Is the Severity of Knee Osteoarthritis on Magnetic Resonance Imaging Associated With Outcome of Exercise Therapy?

Authors


Abstract

Objective

To evaluate associations between severity of knee osteoarthritis (OA) on magnetic resonance imaging (MRI) and treatment outcomes in knee OA patients treated with exercise therapy in an exploratory study.

Methods

Ninety-five participants with knee OA in a 12-week exercise program had obtained 3.0T MRI scans of the knee joint prior to treatment. MRI data were systematically assessed for OA severity of multiple features (cartilage integrity, bone marrow lesions, osteophyte formation, effusion/synovitis, and meniscal abnormalities) according to the Boston Leeds Osteoarthritis Knee Score method. Regression analyses were performed to analyze associations between OA severity on MRI (for the tibiofemoral and patellofemoral [PF] compartments) and outcome of exercise therapy, i.e., changes in activity limitations (Western Ontario and McMaster Universities Osteoarthritis Index physical function; primary outcome), pain and upper leg muscle strength, and treatment response (Outcome Measures in Rheumatology/Osteoarthritis Research Society International criteria).

Results

Improvements of 24%, 34%, and 21% on average in activity limitations, pain, and muscle strength, respectively, after 12-week exercise therapy were found (P < 0.001). Severity of abnormalities in PF cartilage integrity was significantly associated with fewer improvements in both activity limitations (P = 0.01) and muscle strength (P = 0.04). Severity of PF osteophyte formation was significantly associated with fewer improvements in muscle strength (P < 0.01). All other features on MRI were not associated with treatment outcome.

Conclusion

Effectiveness of exercise therapy seems to be independent of OA severity on MRI, except for abnormalities in cartilage integrity and osteophyte formation, both in the PF compartment. Our study suggests that all grades of OA severity on MRI can benefit from professionally supervised exercise therapy, although the effects might be reduced in patients with advanced PF OA.

INTRODUCTION

Exercise therapy is considered an effective treatment in patients with knee osteoarthritis (OA) for reducing pain and activity limitations ([1, 2]). However, these effects are only small to moderate and vary widely among patients ([2]). Treatment effects might be optimized through identifying subgroups of patients who may not benefit optimally from exercise therapy, for instance, patients with severe knee joint damage, in whom exercise therapy has been suggested to be too painful or even potentially harmful ([1, 3, 4]).

Recently, we conducted a randomized controlled trial ([5]) on the effectiveness of a newly developed, supervised exercise program in 159 knee OA patients (ranging from mild to severe radiographic OA), in which large and clinically relevant improvements in pain and activity limitations, and treatment response in 71% of participants, were found. In a random subsample (n = 95), we additionally obtained baseline magnetic resonance imaging (MRI) scans, enabling us to determine whether the effectiveness of exercise therapy depends on OA severity on MRI. Therefore, the aim of this exploratory study was to evaluate associations between severity of knee OA on MRI and treatment outcome in knee OA patients treated with exercise therapy.

Box 1. Significance & Innovations

  • This exploratory study is the first to evaluate the role of osteoarthritis (OA) severity on magnetic resonance imaging (MRI) in outcomes of exercise therapy in patients with knee OA.
  • Effectiveness of exercise therapy seems to be independent of OA severity on MRI, except for abnormalities in cartilage integrity and osteophyte formation, both in the patellofemoral (PF) compartment.
  • This study suggests that all grades of OA severity on MRI can benefit from professionally supervised exercise therapy, although in patients with advanced PF OA, exercises may need to be adapted or additional interventions might be necessary to optimize treatment outcome.

PATIENTS AND METHODS

Design

We previously conducted a single-blind, randomized controlled trial (STABILITY trial; Dutch Trial Register NTR1475) ([5]) in 159 knee OA patients with knee instability in an outpatient rehabilitation center (Reade, Centre for Rehabilitation and Rheumatology, Amsterdam, The Netherlands). In this trial, 2 exercise programs were compared to evaluate the effectiveness of specific knee joint stabilization training. A research assistant who performed measurements was blinded for group allocation. For the present study, we additionally obtained MRI scans at baseline (i.e., prior to exercise therapy) in 95 consecutive participants. This study was approved by the Medical Ethical Review Board (Reade/Slotervaart Hospital).

Participants

Inclusion criteria for the STABILITY trial were 1) diagnosis of knee OA according to the clinical American College of Rheumatology criteria ([6]), 2) age between 40 and 75 years, and 3) presence of self-reported knee instability (i.e., occurrence of an episode of knee instability in the past 3 months as reported by the patient [5]) and/or biomechanically assessed knee instability (using cutoff points for upper leg muscle weakness, proprioceptive accuracy, and varus–valgus laxity of the knee joint [5]). Exclusion criteria for the trial are described in our previous publication ([5]); for the present study, contraindications for MRI (e.g., pacemaker, claustrophobia) were added. All participants provided written informed consent.

Intervention

Both the experimental and control interventions comprised a 12-week exercise program with sessions of 60 minutes twice weekly in groups of approximately 8 participants, supervised by 2 physical therapists specifically trained to supervise only one of both treatments, and a home exercise program.

A detailed description of the exercise protocol is included in our previous publication ([5]). In summary, the experimental program focused on knee joint stabilization, muscle strengthening, and performance of daily activities, while the control program focused only on muscle strengthening and performance of daily activities. Training intensity and the amount of attention from the physical therapists were similar in both groups.

Measurements

Outcome measures

Detailed information on measurements is included in our previous publication ([5]). For the present study, we used measurements from baseline (T0) and 12-week followup (T12).

Primary outcome

Self-reported activity limitations were assessed by the Dutch translation of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function subscale consisting of 17 items, with a total score ranging from 0–68, where 0 = no limitations and 68 = maximally limited.

Secondary outcomes

Knee pain severity was assessed on a numerical rating scale (NRS; range 0–10, where 0 = no pain and 10 = worst imaginable pain) by the question “What was your average knee pain during the last week?” Upper leg muscle strength was assessed for knee flexion and extension using an isokinetic dynamometer (EnKnee, Enraf-Nonius) in Nm divided by body weight. The Outcome Measures in Rheumatology (OMERACT)/Osteoarthritis Research Society International (OARSI) set of responder criteria was used as a dichotomous measure for treatment response, operationalized as improvement of ≥50% and absolute change of ≥20 points in pain (NRS) or function (WOMAC), or at least 2 of the 3 following items: 1) improvement of ≥20% and absolute change of ≥10 points in pain, 2) improvement of ≥20% and absolute change of ≥10 points in function, and 3) improvement of ≥20% and absolute change of ≥10% in patient's global assessment (PGA) ([7]). For these criteria, NRS pain and WOMAC physical function were rescaled into a 100-point scale. A global perceived effect score ≥5 (i.e., “symptoms slightly improved” or better) ([5]) was regarded as an improvement in PGA.

MRI

Detailed information on MRI assessment is included in our previous publication ([8]). In summary, 3.0T MRI scans (General Electric Medical Systems, 5 sequences) of the most affected knee (index knee) were obtained at baseline (i.e., 1–3 weeks prior to the start of the intervention). MRIs were assessed for cartilage integrity, bone marrow lesions (BMLs), osteophyte formation, effusion, synovitis, and meniscal abnormalities according to the Boston Leeds Osteoarthritis Knee Score (BLOKS) scoring system ([9]) by a radiologist (JPK) with 27 years of musculoskeletal expertise, blinded to the patients' clinical characteristics and radiographic assessment.

Potential baseline confounders

Potential baseline confounders that were included are sex, age and duration of knee symptoms (in years), use of pain medication, presence of obesity (i.e., body mass index ≥30 kg/m2), presence of knee malalignment (i.e., varus–valgus malalignment ≥10°, measured by a goniometer), upper leg muscle strength (mean of knee extension and flexion isokinetic strength, in Nm divided by body weight), proprioceptive accuracy and varus–valgus knee joint laxity (in degrees), and presence of symptoms of anxiety and/or depression (measured by the Hospital Anxiety and Depression Scale) ([5]).

Statistical analyses

Because both interventions were found to be similarly effective ([5]), data from both exercise groups were combined in the present study. Descriptive statistics (mean ± SD or percentage) for baseline patient characteristics and outcome measures (i.e., change between T0 and T12 on WOMAC physical function, NRS pain, and upper leg muscle strength, where positive scores indicate improvement) were calculated. For severity of MRI features, region-specific grades (range 0–3) were combined in tibiofemoral and patellofemoral (PF) compartmental grades (for cartilage integrity, BMLs, and osteophyte formation) or in knee-specific grades (for meniscal abnormalities) using the highest regional grade approach. Subsequently, grades were dichotomized as grade 0/1 (none to mild) versus grade 2/3 (moderate to severe). Effusion grade (range 0–3) and synovitis (presence/absence) were combined in a single feature for inflammation and dichotomized as effusion grade 0/1 in the absence of synovitis versus effusion grade 2/3 and/or the presence of synovitis.

Linear and logistic regression analyses were performed with change scores on the primary outcome WOMAC physical function, NRS pain, and upper leg muscle strength (all adjusted for baseline score) or with treatment response (yes/no) as dependent variables. Independent variables were each of the (dichotomized) MRI features. Crude models and adjusted models, in which relevant baseline confounders were added, were estimated. Standardized (β) and unstandardized (B) regression coefficients and odds ratios for linear and logistic regression analyses, respectively, with adjunctive 95% confidence intervals (95% CIs) and P values were estimated. Statistical significance was accepted at P values less than 0.05. Data analyses were performed using PASW Statistics 18.0 (SPSS).

RESULTS

From 112 consecutive, potential participants in the STABILITY trial from January 2010, baseline MRIs could be obtained in 105 persons, of which 10 could not be included in the trial (reasons include not fulfilling the inclusion criteria [n = 6] and withdrawal [n = 4]), while 7 participants could not be planned for MRI prior to treatment. Therefore, 95 patients were included. Because 4 persons dropped out during the intervention (reasons included lack of time [n = 1] and health conditions unrelated to OA or to the intervention [n = 3]), regression analyses could be performed in 91 participants.

As shown in Table 1, baseline characteristics of the total sample from the STABILTY trial (n = 159) and present subsample (n = 95) are comparable. Our sample represents a severe knee OA group on average, with the highest grade of severity on MRI found in 54%, 31%, and 58% of the participants for cartilage integrity, BMLs, and meniscal abnormalities, respectively. Participants were followed for a mean ± SD of 20.8 ± 4.1 of 24 treatment sessions. Improvements of 24%, 34%, and 21% were found in activity limitations (WOMAC physical function), pain (NRS), and upper leg muscle strength, respectively, after 12-week exercise therapy (P < 0.001 for all), with two-thirds (67%) of the participants fulfilling the OMERACT/OARSI criteria for treatment response.

Table 1. Patient characteristics of the total study sample and the subsample from the present study*
 Total sample (n = 159)Subsample (n = 95)
ValueKneeaTFaPFa
  1. Values are the number (percentage) unless indicated otherwise. TF = tibiofemoral; PF = patellofemoral; BMI = body mass index; K/L = Kellgren/Lawrence; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; NRS = numerical rating scale; MRI = magnetic resonance imaging; N/A = not applicable; BMLs = bone marrow lesions; T0 = baseline; T12 = 12-week followup.
  2. aHighest regional grade per knee or per compartment.
  3. b0 = none, 1 = <10% of surface area, 2 = 10–75% of surface area, and 3 = >75% of surface area.
  4. c0 = none, 1 = mild, 2 = moderate, and 3 = severe.
  5. d0 = none, 1 = <10% of bone volume, 2 = 10–25% of bone volume, and 3 = >25% of bone volume.
  6. e0 = physiologic amount (in suprapatellar bursa only), 1 = small (fluid continuous in retropatellar space), 2 = medium (with slight convexity of suprapatellar bursa), and 3 = large (evidence of capsular distention)
  7. f0 = absent and 1= present.
  8. g0 = none, 1 = signal only, 2 = tear (horizontal, vertical, or complex), and 3 = maceration.
  9. hA positive score indicates improvement.
  10. iAccording to the Outcome Measures in Rheumatology/Osteoarthritis Research Society International set of responder criteria ([6]).
Baseline demographics     
Female sex97 (61)64 (67)   
Age, mean ± SD years61.9 ± 7.161.5 ± 7.0   
Duration of knee symptoms, mean ± SD years10.6 ± 9.311.2 ± 9.5   
BMI, mean ± SD kg/m229.0 ± 4.629.2 ± 4.8   
Radiographic severity     
K/L grade 0/156 (36)27 (28)   
K/L grade 244 (28)26 (27)   
K/L grade 341 (26)27 (28)   
K/L grade 418 (11)15 (16)   
WOMAC physical function (range 0–68), mean ± SD26.2 ± 12.227.2 ± 12.3   
NRS pain (range 0–10), mean ± SD5.0 ± 2.15.1 ± 2.1   
Upper leg muscle strength, mean ± SD Nm/kg0.84 ± 0.390.75 ± 0.36   
Baseline MRI feature     
Cartilage integritybN/A    
Grade 0  7 (7)12 (13)31 (33)
Grade 1  8 (8)12 (13)33 (35)
Grade 2  29 (31)28 (30)18 (19)
Grade 3  51 (54)43 (45)13 (14)
BMLscN/A    
Grade 0  16 (17)28 (30)66 (70)
Grade 1  24 (25)21 (22)17 (18)
Grade 2  26 (27)20 (21)9 (10)
Grade 3  29 (31)26 (27)3 (3)
Osteophyte formationdN/A    
Grade 0  14 (15)18 (19)18 (19)
Grade 1  36 (38)36 (38)44 (46)
Grade 2  31 (33)31 (33)27 (28)
Grade 3  14 (15)10 (11)6 (6)
EffusioneN/A  N/AN/A
Grade 0  32 (34)  
Grade 1  29 (31)  
Grade 2  22 (23)  
Grade 3  12 (13)  
SynovitisfN/A  N/AN/A
Grade 0  63 (66)  
Grade 1  32 (34)  
Meniscal abnormalitiesgN/A  N/AN/A
Grade 0  7 (7)  
Grade 1  12 (13)  
Grade 2  21 (22)  
Grade 3  55 (58)  
Outcome of exercise therapy     
WOMAC physical function (range 0–68)     
ΔT0 to T12, mean ± SDh7.6 ± 10.26.4 ± 9.5   
T0 to T12, % changeh2924   
NRS pain (range 0–10)     
ΔT0 to T12, mean ± SDh1.9 ± 2.11.7 ± 2.2   
T0 to T12, % changeh3934   
Upper leg muscle strength, Nm/kg     
ΔT0 to T12, mean ± SDh0.16 ± 0.210.16 ± 0.19   
T0 to T12, % changeh1921   
Treatment responsei113 (71)64 (67)   

Results from the regression analyses of OA severity on MRI and outcome of exercise therapy (n = 91) are shown in Table 2. Severity of abnormalities in cartilage integrity in the PF compartment (i.e., grade 2/3) was significantly associated with fewer improvements in both the primary outcome WOMAC physical function and upper leg muscle strength (adjusted B = −5.0 [95% CI −8.8, 1.2] and −0.08 [95% CI −0.16, −0.01], respectively). Severity of PF osteophyte formation (i.e., grade 2/3) was significantly associated with fewer improvements in upper leg muscle strength (adjusted B = −0.12 [95% CI −0.20, −0.04]). No associations were found for OA severity of any other MRI features in crude or adjusted models.

Table 2. Results of regression analyses of outcome of exercise therapy and severity of knee osteoarthritis on multiple MRI features (n = 91)*
MRI featureChange in WOMAC physical function (range 0–68), β (P)aChange in NRS pain (range 0–10), β (P)aChange in upper leg muscle strength (Nm/kg), β (P)aTreatment response, OR (P)b
  1. MRI = magnetic resonance imaging; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; NRS = numerical rating scale; TF = tibiofemoral; PF = patellofemoral; BMLs = bone marrow lesions.
  2. aAdjusted for baseline value of outcome measure and relevant confounding (i.e., change in the B coefficient of the independent variable of >10% after addition of the following potential confounders: age, sex, duration of knee symptoms, use of pain medication, obesity, malalignment, upper leg muscle strength, proprioceptive accuracy, varus–valgus laxity, and symptoms of anxiety/depression). Negative β values indicate less improvement and positive β values indicate more improvement (compared to the reference group).
  3. bAdjusted for relevant confounding (i.e., change in the B coefficient of the independent variable of >10% after addition of the following potential confounders: age, sex, duration of knee symptoms, use of pain medication, obesity, malalignment, upper leg muscle strength, proprioceptive accuracy, varus–valgus laxity, and symptoms of anxiety/depression). Odds ratios (ORs) <1 indicate lower odds for treatment response and ORs >1 indicate higher odds for treatment response (compared to the reference group).
  4. cGrade 2/3 versus 0/1 (reference group).
  5. dSignificant at P < 0.05.
  6. eEffusion grade 2/3 and/or the presence of synovitis versus effusion grade 0/1 and the absence of synovitis (reference group).
Cartilage integrityc    
TF−0.01 (0.92)0.04 (0.69)−0.03 (0.77)2.0 (0.16)
PF−0.25 (0.01)d−0.13 (0.16)−0.21 (0.04)d0.5 (0.11)
BMLsc    
TF−0.10 (0.28)−0.02 (0.86)0.07 (0.50)1.1 (0.81)
PF−0.08 (0.45)−0.11 (0.24)−0.12 (0.24)1.1 (0.89)
Osteophyte formationc    
TF0.02 (0.88)−0.03 (0.75)−0.11 (0.27)1.2 (0.70)
PF0.02 (0.83)−0.08 (0.43)−0.30 (0.002)d0.9 (0.92)
Inflammation, kneee−0.07 (0.45)−0.10 (0.28)0.01 (0.94)0.6 (0.27)
Meniscal abnormalities, kneec0.06 (0.51)0.13 (0.16)0.06 (0.54)1.5 (0.50)

DISCUSSION

This exploratory study is the first to evaluate the role of OA severity on MRI in effectiveness of exercise therapy in patients with knee OA. We demonstrated that outcome of exercise therapy was not affected by severity of OA on MRI in any feature (including BMLs and inflammation), except for abnormalities in cartilage integrity and osteophyte formation, both in the PF compartment. This suggests that all grades of OA severity on MRI can benefit from exercise therapy, although effects might be reduced in patients with advanced PF OA.

Severity of abnormalities in PF cartilage integrity (i.e., moderate to severe loss of cartilage thickness) was significantly associated with poor outcome of exercise therapy, i.e., fewer improvements in WOMAC physical function (P = 0.01) and upper leg muscle strength (P = 0.04). Furthermore, a trend toward less improvement in NRS pain (P = 0.16) and lower odds for treatment response (P = 0.11) was found as well (Table 2). Moderate to severe PF osteophyte formation was significantly associated with fewer improvements in upper leg muscle strength (P < 0.01). These findings are in line with a recent study ([10]) in which PF pain predicted nonresponse to exercise therapy plus manual physiotherapy. As shown in Figure 1, patients with the highest grade of severity for abnormalities in PF cartilage integrity specifically do not seem to benefit optimally from exercise therapy. This trend was similar for PF osteophyte formation. Possibly, exercising is too painful for the subtype of patients with advanced PF OA. These patients may not be able to perform exercises at a level that can be expected to be effective, concordant with our clinical experience. As stated by Bennell et al ([11]), patients with PF joint symptoms may need to exercise “in positions that minimize PF contact forces and knee loading (for example, in lesser degrees of knee flexion and in non-weightbearing positions).” Through these adaptations, as well as by cointerventions aimed to reduce PF pain (e.g., PF taping), effectiveness of exercise therapy might be optimized in this specific subgroup.

Figure 1.

Outcome of exercise therapy (changes between baseline [T0] and 12-week followup [T12], where a positive score indicates improvement and a negative score indicates worsening) on the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function subscale (A) and upper leg muscle strength (B) for each grade of severity of abnormalities in cartilage integrity in the patellofemoral compartment (n = 91). Bars show the mean scores and lines show the 95% confidence intervals.

Except for PF defects in cartilage integrity and osteophyte formation, effectiveness of exercise therapy was found to be independent of OA severity of any other feature in each compartment on MRI. This indicates that referral to exercise therapy, supervised by professionally trained physical therapists, can be considered in patients with all grades of OA severity on MRI. This result is in line with a recent study ([12]) in which exercise therapy was shown to be feasible and safe in patients with end-stage knee or hip OA waiting for total joint replacement. Our study also supports an expert opinion–based recommendation ([13]) proposing that exercise effects do not depend on (radiographic) OA severity with evidence. Conversely, our findings contrast with existing beliefs from some physicians and patients that exercise may be too painful and possibly harmful, especially for patients with severe knee OA ([1, 3, 4]). A longitudinal study using MRIs before and after exercise therapy may unravel the controversy on the impact of exercise therapy on knee joint structures, which may be protective (through improved shock absorption [14], improved cartilage quality [15], and reduced inflammation [16]) or destructive (through persistent overloading [4]).

Major strengths of this study are the use of 3.0T MRIs, which were systematically assessed according to the BLOKS scoring system ([8]), a low dropout rate (4%), and high patient adherence. Study limitations include the absence of a “no exercise” control group, the large number of analyzed associations (n = 32), and the lack of contrast used for assessing effusion/synovitis.

To conclude, effectiveness of exercise therapy seems to be independent of OA severity on MRI, except for severity of abnormalities in cartilage integrity and osteophyte formation, both in the PF compartment. Our study suggests that all grades of OA severity on MRI can benefit from exercise therapy, which was shown to be highly effective, although effects might be reduced in patients with advanced PF OA.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Mr. Knoop had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Knoop, Dekker, van der Leeden, Klein, Steultjens, Lems.

Acquisition of data. Knoop, Klein, Roorda, Lems.

Analysis and interpretation of data. Knoop, Dekker, van der Leeden, van der Esch, Hunter, Roorda, Steultjens.

Acknowledgments

The authors would like to thank Drs. Gerritsen and Voorneman for examining patients, Dr. Reiding for assessing all radiographs, S. Romviel for performing measurements, T. Schweigmann for assisting with MRI data collection, and all physical therapists supervising the exercise therapy.

Ancillary