they were told that they had a low BMD. Some may have stopped smoking, etc. Unfortunately, we did not measure variables covering lifestyle at the end of study. A similar mechanism may explain that BMD in the placebo group remained unchanged during study.
Our study showed no effect of simvastatin treatment on BMD at the lumbar spine, hip, or whole body. However, at the forearm, BMD increased in response to treatment. There are no obvious explanations for this differential effect of simvastatin on BMD at different skeletal sites. However, it is well known that BMD at different skeletal sites may respond differently to various pathological conditions. For example, in primary hyperparathyroidism, most patients have reduced radial and preserved vertebral BMD.(24) Thus, further studies are needed to assess whether specific effects of simvastatin apply to the bone metabolism of the forearm.
By chance, women randomized to the placebo group had a higher body weight (because of a higher fat mass) than the women in the simvastatin group. Normally, body weight correlates positively to BMD. However, we observed no baseline differences in BMD despite differences in body weight, and BMD adjusted for differences in body weight did not differ between groups. Most likely, an effect of body weight on BMD would have been apparent if sample size had been larger. However, our aim was to be able to detect a 2.5% difference in BMD at the lumbar spine. At the time of initiation of our study (May 2000), published studies had suggested that statins might exert bone anabolic as well as antiresorptive effects.(2) In addition, in a cross-sectional study, treatment with statins has been associated with an ∼7% higher BMD at the spine and hip.(5) Thus, before the start of study, we thought that a 2.5% increase in BMD would be reasonable to expect in response to 1 year of treatment with a relatively high dose of simvastatin. Poststudy power calculations showed that our statistical power to detect a 2.5% difference in BMD at the spine was even better than planned. Nevertheless, our results do not exclude an effect of statins. A larger sample size and/or a longer duration of treatment may be needed to detect such effects.
In a randomized controlled design, 1 year of simvastatin treatment did not affect BMD at the lumbar spine, femoral neck, or whole body. Only at the forearm did BMD increase in response to simvastatin treatment. However, simvastatin caused no changes in bone turnover, as measured by biochemical markers of bone formation and resorption. Thus, within 1 year of treatment, our results do not support a general beneficial effect of orally administrated simvastatin on bone metabolism.
The technical assistance of Donna Lund, Lisbeth Flyvbjerg, Charlotte Beck Gylling, Britta Malm, and Tove Stenum is greatly appreciated. The Department of Experimental Clinical Research and Centre for Clinical Pharmacology is acknowledged for financial support. Finally, we thank AM Skjoedtgaard for helping us prepare the study medication.