Effects of strength training on the incidence and progression of knee osteoarthritis




Quadriceps weakness is a risk factor for incident knee osteoarthritis (OA). We describe a randomized controlled trial of effects of lower-extremity strength training on incidence and progression of knee OA.


A total of 221 older adults (mean age 69 years) were stratified by sex, presence of radiographic knee OA, and severity of knee pain, and were randomized to strength training (ST) or range-of-motion (ROM) exercises. Subjects exercised 3 times per week (twice at a fitness facility, once at home) for 12 weeks, followed by transition to home-based exercise after 12 months. Assessments of isokinetic lower-extremity strength and highly standardized knee radiographs were obtained at baseline and 30 months.


Subjects in both groups lost lower-extremity strength over 30 months; however, the rate of loss was slower with ST than with ROM. Compared with ROM, ST decreased the mean rate of joint space narrowing (JSN) in osteoarthritic knees by 26% (P = not significant). However, the difference between ST and ROM groups with respect to frequency of knee OA progression in JSN consensus ratings was marginally significant (18% versus 28%; P = 0.094). In knees that were radiographically normal at baseline, JSN >0.50 mm was more common in ST than in ROM (34% versus 19%; P = 0.038). Incident JSN was unrelated to exercise adherence or changes in quadriceps strength or knee pain.


The ST group retained more strength and exhibited less frequent progressive JSN over 30 months than the ROM group. The increase in incident JSN >0.50 mm in ST is unexplained and requires confirmation.


Moderate to severe osteoarthritis (OA) affects more than 22 million American adults between the ages of 25 and 74 years (1), and more than 12% of the population falls within this age range (2). Knee OA is associated with a characteristic pattern of decrements in function involving transfer from a seated or prone position to a standing position, mobility, and activities of daily living involving the lower extremities (3). Guccione et al (4) found that after adjustment for age, sex, and comorbidity, knee OA in the Framingham cohort was the strongest predictor of disability among 10 diseases (including heart disease, stroke, congestive heart failure, hip fracture, depression, diabetes mellitus, and chronic obstructive lung disease) for several activities studied: stair climbing, walking, housekeeping, and carrying bundles.

Risk factors for knee OA include age, female sex, obesity, trauma, and quadriceps (quads) weakness (5–9). Among these, quads weakness may be the most amenable to treatment for the prevention of knee OA. Numerous studies have documented the symptomatic benefits of isometric and dynamic exercise for individuals with knee OA (10–13). However, no studies have investigated whether quads strengthening exercises prevent incident radiographic changes or slow the progression of knee OA. Accordingly, we conducted a 30-month, randomized, attention-controlled trial of the effects of lower-extremity strength training on the incidence and progression of knee OA in older adults.


The procedures, benefits, risks, and associated safeguards in this trial were approved by the Institutional Review Board affiliated with Indiana University–Purdue University Indianapolis.


Subjects were 221 functionally independent adults ≥55 years of age. Of these, 57 were participants in a population-based cohort of older adults recruited for a previous epidemiologic investigation of lower-extremity muscle weakness in knee OA (8, 9). The remaining 164 subjects were recruited for this trial from the at-large adult community. Exclusion criteria included inability to walk without assistance, amputation of either lower extremity, knee or hip replacement, history of stroke, myocardial infarction, congestive heart failure, uncontrollable hypertension, fibromyalgia, rheumatoid arthritis or other systemic connective tissue disease, lower-extremity neuropathy, and severe cognitive impairment.


We conducted a 30-month, randomized, single-blind clinical trial. Based on a screening assessment that included a standing anteroposterior (AP) knee radiograph and administration of the Western Ontario and McMaster Universities OA Index (WOMAC) (14), subjects were stratified into 8 groups on the basis of sex and the presence of radiographic evidence of knee OA and knee pain. Knee OA was considered to be present if one or both knees exhibited grade 2 or higher OA by Kellgren and Lawrence (K/L) criteria (15). Knee pain was considered to be present if the subject reported moderate or greater knee pain in the past month (i.e., a rating of ≥3 on a 5-point Likert scale) for any of the 5 items of the WOMAC pain scale (14).

Within each of 4 strata (OA/pain, OA/no pain, no OA/pain, no OA/no pain), men and women were randomized separately into either the strength training (ST) group or the range-of-motion (ROM; attention control) group. At baseline and at the 30-month followup, a study coordinator who was blinded to the subject's treatment group assignment assessed isokinetic muscle strength, knee pain, knee function, and general health status. Interim WOMAC assessments were performed at months 12, 18, and 24.

Strength training.

The ST group initiated training at the National Institute for Fitness and Sport (NIFS) on the Indiana University–Purdue University Indianapolis campus. For the first 3 months, subjects were asked to train twice each week at NIFS and once each week at home. During the following 3 months (i.e., months 4–6), subjects were asked to train at NIFS at least once per week, with the remaining 2 workouts to be performed at home. During months 7–9, subjects were asked to perform only 2 training sessions per month at NIFS and 3 workouts per week at home. Finally, in months 10–12, subjects were required to train at NIFS only once each month while performing the remainder of workout sessions at home. After the first year of participation in the study, subjects were no longer required to train at NIFS, but were asked to return to NIFS for strength testing and assessment of pain and function every 6 months for the remainder of the study.

Each workout session, whether at NIFS or at home, had the same basic structure: subjects performed a general warm-up of walking for 5 minutes, followed by the resistance training session, which was followed by a 5-minute cool-down. The following exercises were performed on the same resistance training equipment (CYBEX International, Medway, MA) during each session at NIFS: leg presses, leg curls, seated chest presses, and seated back rows. Upper-body exercises were included in both protocols to provide balance in the exercise program. However, the ST sessions prioritized lower-body training by beginning with quads and hamstring exercises. The intensity of exercise was based on a clinical determination of the maximum resistance that could be overcome in sets of 8–10 repetitions. Each workout session consisted of 3 sets of each exercise. Progression to greater resistance levels was implemented when the subject could perform 12 repetitions on the last training set for 2 consecutive workouts.

In the home sessions, the ST group performed similar exercises (i.e., wall squats, standing leg curls, wall push-ups, and seated rows), working the same muscle groups they exercised in the sessions at NIFS, except that elastic bands were used to provide the appropriate degree of resistance. Subjects were introduced to the at-home exercises while they trained at NIFS and were given exercise booklets and a videotape to guide them through their home ST workouts.

A fitness trainer, who was not blinded to treatment group, supervised all exercise sessions at NIFS. At baseline, month 12, and semiannually thereafter, the trainer assessed clinical symptoms of OA (WOMAC knee pain and function scales) and peak isotonic lower-extremity strength (1 repetition maximum [RM] for horizontal leg press and seated hamstring curl on the CYBEX apparatus). These unblinded assessments served to provide data that were used to reinforce correct technique and adherence to the exercise regimens.

Range-of-motion exercises (attention control).

The ROM group served as controls and performed simple movement exercises involving no external loading. Each ROM workout session had the same structure as the ST sessions: subjects warmed up by walking for 5 minutes, followed by flexibility exercises and a 5-minute cool-down. Flexibility exercises (10 repetitions each) targeted the neck, shoulders, trunk, elbows, wrist, hips, knee, and ankles (Table 1). Each ROM exercise session lasted ∼45 minutes. The ROM group also began training at NIFS and gradually made the transition to home-based exercise on the same schedule as the ST group. Subjects in the ROM group were given exercise booklets to guide them through their at-home workout sessions.

Table 1. Flexibility exercises comprising the range-of-motion intervention
Joint(s)Flexibility exercise
NeckCircumduction of head with neck held in slight flexion
 Flexion and extension
 Lateral flexion
ShouldersShoulder shrugs
 Glenohumeral circumduction (progressing small to large movements)
 Horizontal flexion
 Scapular protraction and retraction
ElbowsFlexion and extension
WristsFlexion and extension
HipsCircumduction (both directions)
 Flexion and extension
 Abduction and adduction
KneeStanding flexion and extension
AnklesDorsi and plantar ankle stretches
 Inversion and eversion movements
 Foot circumduction
TrunkStanding trunk rotations
 High reaches
 Lateral flexions
 Upper trunk flexion and extensions

Measurement of isokinetic strength.

Concentric isokinetic strength was assessed at baseline and 30 months using a Kin-Com III dynamometer (Chattecx, Hixon, TN). Isokinetic strength (knee extension and flexion) was determined for each leg, using 2 speeds: 60°/second and 120°/second. Subjects were tested in the seated position with stabilizing straps around the waist and thigh and over the left shoulder. The range of motion was from 20° to 80° of knee flexion (full extension = 0°). Subjects were allowed 5 practice repetitions, each of which progressively accelerated to maximal effort. Then 3–5 maximal attempts were made and the curve demonstrating the largest peak was used to determine maximal isokinetic strength.

Adherence to treatment.

To minimize the dropout rate in this long-term intervention study, several strategies were used to maintain interest and compliance with the program. Attendance at training sessions at NIFS were recorded using a computerized check-in system, which required subjects to gain entry to the exercise area by swiping their NIFS membership card through an electronic card reader. Subjects who missed workouts were contacted by the fitness trainer to identify and resolve barriers to adherence. Procedures used to promote adherence included distribution of a study newsletter and t-shirts identifying the subject as a participant in the study, provision of a buddy system and group training sessions to help those who wanted training partners, and coordination of social gatherings for holidays such as Thanksgiving and Christmas. After month 12, when exercise protocols were fully home based, subjects self reported how often they exercised ≥2 times per week. Self-reported adherence was assessed at months 18, 24, and 30 during the followup assessments of isotonic strength, which are described above.

Clinical outcomes.

Subjects were permitted to take their usual analgesics/nonsteroidal antiinflammatory drugs during the trial. The WOMAC was used to assess knee pain and functional limitation. The WOMAC yields reliable scores that are highly sensitive to changes in pain and function in persons with OA of the knee or hip (14). Subjects rated each knee separately so that WOMAC scores could be generated for knees with and without radiographic OA and for knees with and without moderate pain at baseline (see Procedures).

The Medical Outcome Study (MOS) Short Form General Health Survey (SF-36) was used to measure overall health status (16). The SF-36 has been shown to be a reliable indicator of health-related quality of life for general populations and for individuals with OA (17, 18).

The level of depressive symptoms was measured with the Center for Epidemiologic Studies Depression Scale (CES-D), a reliable 32-item inventory of symptoms of depression (19). CES-D scores ≥16 are a valid indication of clinical depression.

Radiographic outcomes.

Each subject underwent a radiographic examination at baseline and at 30 months, which included a fluoroscopically standardized, semiflexed AP view of each knee (20). Minimum joint space width (JSW) in the medial tibiofemoral compartment was measured manually with a digital caliper and was corrected for magnification based on the projected diameter of a magnification marker (6.35-mm steel ball) that was affixed with tape over the lateral aspect of the head of the fibula. The 95% confidence interval for individual estimates of JSW is ±0.50 mm (21).

The severity of individual radiographic features of OA (joint space narrowing [JSN] and osteophytosis in the tibiofemoral compartment) was rated independently by 2 readers (KDB and SAM) who were blinded to the sequence of the films and to treatment group. Ratings of severity (grades 0–3) were based on exemplars in standard pictorial atlases (22). Differences in ratings between the 2 readers were discussed until consensus was achieved; if consensus could not be reached, a musculoskeletal radiologist was consulted for adjudication, which was required for <1% of the radiographs.

The overall severity of osteophytosis in the knee was expressed as the sum of the ratings of osteophyte severity at 4 locations in the semiflexed AP view: medial femur, lateral femur, medial tibia, and lateral tibia. Because the individual rating scales for osteophytes ranged from 0 to 3, the overall osteophyte score varied from 0 to 12.

Statistical analysis.

Analysis of variance (ANOVA) models were used to compare subjects in the 2 treatment groups with respect to 12-month and 30-month changes in isotonic strength and 30-month changes in SF-36 and CES-D scores. Repeated-measures ANOVA was used for dependent variables measured serially during the study (i.e., change in WOMAC function and pain scores, isokinetic strength). The ANOVAs for pain and strength outcomes, which were measured separately for each knee, included knee as a repeated measure. To evaluate changes in strength by sex, final models of strength outcomes were fit with sex and treatment group interactions to estimate and compare adjusted means and standard errors of strength changes in women and men. Associations between adherence to the exercise protocol and isokinetic strength at the end of the study were evaluated with Pearson's correlation coefficients.

Comparisons between treatment groups with respect to radiographic outcomes were performed separately for knees with and without radiographic evidence of OA at baseline (i.e., K/L grade 2–4 and K/L grade 0–1, respectively). Knees with grade 4 OA at baseline were omitted from group comparisons. Repeated-measures ANOVA was used to compare treatment groups with respect to mean JSN (i.e., a continuous variable representing loss of articular cartilage thickness) over 30 months. Logistic regression models were used to compare treatment groups with respect to the frequency with which knees exhibited loss of JSW beyond the margin of measurement error (>0.50 mm) or were rated to have an increase in the severity grade of individual radiographic features of OA (i.e., JSN, osteophytosis) over the 30-month interval. Generalized estimating equations were used to account for the correlation between knees within subjects.

All models were adjusted for sex, age, body mass index (BMI), the presence of OA and/or joint pain, and the baseline value of the outcome. Two-way, 3-way, and 4-way interactions (where appropriate) between presence of OA, pain, time, and treatment group were examined for significance in all models and removed in a backwards manner when indicated. In the event of significant interaction terms involving treatment group and OA or pain, separate models were run accordingly.


Demographic, clinical, and radiographic characteristics of the subjects are shown in Table 2. The stratified randomization scheme produced treatment groups that were equivalent at baseline with respect to sex, age, race, BMI, general health status, radiographic and symptomatic severity of knee OA, and lower-extremity strength. The mean baseline CES-D score was higher among women than among men (6.3 versus 4.1, P = 0.009). Isokinetic extensor (quads) strength was greater at baseline in men who were radiographically normal at baseline than in men with radiographic knee OA (P < 0.05), irrespective of the presence of moderate knee pain. In women, isokinetic strength (extension and flexion at both speeds) was significantly greater at baseline in knees without pain than in those with pain (P < 0.05), irrespective of the presence of radiographic OA.

Table 2. Demographic, clinical, and radiographic characteristics of subjects at baseline*
VariableStrength training (n = 113)Range of motion (n = 108)
  • *

    Values are the mean ± SD unless otherwise indicated. The treatment groups did not differ significantly at baseline with respect to any of the variables shown. OA = osteoarthritis; JSW = joint space width; K/L = Kellgren/Lawrence criteria; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; SF-36 = Short Form 36; CES-D = Center for Epidemiologic Studies Depression Scale.

  • In the knee with radiographically more severe OA.

  • In the knee with more severe pain.

Female sex, no. (%)64 (57)64 (59)
Age, years69.4 ± 8.068.6 ± 7.5
White race, no. (%)103 (93)97 (91)
Body mass index, kg/m229.6 ± 5.629.0 ± 5.4
Overall severity of knee OA, no. (%)  
 Grade 048 (42)31 (29)
 Grade 18 (7)14 (13)
 Grade 28 (7)23 (22)
 Grade 332 (28)28 (26)
 Grade 417 (15)11 (10)
Medial JSW, mm (least squares mean ± SE)  
 K/L grade 0–1 OA (n = 254 knees)4.11 ± 0.134.00 ± 0.13
 K/L grade 2–4 OA (n = 182 knees)3.01 ± 0.153.31 ± 0.14
 Knee pain, range 5–257.9 ± 3.97.9 ± 3.6
 Functional limitation, range 17–8528.1 ± 12.529.0 ± 12.6
SF-36 General Health Survey  
 Physical Component Scale, range 0–10045.3 ± 9.244.8 ± 10.3
 Mental Component Scale, range 0–10056.8 ± 6.557.5 ± 7.4
CES-D, range 0–325.4 ± 6.05.4 ± 6.2
Isokinetic strength, Nm (least squares mean ± SE)  
 K/L grade 0–1 OA (n = 252 knees)  
  Extension at 60°/second111.7 ± 4.3112.9 ± 4.3
  Extension at 120°/second83.8 ± 3.788.5 ± 3.7
  Flexion at 60°/second51.5 ± 2.049.8 ± 2.0
  Flexion at 120°/second47.4 ± 2.046.5 ± 2.0
 K/L grade 2–4 OA (n = 182 knees)  
  Extension at 60°/second100.5 ± 4.796.9 ± 4.6
  Extension at 120°/second79.2 ± 4.075.5 ± 3.9
  Flexion at 60°/second47.7 ± 2.248.2 ± 2.1
  Flexion at 120°/second47.2 ± 2.246.5 ± 2.1

Of the 221 subjects, 154 (70%) completed the study per protocol (i.e., underwent followup assessments at months 12, 18, 24, and 30). The most common category of reasons for discontinuation (31%) was the burden of participation, which included time and travel constraints. Dropouts were 11–18% weaker at baseline and attended 33% fewer exercise sessions in the first 12 weeks than subjects who completed the study. There were also more dropouts in the ST group than in the ROM group (36% versus 24%, P < 0.05). Only 1 subject discontinued prematurely because of an adverse effect of strength training (i.e., increased knee pain). However, in accordance with the principles of intent-to-treat, we attempted to evaluate all randomized subjects at 30 months with respect to isokinetic lower-extremity strength and radiographic severity of knee OA. A total of 174 subjects (79%) underwent the 30-month assessment.

Adherence to the exercise regimen.

The measures used to support adherence to exercise regimens were only moderately successful. Subjects in both the ST and ROM treatment groups only attended approximately half of the 24 exercise sessions scheduled during the first 12 weeks of the study (49% and 46%, respectively; P = 0.453). With renewed effort by the fitness trainer, adherence rates increased somewhat in both treatment groups over the initial 12-month interval during which facility-based workouts were scheduled (59% in the ST group and 64% in the ROM group; P = 0.397). Men participated in more exercise sessions than women during both intervals (P = 0.003 at 12 weeks, P = 0.007 at 12 months). Self-report adherence to home-based exercise workouts over months 13–30 was similar in frequency to that measured during months 1–12 by the electronic surveillance system (ST 56%, ROM 62%).

Isotonic strength.

Changes in unblinded measurements of isotonic lower-extremity strength, relative to baseline, are shown in Figure 1. Twelve months after initiation of the interventions, increases in unblinded measurements of peak isotonic quads strength (1 RM) were not apparent in women, although men in the ST group and ROM group showed increases in isotonic quads strength (mean 11.1% and 8.3% relative to baseline, respectively); the difference, however, was not statistically significant. In contrast to measurements of change in quads strength, assessment of isotonic hamstring strength demonstrated 12-month increases in both women and men in the ST group (mean 6.3% and 11.8%, respectively), which were significantly greater than those seen in women and men in the ROM group (−0.7% and 8.5%, respectively; P = 0.021). However, at the followup assessments at months 18, 24, and 30, no differences between treatment groups were apparent with respect to isotonic quads or hamstring strength.

Figure 1.

Percent change in peak isotonic quadriceps and hamstring strength (1 repetition maximum), relative to baseline, by sex and treatment group. ST = strength training group; ROM = range-of-motion exercise group.

Isokinetic strength.

Kin-Com assessment revealed overall decreases in isokinetic strength in women and men in both treatment groups (Table 3). However, the loss of concentric quads strength at 120°/second in the ST group was smaller than that seen in the ROM group (−18.9 Nm versus −22.1 Nm in women, −6.1 Nm versus −12.1 Nm in men); the difference was marginally significant for women and men combined (P = 0.090). Subjects with knee OA lost more strength than subjects without knee OA (mean change −19.7 Nm versus −14.6 Nm of extension at 60°/second; P = 0.041). Furthermore, subjects with moderate or more severe knee pain at baseline lost quads strength to a greater degree than asymptomatic subjects (−17.6 Nm versus −12.0 Nm of extension at 120°/second; P = 0.040).

Table 3. Least squares mean change in isokinetic concentric strength in both knees at 30 months, relative to baseline, in women and men by treatment group*
Isokinetic strength, NmWomenMenP
Strength trainingRange of motionStrength trainingRange of motion
  • *

    Values are the mean ± SE unless otherwise indicated.

  • Least squares mean from repeated-measures analysis of changes in both knees.

  • P value for difference between treatment groups, adjusted for baseline strength, age, sex, body mass index, knee, presence of osteoarthritis, and pain.

Extension at 60°/second−23.8 ± 3.0−24.7 ± 2.8−7.1 ± 3.6−13.1 ± 3.30.260
Extension at 120°/second−18.9 ± 2.9−22.1 ± 2.7−6.1 ± 3.4−12.1 ± 3.30.090
Flexion at 60°/second−12.7 ± 1.7−14.0 ± 1.6−1.1 ± 2.1−6.2 ± 2.00.066
Flexion at 120°/second−12.0 ± 1.7−14.9 ± 1.7−3.2 ± 2.2−5.5 ± 2.10.096

Loss of isokinetic hamstring strength also was observed in both treatment groups (Table 3). However, the ST group again exhibited a slightly slower rate of loss of hamstring strength than the ROM group (flexion at 60°/second: −12.7 Nm versus −14.0 Nm for women, −1.1 Nm versus −6.2 Nm for men; P = 0.063 for men and women combined).

Isokinetic strength at 30 months in the ST group was related to adherence to the facility-based exercise schedule. For example, in assessments of the right lower extremity, the correlation between the number of exercise sessions attended and isokinetic strength at 60°/second was 0.32 for extension and 0.37 for flexion (P < 0.01 for both). Nearly identical results were seen in data for the left lower extremity (r = 0.35 and 0.32, respectively; P < 0.01 for both). In contrast, the frequency of attendance at facility-based exercise sessions by subjects in the ROM group was unrelated to isokinetic lower-extremity strength at month 30 (all r ≤ 0.13; P = not significant).

Radiographic indicators.

The effects of strength training on radiographic indicators of the severity of knee OA are shown in Table 4. Analyses of progressive radiographic changes in OA were performed on knees with K/L grade 2–3 OA at baseline, and analyses of incident OA were performed on those with K/L grade 0–1 OA at baseline.

Table 4. Changes in 30-month radiographic outcomes, relative to baseline, by radiographic severity of osteoarthritis at baseline and treatment group*
Radiographic outcomeK/L grade 2–3K/L grade 0–1
Strength training (n = 45)Range of motion (n = 60)PStrength training (n = 105)Range of motion (n = 89)P
  • *

    Values are the number (percentage) unless otherwise indicated. K/L = Kellgren/Lawrence criteria; JSN = joint space narrowing.

  • Adjusted for age, sex, body mass index, knee, baseline joint space width, and knee pain.

Minimum medial JSN, mm (mean ± SE)0.34 ± 0.110.54 ± 0.090.1360.34 ± 0.060.29 ± 0.060.429
JSN > 0.50 mm19 (42)24 (41)0.85836 (34)17 (19)0.038
Increase in JSN grade8 (18)17 (28)0.09411 (10)12 (13)0.498
Increase in osteophyte grade16 (36)20 (33)0.43713 (12)10 (11)0.990

In subjects with established knee OA at baseline, the mean loss of JSW in the ST group was 37% less than that in the ROM group (0.34 mm versus 0.54 mm); however, this difference was not significant. The frequency of loss of JSW beyond the margin of measurement error (i.e., >0.50 mm) was similar in the osteoarthritic knees of both treatment groups. Progression of medial or lateral tibiofemoral JSN in blinded semiquantitative ratings occurred less often in the ST group than in the ROM group (18% versus 28%); the difference was marginally significant (P = 0.094). However, similar proportions of the 2 groups (33–36%) exhibited an increase in the cumulative osteophyte score.

In knees that were radiographically normal at baseline in the standing AP view (i.e., K/L grade 0–1), the mean loss of JSW at 30 months was slightly greater in the ST group than in the ROM group (0.34 mm versus 0.29 mm; P = 0.429). However, the frequency of loss of JSW beyond the 0.50-mm margin of measurement error in these knees was significantly greater in the ST group than in the ROM group (34% versus 19%; P = 0.038). The treatment groups did not differ with respect to the frequency of incident JSN or incident osteophytosis.

Clinical measures.

The repeated-measures analysis of WOMAC pain scores yielded a significant treatment-group × OA × time interaction (P = 0.033), necessitating separate longitudinal analyses of pain in knees with and without OA (Table 5). However, in neither analysis did strength training have a significant effect on changes in knee pain.

Table 5. Change in WOMAC pain and functional limitation scores relative to baseline*
WOMAC scoreStrength trainingRange of motionP
  • *

    Values are the least squares mean ± SE unless otherwise indicated. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; K/L = Kellgren/Lawrence criteria; OA = osteoarthritis.

  • P values for comparisons of treatment group were adjusted for age, body mass index, sex, time, the presence of OA and knee pain at baseline, and the baseline value of the outcome. Due to a significant treatment-group × time interaction, group comparisons with respect to change in WOMAC functional limitation scores are presented for each followup assessment.

Knee pain   
 K/L grade 2–4 OA0.9 ± 0.41.1 ± 0.30.323
 K/L grade 0–1 OA0.8 ± 0.31.2 ± 0.30.255
Functional limitation   
 Month 124.6 ± 1.03.0 ± 0.90.226
 Month 183.7 ± 1.04.4 ± 0.90.641
 Month 243.7 ± 1.05.6 ± 1.00.161
 Month 301.7 ± 0.93.9 ± 0.90.088

The repeated-measures analysis of the WOMAC functional limitation scores revealed a significant treatment-group × time interaction (P = 0.014). Functional limitation scores increased over time in the ROM group, peaking at month 24 before abating somewhat in the final 6 months of the study (Table 5). In the ST group, functional limitation scores peaked at month 12 and then declined throughout the remainder of the trial. By month 30, both treatment groups still had higher mean WOMAC functional limitation scores than at the onset of the study, indicating poorer function. However, a trend toward better function in the ST group than in the ROM group was observed (P = 0.088).

The mean of the 30-month change scores for the MOS SF-36 General Health Survey are shown in Table 6. The mean Physical Component Scale score decreased slightly more over 30 months in the ROM group than in the ST group, but the difference was not significant. For the SF-36 Mental Component Score (MCS), a significant treatment-group × OA interaction (P = 0.007) demonstrated that the direction of effect of strength training was dependent on the presence of OA. Specifically, in subjects with knee OA at baseline, the mean decrease in MCS scores (indicating a decrease in emotional health status) was significantly larger in the ROM group than in the ST group (−1.6 versus −0.4; P = 0.042). However, the opposite was true of MCS scores in subjects with no knee OA at baseline: the mean decrease in the ST group was significantly greater than that in the ROM group (−5.0 versus −0.4; P = 0.004). The decline of MCS scores in the ST group was not an indicator of incident cartilage damage; the mean ± SD decrease in MCS scores among subjects in the ST group who exhibited JSN >0.50 mm was unremarkable (−2.5 ± 1.3).

Table 6. Thirty-month changes in general health status and depression scores, relative to baseline, by treatment group*
 Strength training (n = 82)Range of motion (n = 80)P
  • *

    Values are the least squares mean ± SE unless otherwise indicated. SF-36 = Short Form 36; OA = osteoarthritis; CES-D = Center for Epidemiologic Studies Depression Scale.

  • P values for treatment effects were adjusted for age, body mass index, sex, the presence of OA and knee pain at baseline, and the baseline value of the outcome.

SF-36 Physical Component Scale−1.6 ± 0.9−3.0 ± 0.90.254
SF-36 Mental Component Scale   
 Subjects with knee OA at baseline−0.4 ± 1.1−1.6 ± 1.00.042
 Subjects without knee OA at baseline−5.0 ± 1.2−0.4 ± 1.30.004
CES-D2.7 ± 0.62.4 ± 0.60.648

The mean of the baseline CES-D scores in both treatment groups (5.4 on a scale of 0–32) indicated a very low level of depressive symptomatology (Table 2). Despite the apparent decline in emotional health status reflected in the SF-36 MCS scores of the ST group relative to the ROM group (see above), CES-D scores worsened only slightly in both treatment groups over the 30-month trial (Table 6). The difference between treatment groups was not statistically significant, nor was change in CES-D scores related to incident JSN (data not shown).


In subjects with knee OA, pain is typically increased by load bearing and relieved by rest. Quads weakness and, often, obvious quads atrophy are common and have been attributed to disuse atrophy of the muscle as the patient minimizes painful weight-bearing activities. In addition, however, quads weakness may precede and serve as a risk factor for incident radiographic changes of knee OA (8, 9).

Decreases in muscle strength are associated with losses in daily physical function (23–25). Fortunately, the trainability of muscles is retained in the older adult, and strength decrements normally associated with aging can be diminished with resistance training (26). Resistance exercise has consistently been shown to maintain or increase muscle mass (27, 28) and to improve strength and power (27–31). Other benefits include improvement in bone mineral density (32), decreased risk of falling (33, 34), increased walking speed (30, 35), better balance (32, 35), and increased stair-climbing ability (36).

Ample evidence exists to demonstrate that lower-extremity exercise also improves symptoms in persons with knee OA (37, 38). Types of exercise regimens that have been found to be effective in the control of pain and maintenance of function in knee OA include aerobic exercise (39–43) and strength training (10–13). Strength training programs have included various forms of isometric and dynamic exercise, the latter using resistance training with elastic bands and isokinetic dynamometers. Although gains in strength and symptomatic benefit have been observed over intervals ranging from 8 weeks to 18 months, there is no evidence that increasing quads strength can either prevent knee OA or delay its progression.

In the present trial, we evaluated the effects of lower-extremity strength training on the incidence and progression of knee OA. In light of several previous positive studies in this area (10–13), the present study's failure to demonstrate gains in isokinetic quadriceps strength in the ST group is difficult to explain. Adherence to the exercise programs was only moderate during the first 12 months of the trial, although it increased slightly over the following 18 months.

Even though strength training did not result in an increase in isokinetic quadriceps strength in the present study, it slowed the rate of loss of isokinetic hamstring strength over 30 months in comparison with ROM exercises. The lack of interim Kin-Com assessments (e.g., at 12 months, when the facility-based exercise sessions ended) limited our ability to detect short-term effects of strength training. However, this trial was designed with a primary focus on radiographic progression of knee OA, and at the time 30 months was considered to be the minimum duration of treatment necessary for the accrual of a significant difference between treatment groups with respect to the rate of JSN. Given the absence of concurrent data on radiographic outcomes, the value of interim assessments of isokinetic strength would have been diminished. This limitation was offset somewhat by semiannual isotonic strength testing that was used to monitor progress and motivate subjects to continue with the exercise regimen at home. Isotonic strength increased over the duration of the study, primarily among men in both treatment groups (Figure 1). However, because the isotonic strength measurements were not blinded to treatment group, the validity of this observation may be questioned.

The apparent contradiction between the results of isotonic and isokinetic strength testing (i.e., mean gains in strength by the former and losses of strength by the latter) may be due, in part, to the fact that the specific maneuvers performed for assessment of isotonic strength more closely matched the exercises performed by the ST group than those required for isokinetic testing. Also, the isotonic strength measurements were obtained with the same equipment that the subject used for exercising. Therefore, differences in the results of isotonic testing and isokinetic testing may have been due to strength specificity, i.e., the largest gains in strength were detected with the mode of testing that most closely resembled the mode of training (44). If the mode of testing is not specific to the mode of training, the ability to detect increases in strength may be substantially reduced and may even be nonexistent (45, 46). In any event, even though the unblinded assessment of isotonic strength may have been biased in favor of showing larger overall gains in strength in the ST group, differences between the 2 treatment groups with respect to changes in isokinetic strength were not subject to such bias.

Given that half of the subjects who entered the present trial did not have knee pain at baseline and that the mean WOMAC pain score at baseline was only 7.9 on a scale of 5–25, it is not surprising that a decrease in knee pain was not seen in either treatment group. Pain scores increased slightly throughout 30 months in both treatment groups, but the difference was not significant, regardless of the severity of knee pain at baseline. In contrast, a trend toward better function in the ST group than in the ROM group began to emerge in the final 6 months of the trial.

It should be emphasized that the main objective of this trial was to examine the effects of strength training on incident and progressive radiographic changes of knee OA. To that end, a highly standardized protocol for fluoroscopically assisted positioning of the knee was used (20). Fluoroscopically assisted positioning of the knee in serial examinations virtually eliminates false changes in the radiographic joint space due to uncontrolled variations in knee flexion or rotation (47) and subtle changes in weight bearing due to knee pain (48).

Despite the lack of significant differences between the ST and ROM groups with respect to changes in isokinetic quads strength, some treatment group differences in radiographic outcomes were observed. The rate of JSN in the knees of subjects in the ST group who had radiographic evidence of OA at baseline was 37% slower than that in the ROM group. This effect was not significant. However, the magnitude of the 30-month difference between treatment groups with respect to mean JSN (0.13 mm) was as large as that observed over the same interval in our recently published placebo-controlled trial of the structure-modifying effects of doxycycline (0.20 mm), which used the identical highly standardized protocol for knee radiography (49). However, although strength training did not result in a significant reduction in the rate of JSN in subjects with knee OA, as reflected in quantitative measurements of JSW, data from consensus ratings of JSN suggest that the frequency of progression of knee OA may have been decreased by 35% by ST exercise relative to ROM exercise (18% versus 28%; P = 0.090).

At the start of this trial there was no precedent to suggest how large and variable the effects of strength training on radiographic progression of OA might be. Based on JSN data in the ST and ROM groups in the present trial, we can now estimate that data from 404 knees with OA (202 subjects if each subject contributes data from 2 knees) would have been necessary to attain 80% power to detect a 26% difference between the ST and ROM groups with respect to 30-month JSN. Although >200 subjects were randomized to treatment conditions in the present study, the resulting number of knees with K/L grade 2–3 OA at baseline in these subjects was only ∼25% of the number needed for nominal power.

In comparison with trends indicating beneficial effects of strength training on the progression of established radiographic knee OA suggested by these results, the negative effect of strength training on incident radiographic changes was unexpected. In knees with K/L grade 0–1 at baseline, the frequency of incident osteophytes and the rate of JSN were very similar in the 2 treatment groups. However, instances of JSN exceeding 0.50 mm (i.e., the margin of measurement error) were 79% more common in the ST group than in the ROM group (Table 4). Sharma et al (50) reported that quads strength was a significant risk factor for radiographic progression of OA in malaligned and lax OA knees, implying that strength training may lead to damage of at-risk OA joints. Although we did not characterize the knees in this study with respect to malalignment and laxity, it must be pointed out that strength training was associated with an increase in the frequency of JSN >0.50 mm only in knees that were radiographically normal at baseline. Moreover, neither compliance with the strength training regimen nor the observed changes in quads or hamstring strength was associated with incident JSN >0.50 mm.

It is difficult to accept, therefore, that the strength training used in this study may have been harmful for adults without knee OA. Otterness et al (51) demonstrated that in hamsters, daily wheel-running exercise resulted in a higher cartilage proteoglycan content and prevented cartilage degeneration, in comparison with sedentary living, and concluded that exercise was required for maintenance of healthy cartilage. Palmoski et al (52) reported that immobilization of the hind limb of a normal dog in an orthopedic cast led to marked changes of articular cartilage atrophy, with significant decreases in cartilage thickness, proteoglycan content and concentration, and net rate of proteoglycan synthesis. Notably, these changes were fully reversible if the cast was removed within a reasonable period. Furthermore, the data suggest that the changes of cartilage atrophy were mediated primarily by a reduction in the strength of contraction of the periarticular muscles (i.e., quads and hamstrings) that normally contract to stabilize the knee joint during gait, rather than by a decrease in range of motion of the joint (53). In a recent preliminary report of a placebo-controlled study of subjects who had undergone partial medial meniscectomy and were at risk for knee OA, Dahlberg and Roos (54) found that a 4-month program of progressive aerobic and weight-bearing exercise resulted in an increase in the proteoglycan content of the articular cartilage in the knee, as shown by delayed gadolinium-enhanced magnetic resonance imaging (MRI).

It is possible that the increased frequency of incident medial compartment JSN associated with strength training in this study reflects, in part, meniscal subluxation, which MRI studies have shown to precede the thinning of knee articular cartilage and to account for a significant proportion of cases of early radiographic JSN (55, 56). In any event, this finding requires confirmation in future trials of resistance exercise programs for older adults, which should include serial standardized radiographic or MRI examinations to monitor possible adverse effects of lower-extremity resistance training on articular cartilage in the knee.