aParticipant may have been ineligible for more than 1 reason.
bPersons excluded due to x-ray for multiple reasons.
cStaff considered participation a danger to themselves or others.
dA stratified block randomization method, stratified by body mass index (BMI) and sex, was used to assign all eligible persons to 1 of 3 intervention groups.
The middle line in the plot boxes represents the median values; the X, the mean values; and the boxes, the interquartile range. The whiskers extend to the most extreme observed values within 1.5 × the interquartile range of the nearer quartile, and the dots represent observed values outside the range. WOMAC indicates Western Ontario and McMaster Universities Osteoarthritic Index.
B, A knee compressive force of 872 N was equivalent to mean body weight. One pound (0.45 kg) of force is equivalent to 4.45 N of force.
eTable 1 Comparison of Baseline Values of Completers and Noncompleters
eTable 2. Pairwise Between-Group Differences At 18-Month Follow-Up For Pre-Specified Primary And Secondary Outcomes, And Post Hoc Outcome From Multiply Imputed Data
eTable 3. Primary Outcome Analyses Mean Values Adjusted For Baseline Frontal Plane Alignment Or Patellofemoral OA
eTable 4. Pre-Specified Secondary Outcomes At 6-month Follow-Up
eTable 5. Unadjusted Changes In Pre-Specified Secondary And Post-Hoc Outcomes At 18-Month Follow-Up
eTable 6. Serious Adverse Events (SAE)
eTable 7. Most Commonly Occurring Adverse Events (AEs) By Group Assignment
eTable 8. Adverse Events Related To The Study By Group
eFigure 1. Unadjusted Mean WOMAC Function, and Unadjusted Mean 6-Minute Walk Distance
eFigure 2. Unadjusted Mean Knee Extensor Strength, Unadjusted Mean Knee Flexor Strength, and Unadjusted Mean Hip Abductor Strength
eFigure 3. Unadjusted Mean Medial Tibiofemoral Joint Space Width, and Unadjusted Mean Thigh Muscle Volume, Unadjusted Mean Thigh Fat Volume, and Unadjusted Mean Log IL-6 Levels
eFigure 4. Percent Pain Medication Use By Group
Data Sharing Statement
Customize your JAMA Network experience by selecting one or more topics from the list below.
Messier SP, Mihalko SL, Beavers DP, et al. Effect of High-Intensity Strength Training on Knee Pain and Knee Joint Compressive Forces Among Adults With Knee Osteoarthritis: The START Randomized Clinical Trial. JAMA. 2021;325(7):646–657. doi:10.1001/jama.2021.0411
Is high-intensity strength training more effective than low-intensity strength training and attention control for the outcomes of knee pain and knee joint compressive forces in participants with knee osteoarthritis?
In this randomized clinical trial involving 377 participants with knee osteoarthritis, high-intensity strength training, compared with low-intensity strength training and an attention control, did not significantly improve knee pain or knee joint compressive forces at 18 months.
These findings do not support the use of high-intensity strength training over low-intensity strength training or attention control in adults with knee osteoarthritis.
Thigh muscle weakness is associated with knee discomfort and osteoarthritis disease progression. Little is known about the efficacy of high-intensity strength training in patients with knee osteoarthritis or whether it may worsen knee symptoms.
To determine whether high-intensity strength training reduces knee pain and knee joint compressive forces more than low-intensity strength training and more than attention control in patients with knee osteoarthritis.
Design, Setting, and Participants
Assessor-blinded randomized clinical trial conducted at a university research center in North Carolina that included 377 community-dwelling adults (≥50 years) with body mass index (BMI) ranging from 20 to 45 and with knee pain and radiographic knee osteoarthritis. Enrollment occurred between July 2012 and February 2016, and follow-up was completed September 2017.
Participants were randomized to high-intensity strength training (n = 127), low-intensity strength training (n = 126), or attention control (n = 124).
Main Outcomes and Measures
Primary outcomes at the 18-month follow-up were Western Ontario McMaster Universities Osteoarthritis Index (WOMAC) knee pain (0 best-20 worst; minimally clinically important difference [MCID, 2]) and knee joint compressive force, defined as the maximal tibiofemoral contact force exerted along the long axis of the tibia during walking (MCID, unknown).
Among 377 randomized participants (mean age, 65 years; 151 women [40%]), 320 (85%) completed the trial. Mean adjusted (sex, baseline BMI, baseline outcome values) WOMAC pain scores at the 18-month follow-up were not statistically significantly different between the high-intensity group and the control group (5.1 vs 4.9; adjusted difference, 0.2; 95% CI, −0.6 to 1.1; P = .61) or between the high-intensity and low-intensity groups (5.1 vs 4.4; adjusted difference, 0.7; 95% CI, −0.1 to 1.6; P = .08). Mean knee joint compressive forces were not statistically significantly different between the high-intensity group and the control group (2453 N vs 2512 N; adjusted difference, −58; 95% CI, −282 to 165 N; P = .61), or between the high-intensity and low-intensity groups (2453 N vs 2475 N; adjusted difference, −21; 95% CI, −235 to 193 N; P = .85). There were 87 nonserious adverse events (high-intensity, 53; low-intensity, 30; control, 4) and 13 serious adverse events unrelated to the study (high-intensity, 5; low-intensity, 3; control, 5).
Conclusions and Relevance
Among patients with knee osteoarthritis, high-intensity strength training compared with low-intensity strength training or an attention control did not significantly reduce knee pain or knee joint compressive forces at 18 months. The findings do not support the use of high-intensity strength training over low-intensity strength training or an attention control in adults with knee osteoarthritis.
ClinicalTrials.gov Identifier: NCT01489462
Osteoarthritis is the most common form of arthritis and a leading cause of disability among adults.1 A report from 2012 estimated that knee osteoarthritis affected more than 250 million people worldwide.2 Clinical guidelines recommend strength training for patients with knee osteoarthritis based on clinical trial evidence of effectiveness.3 Improved strength mediates pain relief,4 enhances psychological well-being,5 maintains cartilage integrity in animal models,6,7 and may increase the shock absorbing capability of lower extremity muscles during walking.8 Limitations of prior randomized trials of strength training for patients with osteoarthritis include the use of training intensities below those recommended by practice guidelines9-13 and use of short-duration interventions, lasting between 6 and 24 weeks.11,12
High-intensity strength training may be detrimental for knee osteoarthritis symptoms due to the greater contact forces exerted on the joint.14 However, a prior study suggested that short-term, high-intensity strength training was safe and well tolerated by older adults with knee osteoarthritis.15 Improving strength gains with intense exercise over the long-term could improve clinical outcomes. The objective of this study was to determine whether high-intensity strength training lowered pain levels and reduced knee joint compressive forces, defined as the maximal tibiofemoral contact force exerted along the long axis of the tibia during walking, compared with low-intensity strength training and attention control.
The Strength Training for Arthritis Trial (START) was an assessor-blinded, single-center (Wake Forest University, Winston-Salem, North Carolina), randomized clinical trial with 3 parallel groups followed up for 18 months. The trial protocol appears in Supplement 1. The Human Subjects Committee of Wake Forest Health Sciences approved this protocol. Participants provided written informed consent.
The study included ambulatory, community-dwelling adults, 50 years or older with self-reported disability due to knee osteoarthritis (Figure 1). Enrollment occurred between July 2012 and February 2016. Inclusion criteria were (1) mild to moderate radiographic medial tibiofemoral osteoarthritis (Kellgren-Lawrence score, 2 or 3) in at least 1 knee associated with radiographic skyline views demonstrating patellofemoral osteoarthritis (none to moderate severity); (2) neutral (−2° valgus to 2° varus), or varus (≤10° varus) mechanically aligned knee16; (3) body mass index (BMI, calculated as weight in kilograms divided by height in meters squared) ranging from 20 to 45; (4) and no participation in formal strength training more than 30 minutes a week in the past 6 months. All participants continued their medications, including nonsteroidal anti-inflammatory drugs (NSAIDs). If pain decreased, they could reduce them with their physician’s approval.
Exclusion criteria included (1) severe patellofemoral osteoarthritis (joint space width, 3 using the Osteoarthritis Research Society International atlas from skyline x-ray views17) (2) lateral more than medial tibiofemoral osteoarthritis; (3) or less than 20 on the Montreal Cognitive Assessment.18
Participants were randomly assigned to treatment groups via a computer-generated permuted block randomization scheme with randomly selected sized blocks of 3, 6, 9, or 12 stratified by sex and baseline BMI (calculated as weight in kilograms divided by height in meters squared). Participants were not aware of the hypotheses tested.
Quiz Ref IDStrength training sessions were conducted in groups of participants 3 times weekly over 18 months. Sessions consisted of a 5-minute warm-up, 40-minute training, and 15-minute cooldown conducted 3 times a week at the Wake Forest Clinical Research Center and supervised by staff interventionists. After the first 4 sessions during which participants were taught proper techniques, the maximum amount a participant was capable of lifting in a single repetition was measured (1 repetition maximum [1RM] tests) to determine their starting resistance. The program included 6 lower body exercises, with each leg exercised separately to prevent unequal load distribution between limbs: hip abduction and adduction; leg curl, leg extension, and leg press; and seated calf; and 4 upper body and core exercises: compound row, vertical chest, lower back, and abdomen.
The high-intensity group performed 3 sets of each exercise beginning at 75% of 1RM with 8 repetitions per set for 2 weeks, progressing to 80% of the 1RM with 8 repetitions per set for weeks 3 and 4, 85% of the 1RM with 6 repetitions per set for weeks 5 and 6, and 90% of the 1RM with 4 repetitions per set for weeks 7 and 8. Week 9 was a taper week with alternate exercises and establishing new 1RMs for each exercise. This 9-week block was repeated using the new 1RM values. The low-intensity group used the same 9-week block pattern but performed 3 sets of 15 repetitions at 30% to 40% 1RM of the exercises described above. The target workload per total volume performed during these 9-week cycles was the same regardless of whether the participant was assigned to the high-intensity or low-intensity group.19 To improve adherence and retention, interventionists were trained in standardized behavioral techniques developed in a social cognitive framework.20
Quiz Ref IDParticipants attended 60-minute group workshops biweekly for the first 6 months and monthly thereafter (total of 24 sessions over 18 months). Details of the control intervention are included in Supplement 1 and the design publication.19
The 2 primary outcomes at the 18-month (long-term) follow-up were self-reported knee pain using the Likert version of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)21 and maximum knee joint compressive force during walking (minimally clinically important difference [MCID] unknown).
Excessive knee joint compressive forces are an important mechanism in the osteoarthritis disease pathway and are associated with poor clinical and structural outcomes.22,23 The WOMAC knee pain score assesses knee pain over the last 48 hours.24 The total score ranges from 0 to 20 (higher scores indicate greater pain). The MCID in WOMAC pain between groups is 2 on a 20-point Likert scale.25 The pain categories on a 0 to 20 scale are 2 to 8, mild; more than 8 to 14, moderate; more than 14 to 20, severe (transformed from 0-10 scale reported by Kapstad et al26).
The 2 primary outcomes were also included as prespecified secondary outcomes at the 6-month (short-term) follow-up time point. Prespecified secondary outcomes measured at the 6-month and 18-month follow-up were WOMAC physical function, which assesses the degree of difficulty with activities of daily living in the last 48 hours with the total score range of 0 to 68; higher scores indicate poorer function with a score of 21 or higher indicating physical work limitations27 and the 6-minute walk distance, referred to as a clinical measure of mobility in the study protocol, assessed the maximum distance a participant could walk along a standardized walkway in 6 minutes (MCID, 26-55 m).28
Prespecified secondary outcomes included at 18 months only were disease progression assessed as change in medial tibiofemoral joint space width from bilateral posteroanterior weight-bearing fixed-flexion knee x-rays using a positioning device (SynaFlexer, BioClinica Inc) and modified Lyon-Schuss technique,29 and thigh skeletal muscle volume and thigh fat volume (sum of subcutaneous and intermuscular fat depots) determined using volumetric noncontrast enhanced computed tomography (CT) (VCT 64, GE Healthcare). IL-6 serum levels, knee extensor strength, and hip abductor strength were secondary outcomes for which testing time points were not specified in the protocol. Plasma IL-6 was measured with morning venipuncture samples (50 mL per visit) at least 2 hours after rising and after a 10-hour overnight fast. The IL-6 measure was performed in duplicate using an IL-6 enzyme-linked immunosorbent assay (Quantikine ELISA kits, R&D Systems). The mean value was used in data analysis. Knee extensor strength (termed quadriceps strength in the protocol) and hip abductor strength were measured using a Humac NORM isokinetic dynamometer at 30° per second (Computer Sports Medicine Inc). Prespecified secondary outcomes not included in this report were knee adductor moment; knee anteroposterior shear force; tumor necrosis factor, soluble tumor necrosis factor receptor 1, leptin, serum N-propeptide of collagen IIA, cartilage oligomeric matrix protein, and urinary type II collagen telopeptide levels; and muscle power.
Post hoc outcomes at 6 and 18 months were knee flexor strength and pain medication use (NSAIDS, analgesics, selective serotonin releasing agents, anticonvulsants, and opioids).30 A responder analysis used the Outcome Measures in Rheumatology Clinical Trials and the Osteoarthritis Research Society International (OMERACT-OARSI) criteria31 for minimally clinically important improvement of 20% or more for pain and function at 18 months.
Trial stratification factors, baseline BMI, and sex were included in all statistical models. A total sample of 372 provided 80% statistical power, assuming 80% retention to detect between-group differences of 1.1 or more (17.6%) in WOMAC pain (20 point scale) and 316 N or more (9.6%) in knee joint compressive force assuming a 2-sided test at a .05/6 = .0083 significance level (3 pairwise comparisons for 2 coprimary outcomes). These values were based on our previous trial32 and pilot study that showed mean differences between strength training and control groups of 1.12 (18%) for pain, and 657 N (20%) for knee joint compressive force. Power was determined based on absolute effects; the relative effects (%) were listed in the protocol for clarity.
Participants were analyzed according to their assigned randomization group even if they were nonadherent to their assigned intervention. Primary analyses were conducted using complete-case data. Prespecified sensitivity analyses were conducted for all participants using multiple imputation methods for missing observations based on baseline and, when available, follow-up data assuming missing data are missing at random.
The primary aim was to determine if high-intensity strength training lowered pain levels and reduced knee joint compressive forces compared with low-intensity strength training and attention control at the 18-month follow-up. These and all other repeated measure continuous outcomes were assessed for significant treatment effects using a repeated measures analysis of covariance (ANCOVA) model using complete case outcome data at 6 and 18 months, adjusted for sex, baseline BMI, and baseline outcome values. The analytic plan focused primarily on pairwise comparisons between treatment groups at 18 months. Pairwise group differences between the primary outcomes of pain and knee joint compressive forces were deemed significant using a Bonferroni-adjusted .0083 2-sided level of significance, accounting for 3 group pairwise comparisons across 2 outcomes.
An adjusted significance level of .05 was used for pairwise comparison of secondary and post hoc outcomes. Short-term effects were determined by using 6-month treatment effect estimates from the primary and secondary outcome models. Outcomes assessed at baseline and 18 months only (joint space width and thigh muscle and fat volume) were compared using an ANCOVA model for treatment effects adjusted for sex, baseline BMI, and baseline outcome values. The study power was calculated for the coprimary outcomes. Given the large number of secondary outcomes and lack of adjustment for multiple comparisons, the secondary outcomes should be interpreted as exploratory.
Knee flexion strength pairwise comparisons were performed using repeated measures ANCOVA at 6 and 18 months adjusted for sex, baseline BMI, and baseline outcome values. Longitudinal pain medication use was estimated for each group using a repeated-measures negative binomial regression model fit with treatment group, visit, and their interaction assuming an unstructured covariance, and visit-specific estimates were generated using contrast statements. Sensitivity analyses determined whether results were biased due to missing data via multiple imputation methods. Analyses were conducted on the 2 primary outcomes using the methods described above with additional adjustment for knee varus and valgus frontal plane alignment and patellofemoral osteoarthritis severity. Unadjusted pairwise comparisons used complete case changes from baseline at 6 or 18 months using analysis of variance at the same thresholds as the adjusted models. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc).
Table 1 shows the characteristics of randomized participants by group. Among 377 participants randomized (mean age, 65 years; 40% women), 320 (85%) attended the 18-month visit (high-intensity, 109; low-intensity, 108; control, 103) (Figure 1). At 18 months, 311 contributed data for the primary outcome of pain and 192 contributed data for the primary outcome of knee joint compressive force (Table 2). The difference in the number of completers for the 2 primary outcomes was due to missing baseline or follow-up biomechanical gait analyses or both. For the 6-month visit, 310 participants contributed data for the pain outcome, and 196 contributed data for the knee joint compressive force outcome. Noncompleters (no 18-month follow-up data) were comparable with completers, with the exception of age, which was a mean of 65.3 years (SD, 7.7 years) for the completers vs for 63.0 years (SD, 9.6 years) for the noncompleters (P = .04) (eTable 1 in Supplement 2).
During the first 6 months, patients in the high-intensity group had an adherence rate of 78% vs 77% in the low-intensity group. The overall adherence rates were 66% in the high-intensity group vs 69% in the low-intensity group. Adherence rates for the control group were 83% at 6 months and 80% overall.
Quiz Ref IDThere was no statistically significant difference between groups in the mean WOMAC pain at the 18-month (long-term) follow-up, 5.1 in the high-intensity group vs 4.9 in the control group (adjusted difference, 0.2; 95% CI, −0.6 to 1.1; P = .61) and 4.4 in the low-intensity group (adjusted difference, 0.7; 95% CI, −0.1 to 1.6; P = .08).Quiz Ref ID Mean maximum knee joint compressive force was not statistically significantly different between the high-intensity and the 2 other groups: high-intensity vs control, 2453 N vs 2512 N (adjusted difference, −58; 95% CI, −282 to 165 N; P = .61) and high-intensity vs low-intensity, 2453 N vs 2475 N (adjusted difference, −21 N; 95% CI, −235 to 193; P = .85) (Figure 2 and Table 2).
Similar results for the 2 primary outcomes were found without adjustment (Table 2), using multiple imputation, and in post hoc analysis adjusting for knee varus and valgus frontal plane alignment and patellofemoral osteoarthritis severity (eTables 2 and 3 in Supplement 2).
At the 6-month (short-term) follow-up, the low-intensity group had statistically significantly lower WOMAC knee pain (5.6 vs 4.4; adjusted difference, 1.2; 95% CI, 0.5 to 1.9; P = .001) and better WOMAC function (20.8 vs 16.1; adjusted difference, 4.8; 95% CI, 2.4 to 7.2; P < .001) than did the high-intensity group. The knee joint compressive forces were not statistically significantly different at 6 months among groups. The high-intensity group had a mean of 2453 N compared with 2483 N in the control group (adjusted difference, −30 N; 95% CI, −231 to 172; P = .77) and compared with 2536 N in the low-intensity group (adjusted difference, −83 N; 95% CI, −276 to 110; P = .40).
The mean 6-minute walk distance at 6 months was not statistically significantly different between the high-intensity group (493 m) and the control group (508 m; adjusted difference, −14 m; 95% CI, −32 to 3.7; P = .12). But the mean distance walked in the high intensity group was significantly less than in the low-intensity group (mean, 514; m in the low-intensity group (adjusted difference, −21 m; 95% CI, −38 to −3.6; P = .02) (eTable 4 in Supplement 2).
At the 18-month follow-up, comparisons of the mean WOMAC function scores were as follows: high-intensity vs control, 17.8 vs 16.4 (adjusted difference, 1.4; 95% CI, −1.3 to 4.1; P = .32); high-intensity vs low-intensity, 17.8 vs 14.9, adjusted difference, 2.9; 95% CI, 0.2 to 5.6; P = .03) (Table 3 and eFigure 1 in Supplement 2). Changes in the 6-minute walk distance were not statistically significantly different among the groups at 18 months. The mean distance for the high-intensity group was 508 m vs 515 m in the control group (adjusted difference, −6 m; 95% CI, −31 to 18; P = .62) and 514 m in the low-intensity group (adjusted difference, −5 m; 95% CI, −29 to 18; P = .67) (Table 3 and eFigure 1 in Supplement 2).
Knee extensor mean strength was statistically significantly greater in the high-intensity group (98.5 Newton meter [Nm]) than in the control group (88.5 Nm) at 18 months (adjusted difference, 10.0 Nm; 95% CI, 3.8 to 16.2; P = .002) but was not statistically significantly greater than in the low-intensity group (93.2 Nm; adjusted difference, 5.3 Nm; 95% CI, −0.8 to 11.4; P = .09). The mean hip abductor strength for the high-intensity group was 74.6 Nm vs 69.5 Nm in the control group (adjusted difference, 5.0 Nm; 95% CI, 0.7 to 9.3; P = .02) but was not significantly different from the low-intensity group (adjusted difference, 0.0 Nm; 95% CI, −4.3 to 4.2; P = .99) (Table 3 and eFigure 2 in Supplement 2).
The mean thigh muscle volume for the high-intensity group was 672 cm3 compared with 670 cm3 in the control group (adjusted difference, 2 cm3; 95% CI, −20 to 23; P = .89) and compared with 679 cm3 in the low-intensity group (adjusted difference, −8 cm3; 95% CI, −29 to 13; P = .48). The mean thigh fat volume in the high-intensity group of 712 cm3 compared with 709 cm3 in the control group (adjusted difference, 3 cm3; 95% CI, −37 to 42; P = .89) and compared with the 721 cm3 in the low-intensity group (adjusted difference, −10 cm3; 95% CI, −49 to 30, P = .63) was not statistically significantly different for either comparison at the 18-month follow-up (Table 3 and eFigure 3 in Supplement 2).
The mean log IL-6 levels were the same (0.9) in all 3 groups at 18 months. The adjusted difference between the high-intensity and the control group was 0.0 (95% CI, −0.1 to 0.2; P = .88) and between the high-intensity group vs low-intensity group was 0.0 (95% CI, −0.1 to 0.2; P = .58) (Table 3 and eFigure 3 in Supplement 2).
The mean joint space width at 18 months was not statistically significantly different between the high-intensity group (2.9 mm) and the control group (2.8 mm; adjusted difference, 0.2 mm; 95% CI, −0.0 to 0.4; P = .054) or between the high-intensity group and the low-intensity group (2.8 mm; adjusted difference, 0.1 mm; 95% CI, −0.0 to 0.3; P = .12) (Table 3 and eFigure 3 in Supplement 2).
Quiz Ref IDThere were 87 nonserious adverse events: 53 in the high-intensity, 30 in the low-intensity, and 4 in the control groups. Of those, 29 were related to the study: 20 in the high-intensity, 9 in the low-intensity, none in the control groups. Twenty patients experienced body pain: 12 in the high-intensity, 7 in the low-intensity, and 1 in the control groups; 19 experienced falls: 11 in the high-intensity, 6 in the low-intensity, and 2 in the control groups; and 10 experienced muscle strain: 8 in the high-intensity and 2 in the low-intensity groups, all which were the most frequent adverse events. Three knee replacements occurred, 1 in each group. The 13 serious adverse events (5, high-intensity; 3, low-intensity; 5, control) were determined unrelated to the study by the external safety monitor (eTables 6-8 in Supplement 2).
The mean knee flexor strength at 18 months was statistically significantly greater in both exercise groups than in the control group. The mean high-intensity group measured 51.5 Nm compared with 43.8 Nm in the control group (adjusted difference, 7.6 Nm; 95% CI, 3.6 to 11.7; P < .001). The low-intensity group measured 52.6 Nm, and compared with the control group had an adjusted difference of 8.7 Nm (95% CI, 4.7 to 12.7; P < .001). The mean knee flexor strength was not significantly different between the high-intensity group and the low-intensity group (adjusted difference, −1.1 Nm; 95% CI, −5.1 to 2.9; P = .59) (Table 3; eFigure 2 in Supplement 2).
The proportion of participants using pain medication declined across the 18-month intervention period, with no statistical difference among the groups at the 18-month follow-up: high-intensity, 0.45 (95% CI, 0.35 to 0.58); low-intensity, 0.34 (95% CI, 0.25 to 0.48); and control, 0.55 (95% CI, 0.44 to 0.68; P = .06) (eFigure 4 in Supplement 2). The number and percentage of participants in each group that achieved the OMERACT-OARSI clinical criteria31 of 20% or more improvement from baseline in pain and function at 18 months were 42 of 93 (45%) in the high-intensity; 49 of 88 (56%) in the low-intensity; and 56 of 94 (60%) in the control groups (χ2 P = .12).
Among participants with knee osteoarthritis, high-intensity strength training did not significantly reduce WOMAC knee pain or knee joint compressive forces at 18 months compared with low-intensity strength training or with an attention control group.
One potential explanation for why the outcomes in the high-intensity strength training and control groups were not significantly different relates to the improvement in pain in the control group. In this trial, knee osteoarthritis pain improved by 33% in the control group, compared with from 1% to 17% in prior trials.15,32-34 In these earlier trials, the control intervention consisted of limited healthful lifestyle educational material. In the current study, the control group attended group educational sessions with interventionists trained in social cognitive strategies to maximize adherence, and interacted with other participants in a social environment. The characteristics of the control intervention may have contributed to a larger control group effect than reported previously.
Alternatively, Englund35 suggested the true effect of exercise interventions for knee osteoarthritis is modest, with improvements in pain and function due primarily to the placebo effect, natural history of the disease, and regression to the mean. The large sample size and long duration of the intervention may also have increased the placebo response for the subjective outcome of pain.36 These, together with the substantial reduction in pain in the control group, may explain why the high-intensity strength training group was not significantly different from the control group.31,36
The study premise was that long-term high-intensity strength training would affect the biomechanical pathway via reduced knee joint compressive forces resulting in reduced knee pain and attenuated osteoarthritis disease progression.15,19 However, significant increases in muscle strength were not associated with improvement in knee joint loads.
The decrease in medial joint space width across groups (mean, 0.2 mm; 4%-5% per year) was consistent with the natural history of osteoarthritis progression.37 These data suggest that high-intensity strength training did not exacerbate radiographic disease progression.
Nonserious adverse events were more prevalent in the high-intensity (n = 53) and low-intensity (n = 30) strength training groups than in the control group (n = 4) but did not appear to affect retention adversely (high-intensity and low-intensity, 86% vs control, 83%). Furthermore, the 95% CI around the pairwise difference in WOMAC pain between the high-intensity group and the low-intensity and control groups included worse scores for the high-intensity group of between 1 and 2 points that are of uncertain clinical importance.23 Hence, clinically important levels of increased pain with high-intensity training cannot be ruled out.
This study has several limitations. First, the results may be more generalizable to individuals who are comparable with the study sample, the majority of whom were men, white, obese, and had more than a high school education. Second, the range of eligible frontal plane knee angles was −2° valgus to 10° varus; hence, these results are most generalizable to alignments within this range. Third, musculoskeletal modeling was used to estimate knee joint compressive forces. This method is limited by many simplifying assumptions about joint properties and structures.38
Among participants with knee osteoarthritis, high-intensity strength training compared with low-intensity strength training or an attention control did not significantly reduce WOMAC knee pain or knee joint compressive forces at 18 months. The findings do not support the use of high-intensity strength training over low-intensity strength training or an attention control in adults with knee osteoarthritis.
Corresponding Author: Stephen P. Messier, PhD, J.B. Snow Biomechanics Laboratory, Department of Health & Exercise Science, Worrell Professional Center, Wake Forest University, 1834 Wake Forest Rd, Winston-Salem, NC 27109 (email@example.com).
Accepted for Publication: January 13, 2021.
Author Contributions: Drs Messier and Beavers had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Messier, Mihalko, Beavers, Nicklas, DeVita, Hunter, Bennell, Loeser.
Acquisition, analysis, or interpretation of data: Messier, Mihalko, Beavers, Carr, Hunter, Lyles, Guermazi, Bennell, Loeser.
Drafting of the manuscript: Messier, Mihalko, Beavers, Nicklas, DeVita, Bennell.
Critical revision of the manuscript for important intellectual content: Messier, Mihalko, Nicklas, Carr, Hunter, Lyles, Guermazi, Bennell, Loeser.
Statistical analysis: Beavers.
Obtained funding: Messier, Mihalko.
Administrative, technical, or material support: Messier, Beavers, Nicklas, DeVita, Carr, Hunter, Lyles, Guermazi, Loeser.
Supervision: Messier, Nicklas, Carr, Lyles.
Conflict of Interest Disclosures: Dr Messier reported receiving grants from the National Institute of Arthritis Musculoskeletal and Skin Diseases. Dr Mihalko reported receiving grants from the National Institutes on Health (NIH). Dr Beavers reported receiving grants from the NIH. Dr Nicklas reported grants from Wake Forest School of Medicine. Dr Hunter reported receiving personal fees from Pfizer, Lilly, Merck Serono, and TLCBio. Dr Guermazi reported receiving personal fees from MerckSerono, Pfizer, TissueGene, Galapagos, Roche, AstraZeneca, Boston Imaging Core Lab (BICL) LLC and is the president and founder of BICL. Dr Bennell reported receiving grants from the NIH, the National Health and Medical Research Council, and Medibank Private and personal fees from UpToDate. Dr Loeser reported receiving grants from the National Institute of Arthritis and Musculoskeletal and Skin Diseases and Bioventus and personal fees from Unity Biotechnology. No other disclosures were reported.
Funding/Support: This study was supported by grants 1R01AR059105-01 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases and P30 AG21332 from the National Institute on Aging.
Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Data Sharing Statement: See Supplement 3.
Additional Contributions: We thank the START research staff and the START participants for their important contributions. The participants did not receive compensation for their contributions, and the staff were not compensated besides their salaries.