ARTHRITIS & RHEUMATISM
`Vol. 50, No. 11, November 2004, pp 3516–3525
`DOI 10.1002/art.20627
`© 2004, American College of Rheumatology
`
`The Relationship of Antiresorptive Drug Use to
`Structural Findings and Symptoms of Knee Osteoarthritis
`
`Laura D. Carbone,1 Michael C. Nevitt,2 Kathryn Wildy,3 Karen D. Barrow,1 Fran Harris,2
`David Felson,4 Charles Peterfy,5 Marjolein Visser,6 Tamara B. Harris,7 Benjamin W. E. Wang,1
`and Stephen B. Kritchevsky,8 for the Health, Aging and Body Composition Study
`
`Objective. To examine the cross-sectional associ-
`ation between use of medications that have a bone
`antiresorptive effect (estrogen, raloxifene, and alendro-
`nate) and both the structural features of knee osteoar-
`thritis (OA), assessed by magnetic resonance imaging
`(MRI) and radiography, and the symptoms of knee OA
`in elderly women.
`Methods. Women in the Health, Aging and Body
`Composition Study underwent MRI and radiography of
`the knee if they reported symptoms of knee OA, and
`women without significant knee symptoms were selected
`as controls. MR images of the knee were assessed for
`multiple features of OA using the Whole-Organ MRI
`scoring method, and radiographs were read for Kellgren
`and Lawrence grade and individual features of OA.
`Concurrent medication use and knee symptoms were
`assessed by interview, and knee pain severity was eval-
`
`Supported by grants from the National Institute on Aging
`(N01-AG-6-2010, N01-AG-6-2103, and N01-AG-6-2106).
`1Laura D. Carbone, MD, MS, Karen D. Barrow, MS, Ben-
`jamin W. E. Wang, MD: University of Tennessee Health Sciences
`Center, Memphis; 2Michael C. Nevitt, PhD, Fran Harris, MS: Preven-
`tion Sciences Group, University of California, San Francisco; 3Kathryn
`Wildy, MD: University of Pittsburgh, Pittsburgh, Pennsylvania; 4David
`Felson, MD, MPH: Boston University School of Medicine, Boston,
`Massachusetts; 5Charles Peterfy, MD, PhD: Synarc, Inc., San Fran-
`cisco, California; 6Marjolein Visser, PhD: Vrije University Medical
`Center, Amsterdam, The Netherlands; 7Tamara B. Harris, MD, MS:
`National Institute of Aging, Bethesda, Maryland; 8Stephen B.
`Kritchevsky, PhD: J. Paul Sticht Center on Aging, Wake Forest
`University Health Sciences, Winston-Salem, North Carolina.
`Dr. Carbone has served as a consultant to Eli Lilly, Procter
`and Gamble, Aventis, Wyeth-Ayerst, and Merck. Dr. Nevitt has
`received honoraria from Negma-Lerads. Dr. T. B. Harris has had stock
`ownership in Merck and Shering-Plough. Dr. Wang has served as a
`consultant to Aventis and Merck.
`Address correspondence and reprint requests to Laura D.
`Carbone, MD, MS, University of Tennessee Health Sciences Center,
`956 Court Avenue, Room E336 Coleman Building, Memphis, TN
`38163. E-mail: lcarbone@utmem.edu.
`Submitted for publication March 16, 2004; accepted in revised
`form August 10, 2004.
`
`uated using the Western Ontario and McMaster Uni-
`versities Osteoarthritis Index (WOMAC).
`Results. There were 818 postmenopausal women
`from whom we obtained MR images of the knee and
`data on medication use. Among these women, 214
`(26.2%) were receiving antiresorptive drugs. We found
`no significant association between overall use of antire-
`sorptive drugs and the presence of knee pain and
`radiographic changes of OA of the knee. Use of alen-
`dronate, but not estrogen, was associated with less
`severity of knee pain as assessed by WOMAC scores.
`Both alendronate use and estrogen use were associated
`with significantly less subchondral bone attrition and
`bone marrow edema–like abnormalities in the knee as
`assessed by MRI, as compared with women who had not
`received these medications.
`Conclusion. Elderly women being treated with
`alendronate and estrogen had a significantly decreased
`prevalence of knee OA–related subchondral bone le-
`sions compared with those reporting no use of these
`medications. Alendronate use was also associated with a
`reduction in knee pain according to the WOMAC scores.
`
`Osteoarthritis (OA) of the knee is a common
`condition in the United States and a leading cause of
`disability in elderly people (1–3). Nonoperative thera-
`pies are not curative, and a substantial number of
`affected people ultimately undergo total knee replace-
`ments (4).
`The cartilage degradation that is central to OA is
`accompanied by important periarticular bone abnormal-
`ities, including subchondral bone thickening, deforma-
`tion and cysts, and marginal osteophytosis. A higher rate
`of subchondral bone turnover, as indicated by increased
`uptake of scintigraphic tracer in subarticular bone, is
`associated with more rapid progression of knee OA (5).
`Markers of cartilage degradation, including crosslinked
`
`3516
`
`

`

`ANTIRESORPTIVE DRUG USE AND KNEE OA
`
`3517
`
`C-telopeptide fragments of type II collagen (CTX) (6),
`are elevated in synovial fluid of patients with primary
`knee OA (7), and elevated levels of both CTX and
`urinary N-terminal type I collagen telopeptide, a marker
`of bone resorption (8), have been reported in patients
`with progressive knee OA (9). Knee magnetic resonance
`imaging (MRI) findings that may be related to increased
`bone turnover, subchondral bone marrow edema–like
`lesions, and bone attrition are also very strong predictors
`of structural worsening of knee OA (10–12). Bone
`antiresorptive agents used to prevent or treat osteo-
`porosis, including estrogen, bisphosphonates, calcitonin,
`and raloxifene, may influence the periarticular bone
`changes of OA and could, therefore, have an effect on
`the course of the disease, including the possibility of
`slowing its development and progression (6,13–14). Re-
`cent data also implicate inflammation as playing a
`pivotal role in OA disease progression, chronicity, and
`pain (11,15), and both estrogen (16) and bisphospho-
`nates (17,18) may have antiinflammatory properties.
`The results of epidemiologic studies of estrogen
`use as it relates to OA are conflicting. The majority of
`reports (19–23), although not all (24,25), suggest a
`beneficial effect of estrogen on OA. Studies evaluating
`structural changes of knee OA have found that radio-
`graphic changes are less frequent and less progressive in
`women taking estrogen compared with those not taking
`estrogen (20,26). Interestingly, a large effort is under
`way to test the efficacy of a currently available bisphos-
`phonate as a treatment for knee OA (27,28).
`MRI is one of the most sensitive techniques
`available to assess soft tissue and bone changes in the
`knee (29). Three studies have utilized knee MRI to
`examine the relationship between the use of estrogen
`and cartilage volume in the knee. One cross-sectional
`study demonstrated that estrogen use was associated
`with increased tibial cartilage volume (19); however, a
`more recent longitudinal study did not suggest that tibial
`cartilage volume was different in women receiving estro-
`gen compared with those who had not taken estrogen
`(30). In addition, another study did not find that women
`receiving estrogen had greater patellar cartilage volume
`compared with those who had not received estrogen
`(31). However, no previous study has utilized MRI to
`examine the association of estrogen use with subchon-
`dral bone lesions. Due to the rather recent introduction
`of antiresorptive agents approved for the treatment of
`osteoporosis (including nitrogen-containing bisphospho-
`nates and selective estrogen receptor modulators
`[SERMs]), little is known about their effects on any
`structural or clinical findings of knee OA.
`
`The purpose of this cross-sectional study was to
`examine the association of use of antiresorptive therapy
`with the presence and severity of symptoms of knee
`pain, knee OA radiographic changes, and knee OA–
`related MRI changes in elderly women in the Health,
`Aging and Body Composition (Health ABC) Study
`cohort.
`
`PATIENTS AND METHODS
`
`Subjects. The Health ABC Study is a longitudinal
`study of the factors that contribute to disability in the elderly.
`The Study was approved by the Institutional Review Boards at
`The University of Tennessee and The University of Pittsburgh,
`which are the 2 clinical sites for the Study. To be eligible for
`participation in the Health ABC Study at baseline, participants
`had to report that they had no difficulty walking at least
`one-quarter mile and climbing 10 stairs without resting. A total
`of 3,075 male and female participants were enrolled in Mem-
`phis, Tennessee (n ⫽ 1,548) and in Pittsburgh, Pennsylvania
`(n ⫽ 1,527). Participants in the present analysis were women
`ages 69–81 years who had undergone the year 2 or year 3
`examination of the Health ABC Study, which was conducted
`from July 1998 to July 2000, who underwent MRI or radiog-
`raphy of the knee, and who completed a medication use survey.
`Men were not included in this analysis because there were too
`few men in the antiresorptive user group in years 2 or 3 of the
`Health ABC Study (n ⫽ 9).
`Participants were eligible to obtain bilateral knee
`radiographs and knee MR images at the year 2 or year 3 visits
`if they reported symptoms of OA (defined below) in at least 1
`knee at that visit. In addition, a sample of participants who did
`not meet the study definition for knee OA symptoms were
`chosen at random to obtain knee MR images and radiographs.
`Assessment and definition of knee symptoms. Study
`interviewers assessed knee symptoms at each visit by asking
`participants if they had “pain, aching or stiffness on most days
`for at least one month” at some time during the past year, and
`interviewers also administered a modified Western Ontario
`and McMaster Universities Osteoarthritis Index (WOMAC)
`pain scale (32). The modified WOMAC pain subscale uses a
`5-point Likert scale and asks subjects to rate any knee pain that
`occurs while walking on a flat surface, going up or down stairs,
`at night while in bed, standing upright, getting in or out of a
`chair, and getting in or out of a car. A knee was defined as
`having symptoms of OA if the participant reported pain,
`aching, or stiffness in that knee on most days for at least 1
`month in the past 12 months or if they reported moderate or
`worse knee pain during the last 30 days in association with ⱖ1
`activity listed in the WOMAC pain scale.
`Medication use. As part of their yearly visits, partici-
`pants were asked to bring in to the clinics all of their
`medications,
`including over-the-counter drugs and supple-
`ments that had been taken during the previous 2 weeks.
`Medications were recorded and matched to a dictionary of
`prescription and nonprescription drugs using the Iowa Drug
`Information System (33). This system identifies the ingredients
`of each drug in the dictionary and allows the identification of
`each individual pharmaceutical compound. In addition to the
`
`

`

`3518
`
`CARBONE ET AL
`
`name of the drug, participants were asked to indicate how long
`they had been treated with each drug. The date of the clinic
`visit at which the medication use was recorded was matched to
`the date of the radiograph and the MRI separately.
`The women were included in the antiresorptive drug
`user group if they reported use of estrogen, raloxifene, or any
`bisphosphonate during the 2 weeks prior to the clinic visit.
`Calcitonin was not included in this analysis because only 1
`participant reported its use. None of the participants reported
`use of pamidronate, etidronate, or risedronate in this study;
`therefore, alendronate was the only bisphosphonate included
`in the analysis. Use of calcium supplements, thiazide diuretics,
`and nonsteroidal antiinflammatory drugs (NSAIDs) was also
`ascertained as part of the drug identification process.
`Knee imaging. Bilateral, standing, flexed views of the
`tibiofemoral compartment of the knee joint were obtained
`using the Fixed-Flexion technique (34), and axial (skyline)
`views were obtained of the patellofemoral
`joint. For the
`tibiofemoral compartment, both knees were assessed radio-
`graphically with a posteroanterior projection using a position-
`ing frame (SynaFlexer; Synarc, San Francisco, CA) in order to
`fix knee flexion (between 20° and 30°) and external rotation of
`the feet at 10° for each subject. For the patellofemoral view,
`each knee was imaged separately with the participant in a
`standing position and the limb flexed at 30–40° during weight-
`bearing (35).
`Bilateral MR images of the knee were obtained at each
`clinical center using 1.5T whole-body scanners (General Elec-
`tric Signa, Milwaukee, WI) and standard commercial, unilat-
`eral, circumferential knee coils. The short MRI protocol used
`in the Health ABC Study consisted of 3 sequences: 1) an axial
`T2-weighted fast spin echo (FSE) localizer encompassing the
`entire patella (30 seconds); 2) a sagittal T2-weighted FSE with
`frequency-selective fat suppression, spanning the entire syno-
`vial cavity (4 minutes, 30 seconds); and 3) a coronal T2-
`weighted FSE (4 minutes).
`Assessment of knee images for OA. Radiographs and
`MR images were read centrally. Radiographs of the tibiofemo-
`ral joint were scored for Kellgren and Lawrence severity grade
`of knee OA, and the patellofemoral joint radiographs were
`scored for grade of individual radiographic features (e.g., joint
`space narrowing, osteophytes) based on the Osteoarthritis
`Research Society International Atlas (36). All knee radio-
`graphs obtained at the year 2 and year 3 visits were read at
`Boston University. The presence of tibiofemoral OA in a knee
`was defined as a Kellgren and Lawrence grade of ⱖ2. The
`presence of patellofemoral knee OA was defined as the
`presence of an osteophyte of grade ⱖ2 on a 0–3 scale or the
`presence of moderate to severe joint space narrowing (grade
`ⱖ2 on a 0–3 scale) with co-occurrence of a bone feature in the
`compartment affected (37).
`MR images were evaluated using the multifeature
`semiquantitative Whole-Organ MRI Score (WORMS) method
`(38), conducted by trained readers at the Osteoporosis and
`Arthritis Research Group at the University of California at
`San Francisco (UCSF). Knees were randomly selected for
`reading according to strata defined by the presence and
`absence of knee symptoms.
`Using the WORMS method, 8 separate morphologic
`features of knee OA were evaluated. Five features were
`related to the articular surface: cartilage damage, marginal
`
`osteophytes, subchondral bone marrow abnormality, subchon-
`dral bone cysts, and subchondral bone attrition. Articular
`surface features were scored in up to 15 separate anatomic
`regions of the knee, and the scores were then summed to yield
`total scores for each feature by compartment and for the whole
`knee (38). Three additional articular features that are impor-
`tant to the functional integrity of the knee were also scored:
`meniscal damage (medial, lateral, and total knee), ligament
`damage (total knee), and synovial distention/effusion (total
`knee). All features were scored using ordinal scales that are
`specific to the feature, with higher scores corresponding to
`more severe abnormalities. The interreader reliability between
`3 MRI readers in the Health ABC Study was calculated by
`individual MRI feature, both within compartment and for the
`total knee. Interreader agreement for the WORMS assess-
`ments performed by the Health ABC Study MRI readers on 30
`knees was good to excellent, with intraclass correlation coeffi-
`cients (ICCs) between 0.62 and 0.93. (ICC values between 0
`and 0.40 are interpreted as poor agreement, 0.41 and 0.75 as
`fair to good agreement, and ⱖ0.76 as excellent agreement
`[39]).
`
`For this study, we hypothesized that antiresorptive
`medications would affect cartilage and subchondral bone
`lesions throughout the knee. Therefore, we utilized the total
`knee scores for the features of interest: osteophytes (range
`0–112), subchondral bone marrow abnormality (range 0–45),
`subchondral bone attrition (range 0–42), and cartilage damage
`(range 0–84).
`Measurement of covariates. Bone mineral density
`(BMD) of the total hip was measured using a Hologic QDR
`Model 4500A fan-beam densitometer (Waltham, MA). The
`DXA Quality Assurance Manual for the Health ABC Study
`(UCSF Prevention Sciences Group) was used to standardize
`participant positioning and scan analysis.
`An isokinetic quadriceps strength test was performed
`using a dynamometer (Kin-Com Dynamometer, Chattanooga,
`TN) at visit 2. The maximum torque (moving from 80° to 40°)
`was recorded from averaged curves, with results expressed in
`Nm. There were 151 participants with knee radiographic or
`MRI data who did not or were unable to perform the strength
`test (for example, those with severe knee pain during the test).
`Height was assessed using a Harpenden stadiometer
`(Pembrokeshire, UK). Weight (without shoes) was determined
`using a calibrated balance scale. Body mass index (BMI) was
`calculated by dividing weight (in kg) by height (in m2). Each
`participant’s age, race, history of knee surgery, current smok-
`ing status, self-report of osteoporosis or fracture after the age
`of 45 years, and use of walking device were derived from
`interviewer-administered questionnaire.
`Statistical analysis. We included in these analyses all
`women with data on medication use and either an evaluable
`knee radiograph (n ⫽ 668) or at least 1 knee MRI reading
`completed by October 2002 (n ⫽ 540). Not all of the women
`with an evaluable knee radiograph had a knee MRI reading
`available, nor did all of the women with a knee MRI reading
`have an evaluable knee radiograph available. Women report-
`ing use of multiple antiresorptive agents were excluded, be-
`cause it would not be possible to determine the effect of an
`individual drug on an outcome. There were 8 women taking
`multiple antiresorptive agents who were excluded from the
`radiographic analyses and 5 women taking multiple antiresorp-
`
`

`

`ANTIRESORPTIVE DRUG USE AND KNEE OA
`
`3519
`
`tive agents who were excluded from the MRI analyses. One
`woman taking calcitonin was also excluded.
`Contingency tables were used to assess the statistical
`significance of differences in categorical covariates by antire-
`sorptive use, and analysis of variance was used to assess
`whether there were significant differences in the mean values
`of the continuous covariates by antiresorptive use. We classi-
`fied all women who had knee radiographs by the presence of
`radiographic knee OA and knee symptoms, and compared the
`frequency of antiresorptive use among the resulting 4 sub-
`groups: no knee symptoms and no radiographic OA, no knee
`symptoms with radiographic OA, knee symptoms without
`radiographic OA, and knee symptoms with radiographic OA.
`Logistic regression was used to evaluate the differences in
`antiresorptive drug use among these categories, using the
`group with neither knee symptoms nor radiographic OA as the
`reference group. We then used knees as the unit of analysis to
`compare the MRI feature scores and the WOMAC pain scores
`between groups of knees defined by a subject’s use of antire-
`sorptive drugs.
`Analysis of continuous or near-continuous scores was
`performed using linear regression. We also dichotomized the
`MRI feature scores (0 ⫽ no abnormalities, 1 ⫽ summed
`feature score ⱖ1) and compared the prevalence of any abnor-
`mality by antiresorptive use status using logistic regression. For
`both logistic and linear regression in which knees were the unit
`of analysis, we used generalized estimating equations to adjust
`for interknee correlations. All regression analyses were ad-
`justed for age, race, study site, BMI, use of NSAIDs, use of
`thiazide diuretics, calcium supplementation, BMD of the hip,
`current smoking status, knee extensor strength, self-report of
`history of osteoporosis, self-report of fracture since age 45
`years, and use of a walking device. The latter was not included
`in logistic models of cartilage lesions due to missing data. We
`evaluated possible interactions between antiresorptive use and
`radiographic status, as well as interactions between antiresorp-
`tive use and knee symptoms status, by including interaction
`terms in multivariate models. There were no interactions
`between antiresorptive use and radiographic OA status in their
`association with knee symptoms status and WOMAC scores
`(P ⬎ 0.21 for all), nor between antiresorptive use and knee
`symptoms status in their association with MRI features of OA
`or WOMAC scores (P ⬎ 0.32 for all). Statistical analyses were
`performed using the SAS System for Windows (version 9.1;
`SAS Institute, Cary, NC).
`All of our analyses involving specific drug usage were
`prespecified and preplanned, not post hoc. Our justification
`for this was based on the known different potencies of the
`various antiresorptive agents (SERMs, estrogen, and bisphos-
`phonates) for bone. For example, in a recent head-to-head
`comparison of raloxifene with alendronate (Evista Alendro-
`nate Comparison Trial), it was reported that alendronate had
`a much greater effect on bone biomarkers than did raloxifene
`(40).
`
`RESULTS
`Of the 668 women with radiographs and medica-
`tion use data who were considered eligible for this study,
`there were 504 nonusers (75.4%) and 164 antiresorptive
`
`users (24.6%). Of the antiresorptive users, 125 were
`receiving estrogen, 8 were receiving raloxifene, and 31
`were receiving a bisphosphonate (alendronate only). Of
`the 540 study-eligible women with MRI and medication
`use data, there were 383 nonusers (70.9%) and 157
`antiresorptive users (29.1%) (comprising 114 estrogen
`users, 9 raloxifene users, and 34 bisphosphonate [alen-
`dronate only] users).
`Table 1 shows the selected baseline characteris-
`tics of the combined study population of women with
`knee radiographic and/or MRI reading data (n ⫽ 818)
`by use of antiresorptive drugs. Users of antiresorptive
`therapies, compared with nonusers, were more likely to
`be white, thinner, and reside in Memphis and to take
`calcium supplements. Although knee extensor strength
`was lower in the antiresorptive drug user group com-
`pared with the nonuser group (P ⫽ 0.01), this difference
`disappeared after adjustment for BMI (data not shown).
`There were no significant differences between nonusers
`and users of antiresorptive drugs with respect to age,
`current smoking status, NSAID or thiazide use, total hip
`BMD, frequency of a T score less than or equal to ⫺2.5,
`self-report of osteoporosis, self-report of fracture after
`age 45 years, and use of a walking device. With respect
`to specific antiresorptive drug use, bisphosphonate users
`were older than all other groups (P ⬍ 0.01), had a lower
`BMI than nonusers and estrogen users (P ⬍ 0.01 and
`P ⫽ 0.02, respectively), and had lower BMD of the total
`hip than did nonusers and estrogen users (P ⬍ 0.01 for
`both) (data not shown). Bisphosphonate users also had
`lower knee extensor strength than did nonusers (P ⫽
`0.02); however, this difference was no longer statistically
`significant after adjustment for BMI. Bisphosphonate
`users were not statistically significantly more likely to
`have osteoporosis (a T score less than or equal to ⫺2.5
`at the total hip) than were estrogen or raloxifene users
`(P ⬎ 0.06 compared with both drug use groups).
`Table 2 denotes the association of antiresorptive
`use with the combined presence or absence of knee pain
`and radiographic changes of knee OA. There were no
`significant differences in the proportion of overall anti-
`resorptive use between women classified by the presence
`of knee symptoms and tibiofemoral radiographic OA
`and women classified by knee symptoms and patel-
`lofemoral radiographic OA, after adjustment for covari-
`ates. Results were similar when we classified women
`based on the presence of knee symptoms and radio-
`graphic OA in the tibiofemoral and patellofemoral
`compartments combined (whole-knee OA).
`We also found no association between overall
`antiresorptive use and the presence or absence of knee
`
`

`

`3520
`
`CARBONE ET AL
`
`Table 1. Selected characteristics of study population by use of antiresorptive therapy*
`
`Characteristic
`
`Age, mean ⫾ SD years
`Race, no. (%)
`White
`African American
`Study site, no. (%)
`Memphis
`Pittsburgh
`BMI, mean ⫾ SD kg/m2
`Current smoker, %
`Knee extensor strength, mean ⫾ SD
`maximum torque, Nm
`NSAID use, %
`Thiazide use, %
`BMD of total hip, mean ⫾ SD gm/cm2
`T score ⱕ ⫺2.5, %
`Calcium supplementation, %
`Self-report of osteoporosis, %
`Self-report of fracture after age 45 years, %
`Use of walking device, %
`
`Nonusers
`(n ⫽ 604)
`
`74.78 ⫾ 2.94
`
`253 (41.9)
`351 (58.1)
`
`271 (44.9)
`333 (55.1)
`28.78 ⫾ 5.62
`9.1
`80.35 ⫾ 22.67
`
`48.8
`24.8
`0.817 ⫾ 0.147
`78.5
`21.0
`8.9
`20.4
`7.4
`
`Antiresorptive users
`(n ⫽ 214)
`
`74.81 ⫾ 2.90
`
`158 (73.8)
`56 (26.2)
`
`122 (57.0)
`92 (43.0)
`26.35 ⫾ 4.73
`8.4
`75.53 ⫾ 21.34
`
`55.6
`29.0
`0.788 ⫾ 0.144
`74.8
`33.6
`5.6
`20.7
`4.2
`
`P
`
`0.90
`⬍0.01
`
`⬍0.01
`
`⬍0.01
`0.75
`0.01
`
`0.09
`0.24
`0.19
`0.26
`⬍0.01
`0.13
`0.93
`0.11
`
`* BMI ⫽ body mass index; NSAID ⫽ nonsteroidal antiinflammatory drug; BMD ⫽ bone mineral density.
`
`symptoms (P ⫽ 0.14), nor between overall antiresorptive
`use and the presence or absence of radiographic OA in
`the tibiofemoral (P ⫽ 0.60) or patellofemoral (P ⫽ 0.31)
`compartments or whole knee (P ⫽ 0.31), after adjust-
`ment for covariates (data not shown).
`There was no association between individual
`antiresorptive drug use (estrogen, raloxifene, or bisphos-
`phonates) and the presence or absence of knee symp-
`toms (P ⱖ 0.22 for all drugs), nor between individual
`
`antiresorptive drug use and the presence or absence of
`radiographic OA in the tibiofemoral (P ⱖ 0.90 for all
`drugs) or patellofemoral (P ⱖ 0.12 for all drugs) com-
`partments or whole knee (P ⱖ 0.43 for all drugs), after
`adjustment for covariates. Models were also run follow-
`ing the exclusion of black women from the analyses; this
`did not change any of the radiographic findings (data not
`shown).
`The score ranges for the knee OA features
`
`Table 2. Association between antiresorptive use and the combination of knee pain and radiographic
`evidence of osteoarthritis (OA)
`
`Radiographic OA and knee pain status
`
`Tibiofemoral OA
`No knee pain/no radiographic OA
`No knee pain/radiographic OA
`Knee pain/no radiographic OA
`Knee pain/radiographic OA
`Patellofemoral OA
`No knee pain/no radiographic OA
`No knee pain/radiographic OA
`Knee pain/no radiographic OA
`Knee pain/radiographic OA
`Whole knee OA
`No knee pain/no radiographic OA
`No knee pain/radiographic OA
`Knee pain/no radiographic OA
`Knee pain/radiographic OA
`
`No.
`
`128
`31
`262
`245
`
`135
`23
`311
`198
`
`113
`45
`206
`302
`
`% antiresorptive
`users
`
`P,
`vs. reference*
`
`32.0
`16.1
`24.8
`21.6
`
`31.1
`13.0
`25.1
`20.2
`
`33.6
`15.6
`25.7
`21.5
`
`Reference
`0.63
`0.21
`0.29
`
`Reference
`0.61
`0.21
`0.37
`
`Reference
`0.63
`0.20
`0.31
`
`* Adjusted for age, race, study site, body mass index, use of nonsteroidal antiinflammatory drugs, use of
`thiazides, calcium supplementation, bone mineral density of the total hip, current smoking status, knee
`extensor strength, self-report of osteoporosis, self-report of fracture after age 45 years, and use of walking
`device.
`
`

`

`ANTIRESORPTIVE DRUG USE AND KNEE OA
`
`3521
`
`Table 3. Knee osteoarthritis (OA) magnetic resonance imaging features and WOMAC scores by use and type of antiresorptive therapy*
`
`Knee OA feature (score range)
`
`Bone attrition (0–23)
`Osteophytes (0–85)
`Bone marrow abnormality (0–20)
`Cartilage lesions (0–78)
`WOMAC (0–18)
`
`Nonusers
`(645 knees)
`
`2.0 ⫾ 3.0
`13.5 ⫾ 14.4
`2.6 ⫾ 3.3
`17.5 ⫾ 16.0
`3.3 ⫾ 4.5
`
`Antiresorptive
`users
`(253 knees)†
`
`1.4 ⫾ 2.9
`9.8 ⫾ 14.3
`1.5 ⫾ 2.6
`13.6 ⫾ 14.1
`3.0 ⫾ 4.4
`
`Estrogen
`(178 knees)
`
`1.4 ⫾ 2.7
`9.4 ⫾ 13.2
`1.5 ⫾ 2.4
`12.9 ⫾ 13.7
`2.9 ⫾ 4.4
`
`Antiresorptive agent
`
`Bisphosphonate
`(57 knees)
`
`0.6 ⫾ 1.3‡
`6.9 ⫾ 8.8
`1.0 ⫾ 1.7
`14.1 ⫾ 13.6
`2.9 ⫾ 4.2‡
`
`Raloxifene
`(18 knees)
`
`3.9 ⫾ 6.2
`23.4 ⫾ 26.9
`3.8 ⫾ 5.1
`18.6 ⫾ 18.6
`4.5 ⫾ 4.8
`
`* Values are the mean ⫾ SD score. WOMAC ⫽ Western Ontario and McMaster Universities Osteoarthritis (Index).
`† Estrogen, bisphosphonates, and raloxifene combined.
`‡ P ⬍ 0.05 for linear regression coefficient for bisphosphonate users compared with nonusers, adjusted for age, race, study site, body mass index,
`use of nonsteroidal antiinflammatory drugs, use of thiazides, calcium supplementation, bone mineral density of the total hip, current smoking status,
`knee extensor strength, self-report of osteoporosis, self-report of fracture after age 45 years, and use of walking device.
`
`assessed in our population are shown in Table 3. There
`were no significant differences in WORMS values for
`bone attrition, osteophytes, bone marrow abnormality,
`and cartilage lesions between users of all antiresorptive
`drugs combined compared with nonusers, after adjust-
`ment for covariates (Table 3). When the evaluation was
`focused on specific types of antiresorptive drugs,
`bisphosphonate use was associated with lower WORMS
`values for bone attrition as compared with that in the
`nonuse group (Table 3). There were no differences in
`total cartilage lesions by use of individual antiresorptive
`drugs as compared with nonusers. Among the African
`American group of bisphosphonate users, there were no
`significant differences compared with white bisphospho-
`nate users in any MRI severity score (P ⬎ 0.33 for all
`parameters) (data not shown). WOMAC pain scores
`also did not differ by overall antiresorptive drug use
`(Table 3). However, after adjustment for covariates,
`bisphosphonate use was associated with significantly
`lower WOMAC pain scores compared with nonuse (P ⫽
`0.02) (Table 3).
`When MRI feature scores were dichotomized by
`
`the presence of any knee OA feature abnormality,
`overall antiresorptive use, compared with nonuse, was
`associated with a significantly decreased likelihood (af-
`ter adjustment for covariates) of having subchondral
`bone attrition and subchondral bone marrow abnormal-
`ities (Table 4). When comparing users of specific anti-
`resorptive agents with nonusers of antiresorptives (Table
`4), there was also a significantly decreased likelihood
`(after adjustment for covariates) of having bone attrition
`and bone marrow abnormalities in the estrogen users
`and bone marrow abnormalities in the bisphosphonate
`users. After adjustment for covariates, raloxifene use
`was not significantly associated with any MRI feature.
`Models were also run after exclusion of African Amer-
`ican women from the analyses; bisphosphonate use
`remained significantly associated with fewer bone mar-
`row abnormalities (P ⬍ 0.01), and a trend toward
`significance was observed in estrogen users (P ⫽ 0.06).
`The mean duration of estrogen use was 13.8 years
`(range ⬍1–45 years), and the mean duration of alendro-
`nate and raloxifene use was 1.8 years (range ⬍1–4 years)
`and 0.3 years (range ⬍1–1 years), respectively. To
`
`Table 4. Adjusted odds ratios (95% confidence intervals) for the presence of knee osteoarthritis (OA) magnetic resonance imaging features by use
`and type of antiresorptive therapy*
`
`Knee OA feature
`
`Nonusers
`(645 knees)
`
`Antiresorptive
`users
`(253 knees)
`
`Estrogen
`(178 knees)
`
`Bone attrition (46.5% of knees)
`Osteophyte (80.3% of knees)
`Bone marrow abnormality (59.9% of knees)
`Cartilage lesions (86.4% of knees)
`
`1.00
`1.00
`1.00
`1.00
`
`0.41 (0.21–0.84)†
`0.57 (0.37–2.44)
`0.45 (0.22–0.89)†
`1.32 (0.50–3.53)
`
`0.36 (0.17–0.79)†
`0.94 (0.33–2.63)
`0.48 (0.23–1.00)†
`1.50 (0.48–4.64)
`
`Antiresorptive agent
`
`Bisphosphonate
`(57 knees)
`
`0.37 (0.08–1.71)
`1.48 (0.19–11.76)
`0.11 (0.01–0.89)†
`0.97 (0.18–5.25)
`
`Raloxifene
`(18 knees)
`
`1.09 (0.19–6.19)
`0.35 (0.07–1.77)
`0.87 (0.15–5.07)
`0.70 (0.07–7.18)
`
`* Adjusted for age, race, study site, body mass index, use of nonsteroidal antiinflammatory drugs, use of thiazides, calcium supplementation, bone
`mineral density of the total hip, current smoking status, knee extensor strength, self-report of osteoporosis, self-report of fracture after age 45 years,
`and use of walking device, compared with nonusers. (Cartilage models do not include use of walking device due to missing data.)
`† P ⱕ 0.05 for logistic regression coefficient from adjusted models.
`
`

`

`3522
`
`CARBONE ET AL
`
`determine the effect of duration of use of antiresorptive
`drugs on MRI parameters and WOMAC pain scores,
`duration of use was dichotomized into short-term and
`long-term use, with the median duration of use of all
`antiresorptive drugs (6 years) as the cutoff. Among
`estrogen users, duration of use was not associated with
`the presence of any MRI lesions (P ⬎ 0.14, using logistic
`regression for all parameters) but did affect the severity
`of bone attrition and bone marrow abnormalities by
`MRI, with women whose duration of estrogen use was
`longer (ⱖ6 years) having less bone attrition (P ⫽ 0.05,
`using linear regression) and less bone marrow abnormal-
`ities (P ⬍ 0.01, using linear regression). There was no
`effect of duration of estrogen use on WOMAC scores
`(P ⫽ 0.15). There was too little variability in duration of
`alendronate or raloxifene use to evaluate the effects of
`their long-term use on MRI and WOMAC parameters.
`
`DISCUSSION
`In this cross-sectional analysis of elderly African
`American and white female participants from the Health
`ABC Study, use of bone antiresorptive agents was
`associated with significa