By Edmund Lau, MS, Kevin Ong, PhD, Steven Kurtz, PhD, Jordana Schmier, MA, and Av Edidin, PhD

Transcription

1 1479 COPYRIGHT Ó 2008 BY THE JOURNAL OF BONE AND JOINT SURGERY, INCORPORATED Mortality Following the Diagnosis of a Vertebral Compression Fracture in the Meicare Population By Emun Lau, MS, Kevin Ong, PhD, Steven Kurtz, PhD, Jorana Schmier, MA, an Av Eiin, PhD Investigation performe at Exponent, Inc., Menlo Park, California Backgroun: Vertebral compression fractures in women are associate with increase mortality, but the generality of this fining, as a function of age, sex, ethnicity, an region, among the entire elerly population in the Unite States remains unclear. The objective of this stuy was to assess the survival of the Meicare population with vertebral compression fractures. Methos: We conucte a retrospective ata analysis of Meicare claims generate by a 5% sample of all Meicare enrollees from 1997 through The patient sample consiste of all 97,142 iniviuals with a new iagnosis of vertebral compression fracture from 1997 through Controls were matche for age, sex, race, an Meicare buyin status, with a five-to-one control-case ratio. The survival of a patient was measure from the earliest ate of a new fracture until eath or until the en of the stuy. The patients with a fracture were compare with the controls by calculation of the mortality rates, with use of Kaplan-Meier analysis an the Cox regression metho. Demographic subpopulation analysis an analysis by comorbiity levels were performe as well. Results: Meicare patients with a vertebral fracture ha an overall mortality rate that was approximately twice that of the matche controls. The survival rates following a fracture iagnosis, as estimate with the Kaplan-Meier metho, were 53.9%, 30.9%, an 10.5% at three, five, an seven years, respectively, which were consistently an significantly lower than the rates for the controls. The mortality risk following a fracture was greater for men than for women. The ifference in mortality between the patients with a vertebral compression fracture an the controls was greatest when the patients were younger at the time of the fracture; this ifference ecline as the age at the time of the fracture increase. Conclusions: This stuy establishes the mortality risk associate with vertebral fractures for elerly patients of all ages an ethnicities an both sexes in the Meicare population; however, it oes not imply a causal relationship. The ifference in mortality between patients with a fracture an controls is higher than previously reporte, even after controlling for comorbiities. Level of Evience: Prognostic Level III. See Instructions to Authors for a complete escription of levels of evience. Avertebral compression fracture is a serious complication associate with osteoporosis of the spine. The prevalence of osteoporosis in the Unite States is high, especially among women, an it increases rapily with age. Accoring to one estimate, osteoporosis affects as many as one in three women an one in eight men over the age of fifty years worlwie 1. In patients with a vertebral compression fracture, the structural strength of the vertebral boy is weakene by the loss of mineral an bone mass, an it is no longer able to sustain the compressive loa from the weight of the upper boy an aitional stress brought on by boy movements an other activities. Vertebral boy fractures ue to compression cause pain, eformity, an restricte mobility. In the U.S., about 1.5 million fractures occur annually, an 700,000 of these occur in the spine 2. Each year, 250,000 new vertebral compression fractures are iagnose, an >80% of these Disclosure: In support of their research for or preparation of this work, one or more of the authors receive, in any one year, outsie funing or grants in excess of $10,000 from Kyphon, Inc. Neither they nor a member of their immeiate families receive payments or other benefits or a commitment or agreement to provie such benefits from a commercial entity. No commercial entity pai or irecte, or agree to pay or irect, any benefits to any research fun, founation, ivision, center, clinical practice, or other charitable or nonprofit organization with which the authors, or a member of their immeiate families, are affiliate or associate. A commentary is available with the electronic versions of this article, on our web site ( an on our quarterly CD-ROM (call our subscription epartment, at , to orer the CD-ROM). J Bone Joint Surg Am. 2008;90: oi: /jbjs.g.00675

2 1480 are relate to weakene vertebral boies seconary to osteoporosis 3. The averse effects associate with a vertebral compression fracture have been stuie an reporte by several researchers. In aition to increase back pain, the overall quality of life 4,5 may be iminishe. Patients with a vertebral compression fracture are more likely to fall an are five times more likely to sustain subsequent fractures than iniviuals without such a fracture 6. Although a ecrease in the quality of life has been ocumente in a number of stuies, the association between vertebral compression fracture an patient longevity has been less well stuie. In one stuy of the effect of vertebral compression fracture on survival 7, the authors foun a 23% increase in mortality risk after a mean uration of follow-up of 8.3 years an an even higher increase (110%) in the risk of eath from pulmonary isease. The purpose of the present stuy was to examine the survival of a broa group of patients with a vertebral compression fracture, incluing men, nonwhite patients, an patients from rural an suburban communities, an to examine ifferences in survival among these subpopulations. Materials an Methos Patient Cohort The cohort of patients with a symptomatic vertebral compression fracture was ientifie from Meicare claims submitte by physicians, hospitals, an outpatient clinics from January 1, 1997, through December 31, We inclue both pathologic an traumatic fractures. One or more claims listing an International Classification of Diseases, Ninth Revision, Clinical Moification (ICD-9-CM) iagnosis coe of (pathologic fracture of vertebrae) or 805.0, 805.2, 805.4, 805.6, or (cervical, thoracic, lumbar, sacrum/coccyx, an other unspecifie vertebral fractures) was require for inclusion in the stuy. Claims recors from the 5% Meicare Beneficiary Encrypte File (BEF) were use in this stuy. This file is a ranom sample compile from claims from beneficiaries whose health insurance claims number contains the requisite igits in the sampling plan evelope by the Centers for Meicare an Meicai Services. Meicare is the largest nationwie insurance program that provies hospital insurance (Part A) or meical insurance (Part B) to the elerly. The 5% systematic sample reflects the claims generate by approximately two million Meicare enrollees. Although the Meicare ata inclue some claims from nonelerly patients with a vertebral compression fracture who belong to special populations (for example, those with en-stage renal isease), this stuy inclue only patients with a vertebral compression fracture who were sixty-five years of age or oler at the time of the iagnosis. Patients enrolle in health maintenance organizations, those not enrolle in both Part A an Part B of Meicare, an those resiing outsie of the U.S. were exclue. For each patient iagnose with a vertebral compression fracture, we efine an inex event, which was the earliest iagnosis of a vertebral compression fracture ientifie from TABLE I Demographic Characteristics of Patients with Vertebral Compression Fracture an Controls Patients with Vertebral Compression Fracture (N = 97,142) Controls (N = 428,956) No. % No. % Sex Male 24, , Female 72, , Age yr 10, , yr 15, , yr 21, , yr 22, , yr 28, , Race White 91, , Black , Other/unknown , Charlson inex categories 0 17, , , , , , , , Buy-in Yes 19, , No 78, , any of the three claim sources examine. To ensure that the inex event inee represente a new fracture, we require that the inex event ha occurre after the patient was enrolle in Meicare an that the patient ha not ha a vertebral compression fracture for at least 180 ays prior to the inex event. On the basis of the above criteria, 97,142 elerly patients with a new iagnosis of vertebral compression fracture were ientifie for the stuy. Comparison Cohort To evaluate the survival of patients following a vertebral compression fracture, a comparison cohort, or control group, was also selecte. The ratio of controls to patients with a vertebral compression fracture was five to one. Eligible controls were selecte from the same 5% sample of Meicare enrollees. Each patient in the final fracture cohort was matche with patients, from the eligible pool of controls, whose ages were within three years of the age of the case patient an who were at least sixtyfive years of age. If more eligible controls were available than the five require for each case, the controls closest in age were selecte. Eligible controls were also of the same sex an race (white, black, or other/unknown) as the case patient. Eligible controls ha to be enrolle in both Part A an Part B of Meicare an coul not be enrolle in a health maintenance

3 1481 TABLE II Summary of Mortality of Patients with Vertebral Compression Fracture an Controls Patients with Vertebral Compression Fracture Controls Person-Years Deaths Rate Person-Years Deaths Rate Ratio Overall 171,444 39, ,109,763 94, Sex Male 38,039 11, ,826 27, Female 133,405 27, ,937 67, Age yr 17,768 2, , yr 28,767 4, , yr 38,880 7, ,473 13, yr 39,786 9, ,464 22, yr 46,244 15, ,101 48, Race White 161,430 37, ,041,154 90, Black , Other/unknown , Charlson inex categories 0 29, ,936 24, ,660 10, ,201 30, ,233 11, ,354 21, ,933 15, ,272 17, Buy-in Yes 37, ,070 26, No 134,315 30, ,693 68, organization. Patients with a vertebral compression fracture who receive state payment assistance with their Meicare premium (that is, state buy-in) were matche with controls with the same status. Although not ieal, this buy-in inicator was use as a marker of iniviuals with limite financial resources, which may affect survival. Eligible controls coul not have been iagnose with a vertebral compression fracture at any time uring the stuy perio. The sampling plan for the controls was esigne so that the controls for each patient with a vertebral compression fracture were not use again i.e., a control in a case-control set appeare only once in the stuy. A small number of patients with a vertebral compression fracture (for example, black ninety-year-ol men in Western states) i not have a sufficient number of matching controls for the five-to-one sampling plan. In those cases, all of the eligible controls were use. A total of 428,956 controls were selecte for the 97,142 patients with a vertebral compression fracture (average, 4.4 controls per case). These frequency-matche case-control groups forme the basis of the survival comparison. The general characteristics of the patients an controls are summarize in Table I. Follow-up The eath of an enrollee was ascertaine by the Centers for Meicare an Meicai Services with use of ata from the Social Security Aministration an other sources. The ate of eath was available from the annual Meicare enominator file for enrollees in the 5% sample who haie in that year. Survival of a patient with a vertebral compression fracture was calculate as the time from the inex ate until eath or until the en of follow-up on December 31, Patients with a vertebral compression fracture who were alive on December 31, 2004, were consiere censore. Similarly, survival of the five matching controls was calculate from the inex ate of the case patient in each of the case-control groups. This stuy affore a minimum uration of follow-up of one ay an a maximum uration of 7.5 years from July 1, 1997, until December 31, Comorbiity Ajustment To account for the health status of each patient in this comparison, the Charlson comorbiity inex (Dartmouth-Manitoba version 8 ) was calculate. Because all of the iagnoses an proceures receive by a Meicare beneficiary are ientifie in the claims ata, the Meicare ata are well suite for characterizing the overall health status of a patient. All isease iagnoses an surgical proceures performe in a six-month perio prior to the inex event were compile from the Meicare claims. The Charlson comorbiity algorithm utilizes nineteen categories of iseases base on iagnoses an surgical proceures receive by a patient, an the final inex is a composite value representing the overall egree of comorbiity. In this analysis, the Charlson inex values were groupe into four categories 9 :0(none),1to2 (low), 3 to 4 (moerate), an 5 (high).

4 1482 TABLE III Results of Kaplan-Meier Analysis Comparing Patients with Vertebral Compression Fracture an Controls Patients with Vertebral Compression Fracture Controls 8-Year Survival Rate (%) Mean Duration of Survival (mo) Meian Duration of Survival (mo) 8-Year Survival Rate (%) Mean Duration of Survival (mo) Meian Duration of Survival (mo) Overall Sex Male Female Age yr yr yr yr yr Race White Black Other/unknown Charlson inex categories Buy-in Yes No Statistical Analysis Two types of analyses were performe: calculations of mortality rates an multivariate proportional hazar regressions. The mortality status of patients with a vertebral compression fracture an controls was summarize by calculating the total person-years of follow-up, total eaths, aneath rates (eaths per 1000 person-years). The overall survival of the patients with a vertebral compression fracture an the controls was Fig. 1 Survival curves for men (left) an women (right). VCF = vertebral compression fracture.

5 1483 Fig. 2 Survival curves for ifferent age groups. VCF = vertebral compression fracture, an y.o. = years ol. estimate with the Kaplan-Meier metho along with means an stanareviations of survival time. Kaplan-Meier survival curves following the iagnosis of a vertebral compression fracture were constructe for the fracture an control cohorts an various emographic subpopulations. Differences in survival were assesse with the log-rank test. The Kaplan-Meier survival curves provie a graphical view of the ifference in mortality between the patients with a vertebral compression

6 1484 TABLE IV Summary of Hazar Ratios Determine with Cox Regression Analysis Comparing Patients with Vertebral Compression Fracture an Controls Unajuste Ajuste for Comorbiity Ratio 95% Confience Interval Ratio 95% Confience Interval Overall Sex Male Female Age yr yr yr yr yr Race White Black Other/unknown Charlson inex categories Buy-in No Yes fracture an the controls. The log-rank test results an the survival curves escribe variations in survival characteristics among these subgroups. A more comprehensive assessment of the survival characteristics was accomplishe with the Cox regression analysis. Consistent with the stuy esign, the Cox regression was stratifie by the iniviual matche case-control set. This use of the Cox regression allowe non-proportionality to exist across ifferent strata while still proviing an estimate of the overall hazar ratio for the ifference between the patients with a vertebral compression fracture an the controls. Results Description of Cohort The emographic characteristics of the patients with a vertebral compression fracture an their controls are summarize in Table I. Because the controls were selecte by matching the case on the basis of several characteristics, the istributions of these characteristics in the patient an control groups were nearly ientical. Of the 97,142 patients with a vertebral compression fracture, 74.4% were women an most (93.8%) were white. The average age of the patients with a vertebral compression fracture was 79.9 years, with a stanar eviation of 7.6 years. The average age of the controls was 79.0 years, with a stanareviation of 7.5 years. Mortality an Mortality Rates From 1997 through 2004, the 97,142 patients with a vertebral compression fracture accumulate a total of 171,444 personyears of observation, with an average of 21.2 months per patient. The 428,956 controls accumulate a total of 1,109,763 person-years of follow-up, with an average of 31.0 months per person. By December 31, 2004, a total of 39,707 eaths of patients with a vertebral compression fracture were ientifie, for an overall eath rate of eaths per 1000 person-years of observation. The total number of eaths in the control group was 94,982, a eath rate of 85.6 per 1000 person-years of observation. The ratio of the two mortality rates was The person-years, eaths, eath rates, an mortality ratios for the subpopulations are shown in Table II. The survival of the patients with a vertebral compression fracture an the controls was estimate with use of the Kaplan- Meier metho. The results of the log-rank test comparing the patients with a vertebral compression fracture with the controls, in the total group an in the various subpopulations, were consistently highly significant (p < 0.001) an therefore are not iniviually reporte. Table III shows the percentages of subjects who were alive at the en of the stuy perio. Because these survival percentages were estimate with the Kaplan- Meier metho, the reporte results were base on the last noncensore recor, which may have occurre at less than the

7 1485 maximum of 7.5 years between the start an en of the stuy perio. Overall, 6.8% of the patients with a vertebral compression fracture were still alive at the en of the follow-up perio, as compare with 49.7% of the controls (Table III). Survival rates following the iagnosis of vertebral compression fracture at three, five, an seven years were 53.9%, 30.9%, an 10.5%, respectively. The corresponing survival rates were much higher for the controls (78.2%, 64.5%, an 52.5%). The patients with a vertebral compression fracture an their matching controls survive for a meian of 39.8 an 89.1 months, respectively. As a result of censoring, the average uration of survival (in months) was unerestimate an biaseownwar. Nonetheless, the obvious ifference in the percentages of surviving subjects an the urations of survival, couple with the consistent gap between the survival curves for the patients with a vertebral compression fracture an the controls, emonstrate an increase mortality risk in patients with a symptomatic vertebral compression fracture. The percentages of male an female control subjects who were still alive at the en of the stuy were comparable (47.0% an 50.7%, respectively). The percentages of male an female patients with a vertebral compression fracture who were alive were also similar (7.4% an 7.6%) (Fig. 1). The mortality rate increase with the age at the time of the fracture both for the patients with a vertebral compression fracture an for the controls (Fig. 2 an Table II). As expecte, the ifference between these two sets of rates ecrease as the age at the time of the fracture increase. Patients who were sixty-five to sixty-nine years ol when a vertebral compression fracture was iagnose ha a mortality rate that was 6.41 times higher than that of the age-matche controls. The mortality rate for the patients who were iagnose with a vertebral compression fracture when they were eighty-five years of age or oler was only 1.82 times higher than that for their controls. Similarly, the mortality rate for both sets of subjects increase with increasing Charlson inex values. The ifference between the survival of the patients with a vertebral compression fracture an that of the control subjects also ecrease with increasing Charlson inex values. Survival Comparison with Cox Regression Analysis Overall, without ajustment for ifferences in comorbiity, the hazar ratio between the patients with a vertebral compression fracture an their matche controls was 2.80 (95% confience interval, 2.77 to 2.84) (Table IV). Ajustment for the baseline health of the patients lowere the hazar ratio to 1.83 (95% confience interval, 1.80 to 1.86). Both the ajuste an unajuste hazar ratios were significant (p < 0.001). The hazar (mortality) ratios between the patients with a vertebral compression fracture an the controls in the subpopulations were also calculate, an they were consistently highly significant in all of these subgroup analyses (p < 0.001); therefore, they are not iniviually reporte. Discussion This stuy emonstrate the etrimental association between a vertebral compression fracture an survival in a large systematic sample of elerly patients from the Meicare population. Since enrollment in Meicare is nearly universal among the elerly population in the U.S., the results of this stuy apply well to all persons sixty-five years of age an oler in the U.S. We foun that, over a seven-year perio, an elerly patient with a vertebral compression fracture ha almost two times the risk of eath compare with an elerly person without a vertebral compression fracture, after ajustment for comorbiity. This ae mortality risk was especially etrimental if the vertebral compression fracture occurre at a younger age. Although vertebral compression fractures are more prevalent in women than in men, the mortality risk is higher among men. Even among patients with a vertebral compression fracture an a Charlson comorbiity inex of 0, the hazar ratio was 1.37 (p < 0.001). The survival of the patients with a vertebral compression fracture an no comorbiities was actually comparable with that of their matche controls in the first twelve to twenty-four months, but it became increasingly poorer after about two years. Previous stuies have also ocumente an increase mortality risk following vertebral fractures 7,10,11. Center et al. reporte an increase in mortality risk for both sexes following low-trauma osteoporotic fractures 10, but they also foun that men were at higher risk (a 138% increase risk) compare with women (a 66% increase risk). Mortality rates were also higher for Sweish men an women following clinical vertebral fractures, incluing low-energy fractures 11.Comparewiththe general population, men an women who sustaine such a fracture at the age of sixty years ha a 4.3-fol relative risk of eath at five years after the fracture. Overall, the Sweish patients with a vertebral fracture ha a 28% survival rate at five years, which is comparable with our rate of 30.9% at the same time point. In a stuy involving elerly white women, Kao et al. etermine that the rate of mortality of subjects with one or more fractures was 1.23-fol greater than that of iniviuals without a fracture, an this ifference increase to 1.34-fol in patients with more severe fractures 7. It is unclear to what extent the exclusion of traumatic fractures from our stuy might have affecte the mortality risk. It has been suggeste that women with vertebral fractures following low-energy or no trauma have significantly higher mortality rates (os ratio, 2.3; p < 0.05) than those with high-energy trauma 12. The age-relateifferences that we foun in the mortality risk are in agreement with those in a previous Sweish stuy by Johnell et al. 11. Those investigators showe that the risk of eath for men at five years after a fracture was increase 4.3-fol when the fracture ha been sustaine at the age of sixty years compare with 1.3-fol when it ha been sustaine at the age of eighty years. In aition, the Sweish ata showe that the highest mortality risk occurre immeiately following the fracture an that the increase in risk lessene significantly with time (p < 0.001). For example, the increase in the risk of eath for men who ha sustaine a fracture at the age of sixty years was 13.4-fol in the initial year but only 4.3-fol by the fifth year. Our stuy showe comparable results, as the survivorship ecrease markely in the first six months, after which

8 1486 there was a more steaily eclining tren. However, it is unclear if these age-relateifferences apply to the non-meicare an younger patient population. This stuy ha several limitations. We were unable to ientify the severity of each vertebral compression fracture because of the aministrative nature of the Meicare ata. In aition, we i not examine the number of fractures sustaine by each patient. It is likely that the cumulative effect of multiple fractures may lea to progressive ebilitation an ultimately the risk of complications aneath. Kao et al. showe that the increase in mortality risk was 23% in patients with a fracture but 34% in those with a severe fracture an patients with multiple fractures also ha a significantly greater mortality risk (p < 0.001) 7. Our stuy inclue only iagnose, symptomatic vertebral compression fractures, but as many as 80% of vertebral fractures may not be clinically apparent or may remain uniagnose 3,7. Even though we attempte to ajust for comorbiities an other emographic factors, there may still have been other resiual confouning factors ue to inaequate ajustment 13. Part of the observeifference in mortality risk coul have been ue to confouning effects such as smoking, obesity, or other meical problems rather than the effects of the vertebral compression fracture itself. We also relie on aministrative ata, which may not have inclue clinical etails that woul otherwise be available from patient meical recors. Hence, the ifference in mortality coul have been ue to unmeasureifferences between the fracture an control populations. Although patients with a vertebral compression fracture were foun to have greater mortality rates, the present stuy i not inclue an investigation of the cascae of events leaing to their premature eaths. A iagnosis of vertebral compression fracture coul be a marker of poor health, although a vertebral compression fracture itself is selom a irect cause of eath. Further investigation is neee to clarify the sequence of health events that result in premature eath of these patients. Despite these limitations, our stuy encompasse a large population of patients with a vertebral compression fracture in all emographic segments who were matche with corresponing controls. This allowe us to etermine the relative mortality risk for this well-controlle group of patients. Although we have emonstrate the substantial negative impact of a vertebral compression fracture on survival, the present stuy i not inclue a comparison of treatment options or efficacy of preventive regimens. n Emun Lau, MS Exponent, Inc., 149 Commonwealth Drive, Menlo Park, CA Kevin Ong, PhD Steven Kurtz, PhD Exponent, Inc., 3401 Market Street, Suite 300, Philaelphia, PA aress for K. Ong: kong@exponent.com Jorana Schmier, MA Exponent, Inc., 1800 Diagonal Roa, Suite 300, Alexanria, VA Av Eiin, PhD Kyphon, Inc., 1221 Crossman Avenue, Sunnyvale, CA References 1. Barlow DH. The report of the Avisory Group on Osteoporosis. Lonon: Department of Health; Riggs BL, Melton LJ 3r. The worlwie problem of osteoporosis: insights affore by epiemiology. Bone. 1995;17(5 Suppl):505S-511S. 3. Cooper C, Atkinson EJ, O Fallon WM, Melton LJ 3r. Incience of clinically iagnose vertebral fractures: a population-base stuy in Rochester, Minnesota, J Bone Miner Res. 1992;7: Cheung G, Chow E, Holen L, Vimar M, Danjoux C, Yee AJ, Connolly R, Finkelstein J. Percutaneous vertebroplasty in patients with intractable pain from osteoporotic or metastatic fractures: a prospective stuy using quality-of-life assessment. Can Assoc Raiol J. 2006;57: Salaffi F, Cimmino MA, Malavolta N, Carotti M, Di Matteo L, Scenoni P, Grassi W; Italian Multicentre Osteoporotic Fracture Stuy Group. The buren of prevalent fractures on health-relate quality of life in postmenopausal women with osteoporosis: the IMOF stuy. J Rheumatol. 2007;34: Ross PD, Davis JW, Epstein RS, Wasnich RD. Pre-existing fractures an bone mass preict vertebral fracture incience in women. Ann Intern Me. 1991;114: Kao DM, Browner WS, Palermo L, Nevitt MC, Genant HK, Cummings SR. Vertebral fractures an mortality in oler women: a prospective stuy. Stuy of Osteoporotic Fractures Research Group. Arch Intern Me. 1999;159: Deyo RA, Cherkin DC, Ciol MA. Aapting a clinical comorbiity inex for use with ICD-9-CM aministrative atabases. J Clin Epiemiol. 1992;45: Murray SB, Bates DW, Ngo L, Ufberg JW, Shapiro NI. Charlson Inex is associate with one-year mortality in emergency epartment patients with suspecte infection. Aca Emerg Me. 2006;13: Center JR, Nguyen TV, Schneier D, Sambrook PN, Eisman JA. Mortality after all major types of osteoporotic fracture in men an women: an observational stuy. Lancet. 1999;353: Johnell O, Kanis JA, Oén A, Sernbo I, Relun-Johnell I, Petterson C, De Laet C, Jönsson B. Mortality after osteoporotic fractures. Osteoporos Int. 2004;15: Hasserius R, Karlsson MK, Jónsson B, Relun-Johnell I, Johnell O. Long-term morbiity an mortality after a clinically iagnose vertebral fracture in the elerly a 12- an 22-year follow-up of 257 patients. Calcif Tissue Int. 2005;76: Barrett J, Losina E, Baron JA, Mahome NN, Wright J, Katz JN. Survival following total hip replacement. J Bone Joint Surg Am. 2005;87: