HMG-CoA Reductase Inhibitors and the Risk of Fractures

Transcription

1 ORIGINAL CONTRIBUTION HMG-CoA Reductase Inhibitors and the Risk of Fractures Christoph R. Meier, PhD, MSc Raymond G. Schlienger, PhD Marius E. Kraenzlin, MD Brigitta Schlegel, MD Hershel Jick, MD THE 3-HYDROXY-3-METHYLglutaryl coenzyme A (HMG- CoA) reductase inhibitors (statins) are effective in reducing the risk of cardiovascular morbidity and mortality in patients with hyperlipidemia. 1-6 In addition to the ability of statins to decrease serum cholesterol levels, recent studies in animals showed that these agents have pharmacologic effects on bones. Mundy et al 7 recently reported substantial increases in bone formation and trabecular bone volume in female rats after 5 weeks of oral simvastatin administration. Additional animal studies indicated that statins may decrease the severity of steroidinduced osteonecrosis. 8,9 These results from animal studies were partly supported in a crosssectional study by Bauer et al 10 in older women taking lipid-lowering agents (mainly lovastatin). Preliminary results of this small study suggested that statin use may be accompanied by increased hip bone mineral density, thereby potentially lowering the risk of hip fractures. Furthermore, Chung et al 11 suggested in a recent study that statins may increase femur mineral density in men with type 2 diabetes mellitus. It currently is not known whether the observed effects on bone density have clinical relevance in preventing bone fractures in patients treated with See also pp 3211 and Context Recent animal studies have suggested that 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors (statins) increase bone formation, volume, and density. It is unknown whether use of statins is associated with a decreased risk of fractures in humans. Objective To determine whether exposure to statins, fibrates, or other lipidlowering drugs is associated with reduced bone fracture risk. Design Population-based, nested case-control analysis. Setting The UK-based General Practice Research Database (GPRD), comprising some 300 practices, with data collection from the late 1980s until September Subjects Within a base population of individuals aged at least 50 years (28340 individuals taking lipid-lowering drugs, untreated individuals with a diagnosis of hyperlipidemia, and randomly selected individuals without diagnosis of hyperlipidemia), we identified 3940 case patients who had a bone fracture and control patients matched for age (±5 years), sex, general practice attended, calendar year, and years since enrollment in the GPRD. Main Outcome Measures Use of statins, fibrates, or other lipid-lowering drugs in case patients vs control patients. Results After controlling for body mass index, smoking, number of physician visits, and corticosteroid and estrogen use, current use of statins was associated with a significantly reduced fracture risk (adjusted odds ratio [OR], 0.55; 95% confidence interval [CI], ) compared with nonuse of lipid-lowering drugs. Current use of fibrates or other lipid-lowering drugs was not related to a significantly decreased bone fracture risk (adjusted OR, 0.87; 95% CI, and adjusted OR, 0.76; 95% CI, , respectively). Conclusions This study suggests that current exposure to statins is associated with a decreased risk of bone fractures in individuals age 50 years and older. This finding has a potentially important public health impact and should be confirmed further in controlled prospective trials. JAMA. 2000;283: statins. If this is the case, elderly patients with hyperlipidemia who are also at increased risk of developing osteoporosis (particularly postmenopausal women) potentially may have an additional benefit from therapy with statins. Author Affiliations: Basel Pharmacoepidemiology Unit, Division of Clinical Pharmacology (Drs Meier, Schlienger, and Schlegel), and Division of Endocrinology, Diabetes, and Clinical Nutrition (Dr Kraenzlin), University Hospital, and Institute of Pharmacotherapy, Department of Pharmacy, University of Basel (Dr Schlienger), Basel, Switzerland; and Boston Collaborative Drug Surveillance We conducted a large nested casecontrol analysis using the UK-based General Practice Research Database (GPRD) to determine whether use of statins, fibrates, or other lipidlowering drugs is associated with a reduced risk of bone fractures. Program, Boston University, School of Medicine, Lexington, Mass (Drs Meier and Jick). Corresponding Author and Reprints: Christoph R. Meier, PhD, MSc, Basel Pharmacoepidemiology Unit, Division of Clinical Pharmacology, University Hospital of Basel, Petersgraben 4, CH-4031, Basel, Switzerland ( Christoph.Meier@unibas.ch) American Medical Association. All rights reserved. (Reprinted) JAMA, June 28, 2000 Vol 283, No

2 METHODS Study Population and Data Source Data were derived from the GPRD, which has been described in detail elsewhere Since 1987, more than 3 million residents in the United Kingdom have been enrolled with selected general practitioners who have agreed to provide data for research purposes to the GPRD. The age and sex distribution of the patients enrolled is representative of the entire UK population. The general practitioners received 12 months of instruction on standardized recording via computer of anonymous information, which they agreed to supply continuously to academic researchers. The information recorded includes patient demographics and characteristics (eg, height, weight, and smoking status); symptoms; medical diagnoses; referrals to consultants; hospital admissions; and drug prescriptions, including the specific preparation, route of administration, dose, and number of tablets for each prescription. On request, hospital discharge and referral letters are available for review to validate the diagnoses recorded in the computer record. The GPRD currently encompasses some 30 million person-years of follow-up; it has been the source for numerous epidemiological studies in recent years, and the accuracy and completeness of these data have been well documented and validated. 13,15,16 We analyzed data from the GPRD starting in the late 1980s through September Cohort Definition and Follow-up Within the GPRD, a base population consisting of 3 separate groups was identified: group 1 included all patients who received at least 1 prescription for a statin (ie, atorvastatin, cerivastatin, fluvastatin, pravastatin, or simvastatin), a fibrate (ie, bezafibrate, ciprofibrate, clofibrate, fenofibrate, or gemfibrozil), or a lipid-lowering drug other than statins or fibrates (ie, colestipol hydrochloride, cholestyramine, acipimox, or nicotinic acid) at age 50 through 89 years (at the time of the first prescription). Group 2 comprised patients with a computer-recorded diagnosis of hyperlipidemia (International Classification of Diseases, Eighth Revision [ICD-8], code 272.x) who did not receive any lipid-lowering drug treatment. Group 3 was a random sample of patients who had neither a computer-recorded diagnosis of hyperlipoproteinemia nor a prescription for a lipid-lowering drug at any time. Within a base population consisting of these 3 groups, we followed each patient from the start of follow-up until the person developed a fracture, left the practice, or died. We defined start of follow-up as the date of the first prescription for any lipid-lowering study drug as defined above (group 1) or as the date exactly 1 year after computer recording of prescriptions began (groups 2 and 3). All patients were aged 50 through 89 years. We excluded individuals with a computer-recorded diagnosis of osteoporosis, osteomalacia, cancer (excluding nonmelanoma skin cancer), or alcoholism and patients who used bisphosphonates (considered an indicator for osteoporosis or bone metastases) prior to start of follow-up. Case Definition and Nested Case-Control Analysis Within the base population (ie, the 3 groups combined) we identified by ICD-8 codes all patients who developed a first-time diagnosis of a fracture of the femur; humerus; hand, wrist, or lower arm; vertebrae; clavicle; foot or malleolus; or an unspecified fracture after start of follow-up. The date of the fracture will subsequently be referred to as the index date. We reviewed a random sample of 200 case records by hand to verify the computerrecorded diagnosis and quantify the proportion of patients with fractures due to severe trauma (eg, vehicular collision). Within this sample, there was only 1 case patient with evidence for fracture due to severe trauma, a proportion that we considered negligible, so we included all cases of fracture identified on computer. From the base population, we randomly selected up to 6 control patients per case patient matched for age (±5 years), sex, general practice attended, calendar time (by using the same index date as for cases; ie, the date of the first diagnosis of a bone fracture), and years of prior history in the GPRD (matching on the year of start of follow-up [±1 year]). Furthermore, controls had to be alive and still enrolled at the index date. The same exclusion criteria were applied to control patients as to case patients. Statistical Analysis We conducted a matched analysis (conditional logistic regression) to explore the association between the risk of having a bone fracture and type of exposure (statins, fibrates, other lipid-lowering drugs, mixed use [switched between drug classes or used 2 or more drug classes concomitantly, regardless of timing of exposure] or none), exposure timing (current use, defined as having had the last prescription for a lipid-lowering drug 30 days preceding the index date; recent use, within days; and past use, 90 days prior to the index date), and exposure duration (by number of prescriptions, in categories of 1-4, 5-9, 10-19, and 20 prescriptions). A prescription for a lipid-lowering drug usually lasts for 30 days. In addition to matching for age, sex, general practice, calendar time, and years of history recorded in the GPRD prior to the index date, we controlled the analysis for the potential confounders smoking status (none, current, past, or unknown), body mass index ( 25, , 30 kg/m 2, and unknown), exposure to oral or inhaled corticosteroids, hormone replacement therapy with estrogens, and number of general practice visits prior to the index date. The analysis was performed using SAS, version 6.12 (SAS Institute Inc, Cary, NC). Odds ratios (ORs) are presented with 95% confidence intervals (CIs); P values are 2-tailed. RESULTS The base population comprised individuals, consisting of users of lipid-lowering drugs (group 1), 3206 JAMA, June 28, 2000 Vol 283, No. 24 (Reprinted) 2000 American Medical Association. All rights reserved.

3 13271 individuals with a diagnosis of hyperlipidemia who did not use lipidlowering drugs (group 2), and randomly selected individuals (group 3). During follow-up, 3940 individuals developed a bone fracture and were defined as cases: 705 in group 1 (2.5%), 681 in group 2 (5.1%), and 2554 in group 3 (5.1%). A total of control patients were matched to the 3940 case patients, averaging 5.93 controls per case. For 7 cases (0.18%), no eligible control could be identified; more than 99.4% of cases had 6 matched controls. The average number of years of medical history recorded prior to the index date was similar for cases (3.3 years) and controls (3.2 years). The distributions of age and sex and the independent associations between patient characteristics (body mass index, smoking status, use of corticosteroids or estrogens, and number of general practice visits before the index date) and the risk of having a bone fracture (as assessed in univariate analyses) are displayed in TABLE 1. These covariates have been taken into account in all subsequent multivariate analyses. When we compared cases with a diagnosis of hyperlipidemia who did not use lipid-lowering drugs with cases without a diagnosis of hyperlipidemia, the risk of having a bone fracture was almost identical in the 2 groups. Compared with the referent of normolipidemic nonusers of lipid-lowering drugs, the relative risk estimate for hyperlipidemic nonusers of lipid-lowering drugs was 0.95 (95% CI, ). Since there was no material difference between normolipidemic and hyperlipidemic nonusers of lipid-lowering drugs, we combined all nonusers into 1 reference group for subsequent analyses. Compared with this reference group of nonusers of any lipid-lowering drugs, the ORs for current exposure of 1 to 4, 5 to 19, or 20 or more computerrecorded prescriptions for statins prior to the index date were 0.51 (95% CI, ), 0.62 (95% CI, ), and 0.52 (95% CI, ), respectively, after adjusting for body mass index, smoking status, number of general practice visits, and use of corticosteroids or estrogens. The adjusted ORs for any current, recent, or past exposure to statins, regardless of the total number of prescriptions, were 0.55 (95% CI, ), 0.67 (95% CI, ), and 0.87 (95% CI, ), respectively (TABLE 2). As compared with the same reference group of nonusers of any lipidlowering drugs, current, recent, or past exposure to fibrates (regardless of the Table 1. Characteristics of Case Patients and Control Patients and Association With Fracture Risk by Univariate Analyses* Characteristics Case Patients, No. (%) (n = 3940) Control Patients, No. (%) (n = ) (95% CI) P Value Age, y (18.4) 4343 (18.6) (32.7) 7892 (33.7) (28.2) 6796 (29.0) (20.6) 4348 (18.6) Sex Women 2956 (75.0) (75.0) Men 984 (25.0) 5872 (25.1) Smoking Nonsmoker 1980 (50.2) (50.7) Current 531 (13.5) 2996 (12.8) 1.07 ( ).22 Ex-smoker 393 (10.0) 2142 (9.2) 1.12 ( ).07 Unknown 1036 (26.3) 6383 (27.3) 0.95 ( ).32 Body mass index, kg/m (27.4) 6025 (25.8) (22.3) 5692 (24.3) 0.86 ( ) (10.0) 2572 (11.0) 0.86 ( ).02 Unknown 1586 (40.3) 9090 (38.9) 0.99 ( ).82 Number of general practitioner visits (9.2) 2778 (11.9) (12.6) 3542 (15.1) 1.15 ( ) (39.2) 9222 (39.4) 1.49 ( ) (24.9) 5317 (22.7) 1.81 ( ) (14.1) 2520 (10.8) 2.35 ( ).001 Corticosteroid use None 2979 (75.6) (78.6) Oral (No. of prescriptions) (7.6) 1585 (6.8) 1.19 ( ) (0.7) 136 (0.6) 1.26 ( ) (0.4) 71 (0.3) 1.39 ( ) (0.9) 111 (0.5) 2.03 ( ).001 Inhaled (No. of prescriptions) (6.2) 1406 (6.0) 1.08 ( ) (0.8) 235 (1.0) 0.84 ( ) (0.5) 99 (0.4) 1.33 ( ) (0.9) 126 (0.5) 1.62 ( ).01 Mixed 248 (6.3) 1236 (5.3) 1.25 ( ).002 Estrogen use (No. of prescriptions) None 3708 (94.1) (92.8) Current, (1.0) 323 (1.4) 0.70 ( ).04 Current, (0.5) 219 (0.9) 0.53 ( ).007 Current, (0.4) 198 (0.9) 0.41 ( ).001 Past, (4.0) 948 (4.0) 0.92 ( ).38 *CI indicates confidence interval; ellipses, not applicable. Number of general practitioner visits indicate number of visits before the index date (date of fracture) American Medical Association. All rights reserved. (Reprinted) JAMA, June 28, 2000 Vol 283, No

4 number of prescriptions) yielded adjusted ORs of 0.87 (95% CI, ), 1.05 (95% CI, ), and 0.85 (95% CI, ), respectively. Current, recent, or past use of other lipid-lowering agents resulted in adjusted ORs of 0.76 (95% CI, ), 1.19 (95% CI, ), and 0.97 (95% CI, ), respectively (Table 2). We further stratified the analysis of current use of statins or fibrates by age (50-69 and years), sex, and outcome (fracture of femur, humerus or arm, vertebrae, malleolus or foot, or other [including clavicle and unspecified fractures]) to detect possible effect modification. The results of these subanalyses did not suggest that the effects of statins on fracture risk differed materially by bones affected, age groups, or sex. Odds of fracture among those with current exposure to fibrates differed between men (OR, 0.50; 95% CI, ) and women (OR, 1.01; 95% CI, ) (TABLE 3). We also explored the effect of current exposure to individual lipid-lowering drugs on the risk of developing bone fractures. The effects were consistent within groups of lipid-lowering drugs; ie, all individual statins were associated with decreased fracture risks. From the patient records, we assessed additional covariates that have been related to an altered risk of having a bone fracture and explored whether their inclusion in the multivariate regression model changed the results. There was no evidence of confounding of the association between lipid-lowering drugs and the bone fracture risk by including dichotomous variables for a history of chronic renal failure, hyperthyroidism, hyperparathyroidism, inflammatory bowel disease, malnutrition or malabsorption, or current exposure to benzodiazepines, neuroleptics, antidepressants, antihypertensives, calcium, fluoride, or vitamin D preparations, although current exposure to drugs affecting the central nervous system (ie, benzodiazepines, antidepressants, and antipsychotics) was associated with a slightly increased risk of bone fractures (OR, 1.20; 95% CI, ). COMMENT The findings of this large nested casecontrol analysis indicate that exposure to statins (HMG-CoA reductase inhibitors) is associated with a substantially lower risk of developing fractures in humans. The association was mainly present for current users of statins (OR, 0.55; 95% CI, ) and could be identified even after relatively short exposure duration (ie, 1-4 prescriptions, corresponding to a treat- Table 2. Exposure to Lipid-Lowering Drugs and Association With Fracture Risk in Multivariate Analysis* Exposure No. of Case Patients (n = 3940) No. of Control Patients (n = ) Unadjusted (95% CI) Adjusted (95% CI) None P Value Statins only Current use ( ) 0.55 ( ).001 No. of prescriptions ( ) 0.51 ( ) ( ) 0.62 ( ) ( ) 0.52 ( ).001 Recent use ( ) 0.67 ( ).01 Past use ( ) 0.87 ( ).37 Fibrates only Current use ( ) 0.87 ( ).19 No. of prescriptions ( ) 0.78 ( ) ( ) 0.79 ( ) ( ) 0.95 ( ).71 Recent use ( ) 1.05 ( ).72 Past use ( ) 0.85 ( ).27 Other lipid-lowering drugs only Current use ( ) 0.76 ( ).37 No. of prescriptions ( ) 0.90 ( ) ( ) 0.95 ( ) ( ) 0.49 ( ).24 Recent use ( ) 1.19 ( ).56 Past use ( ) 0.97 ( ).87 Mixed use ( ) 1.11 ( ).25 *CI indicates confidence interval. Odds ratios are adjusted for body mass index, smoking, number of general practitioner visits, and steroid or estrogen use. P values are for adjusted odds ratios. For definition, see Methods section JAMA, June 28, 2000 Vol 283, No. 24 (Reprinted) 2000 American Medical Association. All rights reserved.

5 ment duration of approximately 1-4 months). The reduced fracture risk was observed in various skeletal sites. There was no evidence for material effect modification by age or sex in this study population because the association between statin exposure and fracture risk did not differ by age group or sex. In contrast, there was little evidence that fibrates or other lipidlowering drugs are related to altered bone fracture risk, even though the current analysis does not entirely rule out an effect of fibrates on fracture risk. Current fibrate use was associated with a reduced odds of fracture in men and of vertebral fractures. However, some of these findings were based on small numbers, and the findings overall suggest that the association between drug use and decreased fracture risk may be most pronounced for statins. Exposure to statins was associated with a decreased fracture risk even after a short exposure duration of a few weeks to a few months. Mundy et al 7 reported an increase in bone formation and some inhibition of osteoclastic activity by statins in their rodent model after 5 weeks of treatment. Other studies have shown that intervention with bisphosphonates reduces osteoclastic bone resorption significantly after 1 week of treatment. 17,18 Thus, a rapid change in bone remodeling seems possible, but the dynamics of the effect of statins on bone remodeling needs further investigation. As in previous investigations, we also found independent effects of several covariates on the fracture risk, such as a small decreased risk of osteoporotic fractures in relation to increasing body mass index, substantially decreased fracture risk associated with estrogen use, and increased fracture risk for longer-term use of oral corticosteroids and drugs and affecting the central nervous system (ie, benzodiazepines, antidepressants, and antipsychotics) in the current study population. As with all observational studies, biases or unknown confounders cannot be completely ruled out as alternative explanations for the findings. A limitation of this study is the lack of information on physical activity of cases and controls. However, the proposition that individuals may dramatically change their lifestyle habits and begin exercising after a diagnosis of hyperlipidemia cannot sufficiently explain the findings because physical activity has only a moderate effect in slowing age-related bone loss, and its effect in reducing the fracture risk can only be expected in the long-term. 19,20 In addition, it seems unlikely that such substantial confounding would selectively affect statins but not other classes of lipid-lowering drugs. The same is true for a possible confounding effect of low socioeconomic status, which may be related to both decreased access to medical care and to the likelihood of getting treatment with lipid-lowering drugs. We attempted to control for this possibility at least in part by matching for general practice. Regardless, it seems unlikely that potential confounding by this parameter would selectively affect the findings for statins but not for fibrates or other lipidlowering drugs. In addition, subtle differences in access to medical care do not seem to be particularly relevant for an acute outcome that requires medical attention regardless of socioeconomic status, such as bone fracture. An exception may be vertebral fractures because a considerable number of vertebral fractures remain undiagnosed; it has been reported that less than one third of patients with vertebral deformities present to medical practitioners. 21 The number of vertebral fractures was quite low in the current study; however, possible underdiagnosis of this type of fracture does not affect the validity of the diagnosed and computer-recorded fractures in general. In epidemiological research, it is generally important to include only wellvalidated cases in an analysis. For this particular study, we did not validate the case diagnoses (ie, fractures) by review of original medical records because this Table 3. Association of Current Use of Statins or Fibrates With Fracture Risk, Stratified by Type of Fracture, Age, and Sex in a Multivariate Analysis* Characteristics Case Patients, No. (%) (n = 3940) Cases Current Use of Statins Controls (95% CI) P Value Cases Controls Current Use of Fibrates (95% CI) P Value Type of fracture Femur 678 (17.2) ( ) ( ).21 Hand, wrist, or arm 1906 (48.4) ( ) ( ).38 Vertebrae 98 (2.5) ( ) ( ).05 Foot 748 (19.0) ( ) ( ).76 Other 510 (12.9) ( ) ( ).63 Age, y (51.2) ( ) ( ) (48.8) ( ) ( ).62 Sex Men 984 (25.0) ( ) ( ).007 Women 2956 (75.0) ( ) ( ).94 *CI indicates confidence interval. Odds ratios are for comparison with nonusers of lipid-lowering drugs (referent) and are adjusted for body mass index, smoking, number of general practitioner visits, and steroid or estrogen use. Includes clavicle and unspecified fracture American Medical Association. All rights reserved. (Reprinted) JAMA, June 28, 2000 Vol 283, No

6 outcome of interest has been well documented in the computer records and left little room for misclassification. In conclusion, we found that current use of statins in men and women aged 50 years or older was associated with a reduced risk of fracture. This association was much weaker or not present for fibrates and other lipid-lowering drugs, and there was no risk difference between nonusers of lipid-lowering drugs with or without a recorded diagnosis of hyperlipidemia. Despite strong evidence from the current analysis, it is necessary to obtain additional information from controlled trials to further investigate a possible causal effect of statins on bone fracture risk, and, if an effect exists, further assess the timing of onset of such an effect. Funding/Support: The Boston Collaborative Drug Surveillance Program is supported in part by grants from AstraZeneca, Berlex Laboratories, Boehringer Ingelheim Pharmaceuticals Inc, Boots Healthcare International, Bristol-Myers Squibb Pharmaceutical Research Institute, Glaxo Wellcome Inc, Hoffmann-La Roche Ltd, Janssen Pharmaceutica LP, RW Johnson Pharmaceutical Research Institute, McNeil Consumer Products Company, and Novartis Farmacéutica SA. This particular study was not funded. Dr Meier is a recipient of a grant from the Swiss National Science Foundation (grant number ). Acknowledgment: We thank the participating general practitioners for their excellent cooperation. REFERENCES 1. Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet. 1994;344: Byington RP, Jukema W, Salonen JT, et al. Reduction in cardiovascular events during pravastatin therapy: pooled analysis of clinical events of the pravastatin atherosclerosis intervention program. Circulation. 1995; 92: West of Scotland Coronary Prevention Group. Influence of pravastatin and plasma lipids on clinical events in the West of Scotland Coronary Prevention Study (WOSCOPS). Circulation. 1998;97: Shepherd J, Cobbe SM, Ford I, et al. Prevention of coronary heart disease with pravastatin in men with hypercholesterolemia.nengljmed.1995;333: Downs JR, Clearfield M, Weis S, et al. Primary prevention of acute coronary events with lovastatin in men and women with average cholesterol levels: results of AFCAPS/TexCAPS: Air Force/Texas Coronary AtherosclerosisPreventionStudy.JAMA.1998;279: LaRosa JC, He J, Vupputuri S. Effect of statins on risk of coronary disease: a meta-analysis of randomized controlled trials. JAMA. 1999;282: Mundy G, Garrett R, Harris S, et al. Stimulation of bone formation in vitro and in rodents by statins. Science. 1999;286: Wang GJ, Chung KC, Shen WJ. Lipid clearing agents in steroid-induced osteoporosis. J Formos Med Assoc. 1995;94: Cui Q, Wang GJ, Su CC, Balaian G. The Otto Aufranc award: lovastatin prevents steroid induced adipogenesis and osteonecrosis. Clin Orthop. 1997;344: Bauer DC, Mundy GR, Jamal SA, et al. Statin use, bone mass and fracture: an analysis of two prospective studies [abstract]. J Bone Miner Res. 1999;14 (suppl 1):S Chung YS, Lee MD, Lee SK, Kim HM, Fitzpatrick LA. HMG-CoA reductase inhibitors increase BMD in type 2 diabetes mellitus patients. J Clin Endocrinol Metab. 2000;85: Jick H. A database worth saving. Lancet. 1997; 350: Walley T, Mantgani A. The UK General Practice Research Database. Lancet. 1997;350: Garcia Rodriguez LA, Perez Gutthann S. Use of the UK General Practice Research Database for pharmacoepidemiology. Br J Clin Pharmacol. 1998;45: Jick H, Jick SS, Derby LE. Validation of information recorded on general practitioner based computerized data resource in the United Kingdom. BMJ. 1991;302: Jick H, Terris BZ, Derby LE, Jick SS. Further validation of information recorded on a general practitioner based computerized data resource in the United Kingdom. Pharmacoepidemiol Drug Saf. 1992;1: Pedrazzoni M, Alfano FS, Gatti C, et al. Acute effects of biphosphonates on new and traditional markers of bone resorption. Calcif Tissue Int. 1995;57: Gamero P, Shih WJ, Gineyts E, Karpf DB, Delmas PD. Comparison of new biochemical markers of bone turnover in late postmenopausal osteoporotic women in response to alendronate treatment. J Clin Endocrinol Metab. 1994;79: Wolff I, van Croonenborg JJ, Kemper HC, Kostense PJ, Twisk JW. The effect of exercise training programs on bone mass: a meta-analysis of published controlled trials in pre- and postmenopausal women. Osteoporos Int. 1999;9: Prince RL, Smith M, Dick IM, et al. Prevention of postmenopausal osteoporosis: a comparative study of exercise, calcium supplementation, and hormonereplacement therapy. N Engl J Med. 1991;325: Cooper C, Atkinson EJ, O Fallon WM, Melton LJ. Incidence of clinically diagnosed vertebral fractures: a population based study in Rochester, Minnesota, J Bone Miner Res. 1992;7: JAMA, June 28, 2000 Vol 283, No. 24 (Reprinted) 2000 American Medical Association. All rights reserved.