Use of Statins and Risk of Fractures FREE

Synthesis of cholesterol is reduced by inhibition of 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase in the liver, which catalyzes conversion of HMG-CoA to mevalonic acid. Statins have been shown to be reversible, competitive inhibitors of this enzyme and, thus, are currently used for clinical treatment of hypercholesterolemia.^1- 2 Mevalonic acid has been found to be a precursor not only of cholesterol but also of proteins such as geranylgeranyl pyrophosphate, which is important in the control of osteoclast-mediated bone resorption. Nitrogen-containing bisphosphonates, such as alendronate and risedronate, may exert their antiresorptive action by suppressing the generation of geranylgeranyl pyrophosphate in the mevalonate pathway.³ Furthermore, recent animal and in vitro research has led to a hypothesis that statins may reduce risk of fractures.⁴ Some epidemiological studies have reported reductions in fracture risk,^5- 8 but the preliminary results of a large follow-up study have not confirmed these observations.⁹ Given the widespread use of statins, the possible effects on fracture is an important clinical issue. We performed a population-based case-control study in which a history of statin use in fracture cases was compared with that among age- and sex-matched control patients.

In the United Kingdom, health care delivery is centered around general practitioners, who are responsible for primary health care and specialist referrals.¹⁰ The information in this study was obtained from the General Practice Research Database (GPRD), which contains computerized medical records of 683 general practices in the United Kingdom. Approximately 6.5% of the total registered population of England and Wales is represented in the database. The age and sex distribution of the patients enrolled is representative of the population of England and Wales. The data accrued in the GPRD include demographic information about patients, prescription details, clinical events, preventive care, referrals to specialist care, and hospital admissions and their major outcomes.^10- 15 Clinical data are stored and retrieved by means of Oxford Medical Information Systems codes for diseases or causes of morbidity and mortality that are cross-referenced to the International Classification of Diseases, Ninth Revision (ICD-9). Each entry into the GPRD is internally validated by cross-checking within the practice and by comparison with external statistics.^10- 15 Only data from practices that pass this quality control check are compiled to form the GPRD database. Several independent validation studies have shown that the GPRD database has a high level of completeness and validity.^16- 20 The GPRD is owned by the UK Department of Health and managed by the Medicines Control Agency. This study was conducted in all practices that have contributed data to the GPRD.

Within the GPRD, a case-control study was conducted. Cases were permanently registered patients (ie, with residence in the practice neighborhood) aged 50 years or older who had a fracture of the vertebrae, clavicle, humerus, radius/ulna, carpus, hip, ankle, or foot recorded in their medical records between the enrollment date of their practice in GPRD and the end of data collection (July 1999). Data collection for the GPRD began in 1987. The classification of fractures was based on ICD-9 categories. Previous studies of GPRD data reported a high level of validity of the GPRD with respect to fractures (>90% of fractures were confirmed).^19- 20

Control patients were adults without a history of any type of fracture in their medical records who were matched to case patients by age (within 1 year), sex, and medical practice. If no control patient was found, the age criterion was expanded consecutively at 1-year intervals to a maximum of 10 years. If no eligible control patient within 10 years of age was identified, an age- and sex-matched control patient was selected from another practice. The index date of each control patient was that of their matched case patient (ie, first fracture after enrollment in GPRD). For control patients who had transferred to another practice or died prior to this date, an index date between registration and transfer dates was randomly selected. About 80% of cases were matched by practice, sex, calendar year of the index date, and within 1 year of age.

Exposure to statins was determined by reviewing prescription information from prior to the index date. Current statin users were patients who had received a prescription for a statin in the 6 months before the index date. Past users were patients who had received 1 or more statin prescriptions but who had stopped treatment more than 6 months before the index date. A similar method was used to determine exposure to nonstatin lipid-lowering drugs (ie, bezafibrate, ciprofibrate, clofibrate, fenofibrate, gemfibrozil, cholestyramine, colestipol hydrochloride, acipimox, nicotinic acid, or omega-3 marine triglycerides). Time since first statin use was determined by calculating the period between the index date and first prescription. The daily dosage of statins was obtained from the written dosage instructions for the last statin prescription prior to the index date and the strength of the tablets. Simvastatin, atorvastatin, pravastatin, and fluvastatin were considered dose-equivalent, while the equivalent dose of cerivastatin was considered to be 100 times lower. Cumulative statin dose was defined as the total grams of drugs prescribed (ie, the sum of the strength of the tablets prescribed) at any time prior to the index date.

Adjusted odds ratios (ORs) of fractures in statin users compared with nonusers were estimated using conditional logistic regression. Models included current and past use of statins and nonstatin lipid-lowering drugs. We controlled the analysis for a wide range of clinical variables and medication use that have been associated with risk of fractures,^21- 22 including history of diabetes mellitus, rheumatoid arthritis, hyperthyroidism, congestive heart failure, seizures, anemia, dementia, depression, psychotic disorder, cerebrovascular accident, and chronic obstructive pulmonary disease. Prescriptions in the 6 months prior to the index date for anticonvulsants, nonsteroidal anti-inflammatory drugs, methotrexate, hormone replacement therapy, thiazide diuretics, anxiolytics/hypnotics, antipsychotics, antidepressants, anti-Parkinson drugs, systemic and inhaled corticosteroids, and bronchodilators were also considered potential confounding variables. Smoking status (history of smoking, no history of smoking, or unknown) and body mass index (BMI; <25 kg/m², ≥25 kg/m², or unknown) were also included (some practices do not enter these data since it is not part of required data collection).

Final regression models were determined by backward elimination using a significance level of .25. The ORs of the models were compared with the ORs of the models including all variables to identify confounding by an eliminated variable, with inclusion of this variable in cases of confounding.²³ We also conducted an analysis comparing statin users with patients using nonstatin lipid-lowering drugs or those with untreated hypercholesterolemia. This analysis used unconditional logistic regression because of the small number of pairs in which both case and control were using nonstatin or statin lipid-lowering drugs or had hypercholesterolemia.

A total of 81 880 patients in the study population had a fracture. Women comprised 75.6% and the mean age was 70 years; control patients had a similar age and sex distribution (Table 1). Factors strongly associated with fracture risk included systemic corticosteroid use (OR, 1.36; 95% confidence interval [CI], 1.29-1.43) and hormone replacement therapy use (OR, 0.75; 95% CI, 0.71-0.79). Use of thiazides had only a marginally protective effect (OR, 0.95; 95% CI, 0.92-0.98). Prevalence of thiazide or systemic corticosteroid use was similar among statin users and nonusers, but statin users were more likely to be taking estrogens (OR, 1.62; 95% CI, 1.24-2.12). The average duration of GPRD enrollment prior to the index date was 3.1 years for cases and 2.4 years for controls. Information on BMI was present for 61.4% of cases and 55.2% of controls. Information on BMI was more likely to be recorded for patients younger than 75 years (69.3%) compared with older patients (40.9%), for men (61.6%) compared with women (57.3%), and for current statin users (86.3%) compared with nonusers (58.1%). Eighty-five percent and 57.2% of statin users and nonusers, respectively, used hormone replacement therapy, and 58.3% and 58.3% used systemic corticosteroid therapy. Among patients with BMI information, 62.0% of statin users had a BMI of 25 kg/m² or greater compared with 52.2% of nonusers.

Fracture risk was not associated with prior use of statins (Table 2); the adjusted OR for fracture in current statin users compared with nonusers was 1.01 (95% CI, 0.88-1.16). There was also no relationship between time since first statin use and fracture risk. The adjusted OR for fracture associated with long-term (>12 months) statin use was 1.17 (95% CI, 0.99-1.40) compared with nonusers; for short-term use, it was 0.71 (95% CI, 0.50-1.01). Likewise, patients who took higher dosages of statins had fracture risks similar to those who took lower dosages. Risk estimates did not vary markedly by age or sex; adjusted ORs were 0.98 (95% CI, 0.83-1.15) for patients aged 50 to 69 years and 1.09 (95% CI, 0.83-1.43) for patients aged 70 years or older. Adjusted ORs were 0.97 (95% CI, 0.75-1.25) for men and 1.03 (95% CI, 0.87-1.22) for women.

The most frequent type of fracture in the study population involved the forearm (n = 27 417 patients). The adjusted OR for this type of fracture in statin users compared with nonusers was 1.01 (95% CI, 0.80-1.27). Statin use also was not associated with significant reduction in risk of hip or vertebral fractures; the adjusted OR for hip fracture was 0.59 (95% CI, 0.31-1.13) and for vertebral fracture was 1.15 (95% CI, 0.62-2.14) (Table 3). When the analysis for hip fracture was restricted to patients with information on BMI, the OR for hip fracture among statin users moved toward the null (OR, 0.82; 95% CI, 0.37-1.81).

We also compared statin users with users of nonstatin lipid-lowering drugs and with patients with untreated hypercholesterolemia (Table 4). The adjusted OR for fracture was 1.02 (95% CI, 0.83-1.24) for current users of statins compared with users of nonstatin lipid-lowering drugs. A total of 1926 control patients and 2180 case patients had a recorded diagnosis of hypercholesterolemia and had not used lipid-lowering drugs prior to the index date; statin users had a fracture risk that was comparable with that of patients with hypercholesterolemia (OR, 1.06; 95% CI, 0.92-1.23).

Additional analyses were conducted to assess the robustness of our results. Analyses confined to cases and controls who were matched by practice, sex, and age within 1 year did not change our results appreciably (OR, 0.91 for statin use vs nonuse; 95% CI, 0.79-1.06); nor did analyses restricted to practices with ongoing data collection (OR, 1.01; 95% CI, 0.86-1.18), analyses excluding controls with a randomly selected index date (OR, 0.94; 95% CI, 0.81-1.08), adjustment for calendar year of start of GPRD enrollment (OR, 1.02; 95% CI, 0.88-1.17), or adjustment for number of years of GPRD enrollment prior to the index date (OR, 0.95; 95% CI, 0.83-1.10). Alteration of the exposure definition to 3 months before the index date resulted in an OR for fracture of 0.99 (95% CI, 0.86-1.14) with current statin use.

Use of statins was not associated with a reduction in risk of fracture among patients included in this population-based study. We found no reduction in fracture risk among long-term statin users or patients taking higher statin dosages. Finally, patients using nonstatin lipid-lowering drugs (mostly bezafibrate and cholestyramine) and patients with untreated hypercholesterolemia had comparable fracture risk. These findings do not support the hypothesis that statins protect against age-related fracture.

An important consideration in evaluating the biological plausibility for a skeletal effect of statins in humans is the extent of uptake into bone. Statins do not localize preferentially to bone. After absorption, they are biotransformed in the liver, with a high clearance due to the hepatic first-pass effect. Subsequently, they are tightly bound to plasma proteins (with the exception of pravastatin), and the extent of systemic exposure to unbound, pharmacologically active drug is extremely low.²

Three recent studies have reported that statin users are less likely to have fractures compared with nonusers. Meier et al⁸ found a protective effect of current statin use. This effect was apparent for fractures of the hip, vertebral body, and foot. Chan et al⁶ reported that there was no overall effect of statin use on fracture risk but found significant risk reductions in the subgroup of patients who filled the most statin prescriptions. The third study, by Wang et al,⁷ found a significant reduction in hip fracture risk among patients using statins. The results of these studies are inconsistent. Meier et al⁸ found fracture reductions even after relatively short durations of exposure, while Chan et al⁶ observed an effect only in the group of patients with the most exposure to statins. Only 1 study⁷ pointed to the presence of a dose-response relationship, with lower hip fracture risk following increasing use of statins.⁷ Although the a priori hypothesis of 1 of the 3 studies⁶ concerned the overall relationship between statin use and fracture risk, the report focused on the subgroup with a statistically significant reduction in risk. Such analyses have a greater likelihood of yielding false-positive results.²⁴ Preliminary observations based on a large cohort of postmenopausal women⁹ support our own findings that statins have little influence on fracture risk.

It is noteworthy that several reports, including our study, point to a lower risk of hip fracture among statin users.^6- 8 We believe that the most likely explanation for this finding is confounding by obesity. Obesity is a common correlate of hypercholesterolemia and has been associated with a reduced risk of hip fracture.²⁵ This apparent protective effect of adiposity on risk of hip fracture has been attributed to increased bone density and the local shock-absorbing capacity of fat.²² In previous epidemiological studies of statins and fracture, no adjustment was made for BMI^6- 7 or the adjustment was limited due to missing data.⁸ The most likely consequence of not fully adjusting for adiposity is the spurious finding of a lower hip fracture risk among statin users who tend toward obesity. A decreased risk of dementia was found to be associated with statin use,²⁶ which also could preferentially reduce risk of hip fracture.

The association of statin use with a reduced rate of fracture found by Meier et al,⁸ also using data from the GPRD, contrasts with our results. Our study used a conventional case-control design, while Meier et al first compiled 3 cohorts and then identified cases within these cohorts; commencement of follow-up varied among these cohorts.⁸(p3206) We used a design and analysis similar to Meier et al in a sample of 1.6 million patients registered in GPRD who had been prescribed nonsystemic corticosteriods and in whom we have previously described fracture incidence.²⁷ A total of 7743 statin users were identified, of whom 94 experienced the selected fractures. The median calendar year in which follow-up commenced was 1996 in the statin group and 1991 in the random sample of nonusers. Statin users were more likely to be excluded through previous fracture because more medical data were available prior to the start of follow-up (with a corresponding tendency to exclude high-risk patients). When controls were sampled using criteria identical to those used by Meier et al, the unadjusted OR for fracture with current statin use was 0.77 (95% CI, 0.56-1.07); this OR was similar to the unadjusted OR reported by Meier et al (OR, 0.67; 95% CI, 0.53-0.84).⁸ A conventional case-control analysis using this sample (following our methods but limiting inclusion to patients aged 50-89 years as done by Meier et al⁸) yielded an OR of 1.04 (95% CI, 0.74-1.46). In our analysis, all fracture cases were included and matched by age, sex, and calendar year of first fracture. Meier et al,⁸ however, defined follow-up differently between users of lipid-lowering drugs and nonusers (for users of lipid-lowering drugs, it related to the first prescription for a lipid-lowering drug, and for nonusers, to 1 year after computerized recording of prescriptions began, which may have been considerably earlier). Consequently, differential follow-up might have prevented matching fracture cases in the lipid-lowering group to representative controls in the random sample on length of follow-up and could explain the lower frequency of statin use among fracture cases.

There are several potential methodological limitations to our study. Statin users may have been monitored more closely than nonusers, leading to a greater likelihood of fracture ascertainment among statin users and a correspondingly lower likelihood of observing a protective effect of statins. Another limitation was that dosage information was obtained from dosage instructions of the general practitioner that had been entered into the computerized medical record. These data were not systematically recorded and had to be coded by GPRD research staff. Third, exposure to statins increased substantially over time (the study covered a period from 1987 to 1999). However, information on cases and controls was collected from the same practices, and inclusion of the start date of follow-up in regression models did not appreciably alter the results. The exposure assessment represents another possible limitation. Because we did not have data on actual compliance of patients, the statin group may have included patients who did not fill or use their prescriptions. This would lead to a lower likelihood of observing a protective effect of statins.

In conclusion, statin use was not associated with a reduction in risk of fractures. Our results do not support those of earlier epidemiological reports that suggest that statins may prevent fracture. The lower hip fracture rates previously reported among statin users are most likely explained by the residual confounding effect of obesity in these analyses. Other sources of inconsistency include variations in study design. It seems unlikely that these issues will be easily resolved without further information, most importantly data from prospective, randomized controlled trials exploring the effects of statins on bone density, biochemical markers of bone turnover, and fracture incidence.