Statins and Osteoporosis: A Clinical Review

Michael J. Gonyeau, Pharm.D.


Pharmacotherapy. 2005;25(2):228-243. 

In This Article

Clinical Outcome Trials

A number of studies documenting the potential role of statins with clinical outcomes (i.e., fractures) have been conducted. These trials may be of more use to the practicing clinician when reviewing this topic as compared with the preceding surrogate marker studies. A summary of clinical trials available at the time of preparing this review is presented in Table 2 .[50,51,52,53,54,55,56,57,58,59,60]

In a post hoc analysis of 9014 patients taking pravastatin 40 mg or placebo from the 6-year Long-Term Intervention with Pravastatin in Ischemic Disease (LIPID) trial, researchers evaluated the effects of pravastatin on the frequency of fractures, calculated from International Classification of Diseases, Ninth Revision (ICD-9) codes, based on reports of adverse events that resulted in hospitalization during the study.[50] Separate analyses were conducted for women (17% of cohort), and survival curves were reported for the entire population. The researchers found no difference in fracture rates in the pravastatin versus placebo groups in the entire cohort (3.9% vs 4.1%, OR 0.94, 95% CI 0.77-1.16, p=0.58) or in the sub-group of women (6.1% vs 7.8%, OR 0.8, 95% CI 0.55-1.18, p=0.26). Patients aged 65 years or older were also independently assessed, with similar results (4.3% vs 4.6%, OR 0.93, 95% CI 0.68-1.27, p=0.65). Survival curves over the study period were similar (p=0.58). The researchers concluded that pravastatin was not associated with a significant reduction in fracture risk. They also point out limitations to their study, such as the fact that this cohort was not at high risk for fracture because patients were recruited based on cardiovascular disease and hyperlipidemia criteria. It is unknown what percentage of patients studied had documented osteoporosis. Although the researchers did separate analyses for women and the elderly, it would have been interesting to see a separate analysis of obese patients, as these data were captured in the original trial, and obesity is thought to be a potential confounder. In addition, concomitant use of other drugs that may affect osteoporosis or bone metabolism were not accounted for.

A post hoc analysis[51] was performed of 4444 patients taking simvastatin 20-40 mg or placebo for 5.5 years in the Scandinavian Simvastatin Survival Study (4S).[52] Patients were asked to report any adverse experiences at each clinical visit (at least every 6 mo), and these data were the basis for the analysis. One hundred fifty-five subjects were identified with fractures during the study. No statistically significant differences were observed between the placebo and simvastatin groups in regard to hip or femur fractures (0.45% vs 0.45%), vertebral fractures (0.36% vs 0.14%), or all fracture types combined (3.19% vs 3.78%). Specific p values were not presented. Although this analysis adds to our body of evidence, there were limitations. Like the previously mentioned post hoc analysis,[50] a small percentage of patients were women (19% of the 4S population), and the primary end point was all-cause mortality. Fracture data were collected through patient self-reporting, and whether any reports were confirmed by hospital or medical record data is unclear. In addition, osteoporotic fractures were not distinguished from any other fracture type, such as trauma.

Another post hoc case-control study used data derived from the Geelong Osteoporosis Study and included 1375 women (573 with incident fractures and 802 without) to determine any association among statin therapy, BMD, and fracture rates during a 2-year period.[53] All fracture cases were identified through radiology reports. Questionnaires were sent to all participants to collect data on use of statins, other drugs, lifestyle (such as alcohol use, exercise, and smoking status), and diet. Duration of statin use was recorded, but specific statins and dosages were not. Bone mineral density scanning was performed of the femoral neck in 1354 patients, spine in 1373, and whole body in 1311. Sixteen patients taking statins were identified in the fracture group and 53 in the nonfracture group.

No differences were noted in fracture sites among the groups. Baseline characteristics were similar between those who took statins and those who did not. Likewise, patients with fractures and those without fractures had similar baseline characteristics, except for patients taking HRT (fracture 10% vs nonfracture 15%, p=0.005), those with glucocorticoid exposure (fracture 8% vs nonfracture 3%, p=0.001), and those taking calcium and vitamin D (fracture 23% vs nonfracture 17%, p=0.008). The BMD scores were similar in the statin and nonstatin groups after adjusting for age, weight, and fractures (p>0.05 for all comparisons). However, statistically significant reductions in fractures at all sites were observed in those taking statins, which remained after adjustment for age, weight, BMD, dietary calcium intake, alcohol consumption, smoking status, physical activity, HRT use, glucocorticoid exposure, and calcium and vitamin D use (whole body, OR 0.43, 95% CI 0.24-0.78; spine, OR 0.42, 95% CI 0.24-0.75; femoral neck, OR 0.45, 95% CI 0.25-0.8). Although this study did an admirable job of adjusting for several confounders, its main limitation is the small sample of patients taking statins because it limits our ability to extrapolate the data to larger populations.

In an observational case-control study, the authors used automated pharmacy and claims data from six health maintenance organizations between October 1994 and September 1997.[54] Participants included women aged 60 years or older who had continuous pharmacy benefits for 2 or more years before September 1996. The primary end point was to ascertain any relation between statin use and fracture risk among older women. Case patients had a documented fracture of the hip, humerus, distal tibia, vertebrae, or wrist during the time evaluated. Women were excluded from the analysis if their data indicated a diagnostic code for trauma or any cancer, or if they had claims data regarding HRT, bisphosphonates, calcitonin, any anti-convulsant, or thyroid hormones.

Automated pharmacy dispensing records determined the number of statin dispensings in the 2 years before the fracture date. These data were used as the proxy measure for statin exposure to assess any potential relationship between statins and fractures. Exposures to nonstatin lipid-lowering agents were also evaluated. Attempts to control for concomitant drugs that may confound results, such as use of antipsychotics, long-acting hypnotic or antidepressant drugs within 30 days of fracture, thiazide diuretics, hypoglycemic agents, and systemic steroids, were evaluated.

Case patients (928 patients) and control subjects (2747 subjects) were well matched with regard to baseline characteristics. Most fractures were of the hip (28.2%). After adjustment for age, hospital admission, and use of confounding drugs, the 13 or more dispensings of a statin was associated with a 48% reduction in fracture risk (OR 0.52, 95% CI 0.29-0.91). Use of nonstatin lipid-lowering agents was not associated with a significant protective effect at any dispensing level (≥ 13 dispensings, OR 0.90, 95% CI 0.30-2.66).

Although the authors did decrease the recall bias in their study by obtaining data from automated pharmacy records, other potential confounders were not able to be addressed. No information was available to ascertain the percentage of obese individuals in the trial. This is important, as obesity is associated with a reduced risk for osteoporosis and fractures, and obesity may be positively correlated with statin use. Other osteoporotic risk factors, such as smoking history and exercise patterns or other lifestyle differences in study participants were not evaluated. Although attempts were made to control for concomitant use of drugs affecting fracture risk, over-the-counter drugs, such as calcium and vitamin D, were not appraised. Also, the authors did not ascertain dosages of statins used, which may help in determining a dose-response relationship.

Much like the previous study, another group performed a retrospective case-control study of 6110 patients aged 65 years or older who were enrolled in the New Jersey Medicaid program from January 1991-July 1995.[55] Case patients were defined as any patient hospitalized for hip fracture based on claims data. Four age- and sex-matched control subjects per case subject were selected randomly from the study population. Case patients and control subjects were then compared based on prescription records for statin use (fluvastatin, lovastatin, pravastatin, or simvastatin; dosages not recorded) 180 days and 3 years before the index fracture date. The primary end point was fracture rate in patients taking statins versus those without treatment.

Statin use in the previous 180 days was associated with a 50% reduction in hip fracture (OR 0.5, 95% CI 0.33-0.76), and statin use in the previous 3 years was associated with a 43% reduction in hip fracture (OR 0.57, 95% CI 0.40-0.82). These results remained statistically significant after adjustment for potential confounders including comorbidities, use of drugs that may affect osteoporosis, extent of health care utilization, age, sex, race, and insurance status. Patients taking nonstatin antihyperlipidemic agents did not show a statistically significant reduction in the frequency of hip fracture in either period assessed. One limitation of this study was the lack of body mass index (BMI) comparison, which may bias the results in favor of the statin group, as statin users tend to have higher BMIs, and obesity and increased BMI are associated with reduced risk for fractures. Another interesting finding in this trial was the lack of benefit observed with estrogen in this population (OR 0.92, 95% CI 0.55-1.54), although the extent and duration of use were not reported.

In another study from a large database, investigators conducted a retrospective cohort study of Tennessee Medicaid users by comparing hip fracture risk in patients who took statins with that in patients taking other lipid-lowering agents during a 10-year period.[56] Inclusion criteria consisted of patients with statin or other lipid-lowering agent prescriptions, no occurrence of clinically treated osteoporosis or osteoporotic fractures for 1 year preceding study initiation, and aged older than 50 years without severe illness, disability, or dementia. Identified candidates were then matched in a 1:1 fashion to patients not receiving any lipid-lowering drugs by sex, birth year, and study initiation date.

The investigators identified 12,506 patients taking statins and 4798 taking other lipid-lowering agents; these patients were matched to 17,280 control subjects.

No statistically significant differences were observed in baseline characteristics. The frequency of hip fractures was significantly reduced in the statin group (adjusted incidence rate ratio [RR] 0.62, 95% CI 0.45-0.85) and in the group taking other lipid-lowering agents (RR 0.44, 95% CI 0.26-0.72) compared with controls. The reduction in hip fractures between patients taking statins and those taking other lipid-lowering agents was not statistically significant (RR 1.42, 95% CI 0.83-2.43), although a trend was observed that favored the nonstatin group. This study did not control for some very important potential confounders, including BMI and smoking. The number of fractures also limits the results, as only 17 fractures were reported in the nonstatin cohort. Individual statin agents, their doses, and the duration of statin use were not presented.

Same Database, Different Results. Two large, retrospective case-control studies were conducted in the United Kingdom and used the same database, the General Practice Research Database (GPRD), yet each study yielded very different results.[57,58]

The first trial was a nested case-control analysis of 27,319 patients.[57] Three separate groups were analyzed: all patients aged 50-89 years who received at least one lipid-lowering agent prescription during the study, patients with hyperlipidemia receiving no treatment, and a random sample of patients without hyper-lipidemia and no history of any lipid-lowering drug therapy. Patients were excluded from the analysis if they had a diagnosis of osteoporosis, osteomalacia, cancer, or alcoholism, or if they were taking bisphosphonates. Patients were followed from the date of first lipid-lowering agent prescription until the development of a fracture, the patient left the practice, or the patient died. Potential confounders such as BMI, smoking status, use of HRT, and use of oral or inhaled corticosteroids were controlled for in the analysis. The primary end point of this trial was fractures with the use of statins or other lipid-lowering drugs in case patients versus control subjects.

Cases patients (3940 patients) were those who developed a fracture during the study period. Six controls (23,379 subjects) were matched to each case patient based on age, sex, general practice, and all aforementioned potential confounders. After adjustment for confounders, use of statins was associated with statistically significant reductions in fractures at all sites (current use, OR 0.55, 95% CI 0.44-0.69, p<0.001; 1-4 prescriptions, OR 0.51, 95% CI 0.32-0.81, p=0.004; 5-19 prescriptions, OR 0.62, 95% CI 0.45-0.85, p=0.003; ≥ 20 prescriptions, OR 0.52, 95% CI 0.36-0.76, p<0.001). Fibrates were associated with smaller risk reductions when compared with statins and did not reach statistical significance (current use, OR 0.87, 95% CI 0.70-1.08, p=0.19; 1-4 prescriptions, OR 0.78, 95% CI 0.44-1.39, p=0.39; 5-19 prescriptions, OR 0.79, 95% CI 0.55-1.14, p=0.21; ≥ 20 prescriptions, OR 0.95, 95% CI 0.71-1.26, p=0.71), nor did use of any other lipid-lowering agents.

An interesting finding was that a 49% reduction in fracture risk was observed in patients with only 1-4 statin prescriptions, correlating to 1-4 months of therapy. This would suggest a rapid-acting mechanism resulting in dynamic effects in a short period of time. The authors reference the work of another group[33] who found significant increases in bone formation in as little as 5 weeks in a rat model. Similar effects in such a short time in humans would seem improbable but warrant further study.

Limitations of this study, beyond its retrospective design, included lack of information about physical activity of subjects and their socioeconomic status, which may result in underdiagnosis of fractures because of lower utilization of health care resources. Also, statistically significant differences were noted in some baseline characteristics, particularly HRT use, use of oral or inhaled corticosteroids, number of practitioner visits, and BMI, although the authors attempted to control for all of these potential confounders. Furthermore, selection bias may have been introduced into the study by the random pooling of patients with which to select the control population. Also, it is surprising to note that apparently hip fracture was not specified as a fracture type when defining cases in the study, nor were specific hip fracture rates reported in the results.

Using the same GPRD database, a second group of authors selected another group of patients for their analysis based on altered selection criteria ( Table 3 ).[58] These authors used data from all 683 practice sites in the GPRD database, whereas the previously mentioned group of authors[57] limited their population to 340 general practice sites. This second group of authors studied patients from 1987-1999 and identified cases based on ICD-9 codes for fracture. Case patients were then matched with control subjects by age, sex, and medical practice. Current statin use was defined as a statin prescription within 6 months of the index fracture date, and past users were defined as those with one or more statin prescriptions but who had stopped therapy 6 or more months before the index fracture date. Case patients (81,880 patients) were matched to 81,880 control subjects. Their BMIs and use of nonsteroidal antiinflammatory drugs, thiazide diuretics, systemic corticosteroids, and HRT were recorded. The primary end point was risk of fracture in patients currently taking statins versus the risk in those who did not take statins.

The population was 76% women, with an average age of 70 years. Statins were not associated with a statistically significant reduction in fracture risk regardless of past use (OR 1.01, 95% CI 0.78-1.32) or current use (OR 1.01, 95% CI 0.88-1.16). Also, no correlation was observed between time since statin use and fracture risk. Current use of statins showed a trend in hip fracture reduction (OR 0.59, 95% CI 0.31-1.13). No differences were observed between statins and nonstatins with regard to fracture risk (OR 1.02, 95% CI 0.83-1.24). The authors concluded that use of statins was not associated with a reduction in overall fracture risk. Although the ORs were adjusted for potential confounders including smoking, BMI, and use of drugs such as HRT, thiazide diuretics, and systemic corticosteroids, the authors failed to account for patients with a history of osteo-porosis, cancer, or alcoholism, or for patients taking drugs used to treat osteoporosis.

Few prospective trials evaluating the effect of statins on fractures were found. One group[59] performed a prospective observational study of 93,716 postmenopausal women aged 50-79 years from the Women's Health Initiative Observational Study.[61] The primary end point was the rate of hip, lower arm or wrist, and other (including clinical vertebral) fractures among those taking statins (1846 patients) and those not taking statins (85,870 patients) over 3 years. Any differences in BMD were also compared in a subset of 6442 women (422 statin users, 6020 non-statin users). Patients were assessed at baseline and with annual follow-up question-naires to observe any changes in status. Potential confounders assessed included concomitant drugs (thiazide diuretics, alendronate, cortico-steroids, sedative or hypnotics, and HRT), intake of calcium and vitamin D, history of fracture, smoking status, coffee consumption, alcohol consumption, BMI, and physical activity.

Baseline characteristics differed among those taking statins and those who did not with respect to age distribution, obesity (31.3% vs 24.7%), HRT use (40.4% vs 46%), thiazide use (8.4% vs 5.2%), alendronate use (3.6% vs 2.2%), and physical function scores greater than 90 (27.4% vs 38.9%). Overall, statin use was not associated with reductions in any fracture after controlling for potential confounders (hip fracture, multivariate-adjusted hazard ratio [HR] 1.22, 95% CI 0.83-1.81; lower arm or wrist, HR 1.04, 95% CI 0.85-1.27). Likewise, neither duration of statin use nor statin potency affected fracture rates. Statins were actually associated with an increased frequency of "other" clinical fractures (HR 1.11, 95% CI 1.00-1.22). Total hip BMD was slightly increased in those taking statins (0.86 vs 0.84 g/cm2, p=0.0193) after confounder adjustment.

This trial had many strengths including its prospective design, large sample, and completeness of data on potential confounders. It is also important to point out that this trial compared both BMD and fracture rates in the same cohort. The trial is limited, however, by the relatively short duration of statin use, as only 24% of patients were taking statins for more than 3 years. Also, lack of data on past statin use and dosages of statins somewhat detracts from the results. Self-reporting of nonhip clinical fractures is potentially a problem in the trial, as is the absence of spine radiographs.

The other large prospective trial with a prespecified fracture outcome was the Heart Protection Study.[61] All 20,536 subjects in this trial were evaluated for hospitalizations due to fracture in an attempt to discern any protective effects of simvastatin 40 mg. Trauma-related fractures were excluded from the analysis. Baseline characteristics were similar in the statin (10,269 patients) and nonstatin (10,267 patients) groups. No significant differences were noted between the treatment groups in number of patients with any fracture (simvastatin, 241 patients [2.3%] vs nonstatin, 230 patients [2.2%]), or in the number of patients with osteoporotic-related fractures of the hip, wrist, or spine (simvastatin, 109 patients [1.1%] vs nonstatin, 91 patients [0.9%]).

Strengths of this study include its prospective randomized design, the prespecified fracture end points, exclusion of trauma-related fractures, large sample, and long duration. However, specific analyses relating to statins and fracture risk were not presented, including attempts to control for potential confounders, such as use of other agents that would affect fracture rates, BMI, and physical activity. Whether radiographic confirmation of fractures was routinely adjudicated is unclear. Also, the trial possibly was not powered to detect a difference in fracture rates during the time period, as this was not the primary end point. Furthermore, although the study had many thousands of patients, only 25% of the patients were female.

In a recent meta-analysis of osteoporosis studies and a cumulative meta-analysis of selected observational and controlled trials,[62] statins were associated with a trend in lowering hip fractures (relative hazards 0.19-0.62) and in nonspine fractures (relative hazards 0.49-0.95) in the four prospective trials evaluated.[63,64,65,66] After adjustment for age, BMI, physical activity, smoking status, and use of bisphosphonates and estrogen, analysis of the observational studies revealed a 57% reduction in hip fracture (OR 0.43, 95% CI 0.25-0.75) and a 31% reduction in nonspine fractures (OR 0.69, 95% CI 0.75-0.88). Two clinical trials mentioned previously[50,51] were also pooled and analyzed, but results showed no statistically significant differences in hip fractures (OR 0.87, 95% CI 0.48-1.58) or nonspine fractures (OR 1.02, 95% CI 0.83-1.26).

These data add to our growing body of evidence of a potential link between statins and fracture reduction. However, only baseline drug use was examined in each of the four prospective trials evaluated, and the sample overall was fairly small. Although many potential confounders were controlled for in the analyses, other agents such as thiazide diuretics, corticosteroids, calcium and vitamin D, calcitonin, thyroid hormone, and raloxifene were not. Several other trials reviewed in this article were published since this meta-analysis was conducted and may add to our knowledge of this topic.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.